Unlocking Silence: Advances in Brain-Computer Interfaces for Communication in Locked-In Syndrome

Abstract

This article synthesizes current research and development in Brain-Computer Interfaces (BCIs) for restoring communication in individuals with Locked-In Syndrome (LIS). Targeting researchers, scientists, and drug development professionals, it provides a comprehensive overview spanning foundational neuropathology, methodological approaches in invasive and non-invasive BCIs, troubleshooting of technical and clinical implementation hurdles, and validation through recent pioneering case studies. The review highlights the transition from experimental systems to potential clinical tools, emphasizing the role of AI, hybrid systems, and personalized neuroprostheses in achieving naturalistic, real-time communication for a profoundly isolated patient population.

Understanding Locked-In Syndrome: From Pathophysiology to Communication Desperation

Clinical Definition and Classification of Locked-In Syndrome

Locked-In Syndrome (LIS) is a neurological condition characterized by severe whole-body paralysis with preserved consciousness and cognitive function. The condition results from specific damage to the brainstem, particularly the ventral pons, disrupting motor pathways while sparing cognitive and sensory processing areas. Based on the extent of motor preservation, the medical community classifies LIS into three distinct clinical presentations [1].

Classical LIS is characterized by paralysis of all four limbs (quadriplegia) and bilateral facial paralysis, accompanied by loss of voice and speech (aphonia). Crucially, individuals retain the ability to produce vertical eye movements and blinks, which often serve as their primary means of communication [1].

Incomplete LIS describes a condition where patients retain remnants of voluntary movements beyond eye control. These residual movements may include slight finger motions, subtle head movements, or other minimal motor functions that provide additional communication channels beyond eye-based systems [1].

Complete Locked-In State (CLIS) represents the most severe form, characterized by total whole-body paralysis including all eye muscles. This results in complete immobility and an inability to communicate through any muscular activity, creating profound isolation for affected individuals [1] [2].

Table 1: Clinical Classification of Locked-In Syndrome

Classification	Motor Function	Communication Capacity	Common Etiologies
Classical LIS	Quadriplegia, bilateral facial paralysis, aphonia with preserved vertical eye movement and blinking	Yes/No communication via eye blinks; can use eye-tracking devices	Brainstem stroke, traumatic brain injury, amyotrophic lateral sclerosis (ALS)
Incomplete LIS	Remnants of voluntary movements beyond eye control (e.g., slight finger, head, or facial movements)	Multiple potential communication channels depending on residual movement	Brainstem stroke, traumatic brain injury, ALS in progression
Complete LIS (CLIS)	Total paralysis of all voluntary muscles, including ocular motility	No reliable motor-based communication possible	Progressive neurological conditions (e.g., end-stage ALS), brainstem lesions

The progression through these states can vary significantly depending on the underlying etiology. Individuals with brainstem stroke or trauma may maintain a static LIS state or even experience some recovery, while those with neurodegenerative diseases like ALS typically progress through these stages gradually, often ultimately transitioning to complete LIS [1].

BCI Communication Technologies Across the LIS Spectrum

Brain-Computer Interfaces (BCIs) represent a revolutionary technological approach for restoring communication to individuals with LIS by establishing a direct pathway between the brain and external devices. These systems translate measured brain activity into commands for communication software, bypassing the impaired neuromuscular system [3] [1].

Neural Signal Acquisition Modalities

BCI systems employ various signal acquisition methods, each with distinct advantages and limitations for LIS applications:

Non-invasive modalities include scalp electroencephalography (EEG), functional near-infrared spectroscopy (fNIRS), magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI). EEG and fNIRS are particularly relevant for clinical applications due to their portability and relatively low cost [1]. EEG records electrical neural signals at high temporal resolution but suffers from low spatial resolution and susceptibility to artifacts. fNIRS measures hemodynamic responses associated with neural activity, offering better spatial resolution than EEG but slower response times [1] [2].

Invasive modalities include electrocorticography (ECoG), stereotactic EEG (sEEG), and intracortical microelectrode arrays. These approaches offer superior signal quality and spatial resolution but require surgical implantation with associated risks [4] [3]. Intracortical interfaces, such as the Utah array used by companies like Blackrock Neurotech and Paradromics, can record from individual neurons or local field potentials, providing high-fidelity signals for complex control [4] [3].

BCI Paradigms for Communication

Multiple BCI paradigms have been investigated for communication applications in LIS:

Evoked Potential Approaches include the P300 speller, which exploits cortical responses to infrequent stimuli, and steady-state visual evoked potentials (SSVEP), which utilize responses to flickering visual stimuli. These approaches generally offer higher reliability and require less user training than other paradigms [1].

Motor Imagery and Attempted Movement paradigms decode sensorimotor rhythms (mu and beta bands) modulated by imagined or attempted movements. These approaches leverage the brain's natural motor circuitry but require significant user training and calibration [1].

Imagined Speech Decoding represents a cutting-edge approach that aims to decode internal speech processes directly. Recent advances have demonstrated real-time decoding of imagined syllables and words from both invasive and non-invasive recordings [5].

Table 2: BCI Signal Acquisition Modalities for LIS Communication

Modality	Invasiveness	Spatial Resolution	Temporal Resolution	Key Applications in LIS	Limitations
EEG	Non-invasive	Low (widespread signals)	High (milliseconds)	P300 speller, SSVEP, motor imagery spelling	Low signal-to-noise ratio, susceptibility to artifacts
fNIRS	Non-invasive	Moderate (cortical surface)	Low (seconds)	Binary communication in CLIS via frontocentral oxygenation	Slow hemodynamic response, limited depth penetration
ECoG/sEEG	Minimally invasive (subdural)	High (local cortical patches)	High (milliseconds)	High-accuracy spelling, imagined speech decoding	Requires surgical implantation, limited coverage
Intracortical Microelectrodes	Fully invasive (penetrating)	Very high (single neurons)	Very high (milliseconds)	High-speed typing, complex device control	Highest surgical risk, potential signal instability over time

Experimental Protocols and Methodologies

Protocol for fNIRS-Based Communication in CLIS

Functional near-infrared spectroscopy (fNIRS) has emerged as a promising modality for establishing communication in complete LIS, where other approaches often fail. The methodology described by Chaudhary et al. (2017) involves an implicit attentional processing procedure [2]:

Equipment Setup: The system employs fNIRS optodes positioned over frontocentral brain regions to measure oxygenation changes associated with cognitive activity. The system typically uses two or more wavelengths (e.g., 760 nm and 850 nm) to distinguish between oxy- and deoxy-hemoglobin concentrations [2].

Task Structure: Patients are presented with personal questions with known answers and open questions requiring "yes" or "no" responses. The protocol uses auditory presentation of questions to accommodate patients without visual control [2].

Signal Processing: The fNIRS signals are preprocessed to remove motion artifacts and physiological noise (cardiac and respiratory cycles). Features are extracted from the hemodynamic response, typically focusing on oxygenated hemoglobin concentration changes in the 2-8 second window post-stimulus [2].

Classification: A linear support vector machine (SVM) classifier is trained to distinguish between "yes" and "no" responses based on the frontocentral oxygenation patterns. The system provides real-time feedback to facilitate learning [2].

Validation: Performance is assessed through cross-validation and calculation of information transfer rate (ITR). Correct response rates above 70% with statistical significance indicate successful communication [2].

Protocol for Intracortical LFP-Based Long-Term Communication

Milekovic et al. (2018) demonstrated stable long-term BCI communication using local field potentials (LFPs) in individuals with ALS and LIS [4]:

Surgical Implementation: The approach uses intracortical microelectrode arrays (such as the Utah array) implanted in motor cortical areas. The implant records LFPs, which are more stable signals than single-neuron spiking activity [4].

Signal Processing: LFPs are acquired in multiple frequency bands (theta, alpha, beta, gamma). The system extracts power spectral density features and uses discriminant analysis for classification [4].

Decoder Calibration: The initial decoder is calibrated using attempted arm or hand movements. Crucially, the LFP-based decoder maintains stable performance over extended periods (76-138 days in the study) without requiring recalibration [4].

Application Interface: Participants control a computer cursor to select characters from a virtual keyboard. The system achieves spelling rates of 3.07-6.88 correct characters per minute, enabling practical communication for daily use [4].

Performance Validation: Long-term stability is assessed through daily spelling tasks with calculation of accuracy and information transfer rates. System performance remains consistent without significant degradation over time [4].

Protocol for EEG-Based Imagined Speech BCI

A recent study investigated the learning processes involved in operating an imagined speech BCI using EEG [5]:

Participant Training: Fifteen healthy participants trained for five consecutive days to control a binary BCI system using imagery of two syllables (/fɔ/ and /gi/) with contrasting phonetic features [5].

Experimental Design: Each daily session included mental chronometry tests (measuring overt and covert speech timing) followed by BCI-control sessions with real-time feedback [5].

EEG Acquisition: Neural data were recorded using a 64-channel EEG system at 512 Hz sampling rate. EMG signals from facial muscles were simultaneously recorded to monitor for potential subvocalization [5].

Real-Time Decoding: The BCI system provided continuous visual feedback based on the decoded syllable imagery. A control group received discontinuous feedback to assess the importance of feedback quality [5].

Performance Analysis: Learning effects were assessed through daily improvement in classification accuracy. Neural correlates of learning were identified through changes in theta and low-gamma power across training sessions [5].

Diagram 1: BCI Communication Workflow for LIS

Table 3: Key Research Reagents and Solutions for LIS BCI Research

Resource Category	Specific Examples	Function/Application	Key Characteristics
Neural Signal Acquisition Systems	ANT Neuro EEG systems, Blackrock Neurotech Utah array, Synchron Stentrode, Paradromics Connexus	Record neural signals for BCI control	Varying invasiveness, signal quality, and spatial coverage; selection depends on research goals and participant profile
Signal Processing Platforms	OpenBCI, BCI2000, MATLAB with EEGLAB/FieldTrip	Preprocess, filter, and extract features from neural signals	Enable noise reduction, artifact removal, and feature extraction critical for decoding accuracy
Decoding Algorithms	Linear Discriminant Analysis (LDA), Support Vector Machines (SVM), Deep Neural Networks (DNNs)	Translate neural features into control commands	Balance computational efficiency with decoding accuracy; choice depends on signal modality and application
Stimulus Presentation Software	Psychtoolbox, Presentation, Unity-based VR environments	Present visual, auditory, or tactile stimuli for evoked potentials	Precise timing control, flexibility in experimental design, integration with BCI systems
Communication Interfaces	P300 spellers, virtual keyboards, text-to-speech synthesizers	Translate BCI commands into functional communication	User-friendly design, customizable layouts, compatibility with BCI output
Validation Metrics	Information Transfer Rate (ITR), Classification Accuracy, F-score	Quantify BCI system performance and communication efficacy	Standardized measures for comparing systems across studies and laboratories

Current Research Landscape and Future Directions

The field of BCI research for LIS communication is rapidly evolving, with multiple companies and research institutions advancing different technological approaches [3].

Commercial BCI Development: As of 2025, several companies are conducting clinical trials of implantable BCI systems. Neuralink is testing ultra-high-bandwidth implantable chips, while Synchron is pursuing a less invasive endovascular approach with its Stentrode device [3]. Blackrock Neurotech, with years of experience supplying neural arrays for research, is developing new electrode technology called Neuralace [3]. Paradromics and Precision Neuroscience are also advancing high-channel-count implants and minimally invasive cortical interfaces respectively [3].

Integration with Complementary Technologies: Future directions include combining BCIs with functional electrical stimulation (FES) to restore not just communication but also expressive movements [1]. Hybrid BCI systems that combine multiple signal modalities (e.g., EEG with fNIRS) may improve reliability. The integration of virtual and mixed reality with BCIs offers enriched feedback environments that may enhance user engagement and performance [6] [7].

Market Growth and Translation: The global BCI market is projected to grow significantly, from an estimated $2.41 billion in 2025 to $12.11 billion by 2035, representing a compound annual growth rate of 15.8% [8]. This growth reflects increasing investment and technological advancement, positioning BCIs to potentially follow a trajectory similar to other groundbreaking technologies like gene therapy [3].

The continuing evolution of BCI technology offers hope for abolishing the communicative isolation of locked-in syndrome, particularly for those in the complete locked-in state who have historically been without reliable communication options. As technologies mature and clinical trials progress, the translation of these systems from laboratory demonstrations to practical clinical tools represents a crucial frontier in neurorehabilitation and assistive technology.

Locked-in syndrome (LIS) presents a profound paradox in clinical neurology: the preservation of conscious awareness amidst nearly total bodily paralysis. This condition arises from specific lesions in the ventral pons that disrupt corticospinal and corticobulbar pathways while sparing consciousness networks [9]. Understanding the neuroanatomical basis of this dissociation is critical for developing targeted interventions, including brain-computer interface (BCI) technologies that aim to restore communication for affected individuals [10] [11].

The ventral pons serves as a critical bridge connecting cerebral cortical areas with cerebellar coordination centers and spinal cord motor neurons [12]. Damage to this specific brainstem region produces the distinctive LIS clinical profile, making it an essential focus for both pathological understanding and therapeutic development in consciousness research.

Neuroanatomy of the Ventral Pons

Structural Organization

The pons, derived from the embryonic metencephalon, constitutes the largest segment of the brainstem, measuring approximately 27 mm in height with transverse and anteroposterior widths of 38 mm and 25 mm respectively [13] [12]. It is conveniently subdivided into two functionally distinct regions: the ventral basilar pons and the dorsal pontine tegmentum, separated by the trapezoid body [13].

Table 1: Structural Components of the Pons

Component	Location	Major Structures	Primary Functions
Ventral Pons	Anterior portion	Pontine nuclei, corticospinal tracts, corticopontine tracts, transverse pontine fibers	Motor coordination, relay between cortex and cerebellum
Dorsal Tegmentum	Posterior portion	Reticular formation, cranial nerve nuclei, ascending sensory tracts	Arousal, sensory integration, cranial nerve functions
Trapezoid Body	Boundary between ventral and dorsal pons	Decussating auditory fibers	Separation of motor and sensory regions

The ventral pons contains several crucial structures including pontine nuclei, vertically traversing corticopontine and corticospinal tracts, and transversely crossing fibers that project to the contralateral cerebellum via the middle cerebellar peduncle [13]. This region appears bulky precisely because of the extensive pontine nuclei and their connections to the corticopontocerebellar pathway [13].

Vascular Supply and Vulnerability

The ventral pons receives its blood supply primarily from the vertebrobasilar system, with most of the pons supplied by pontine arteries branching from the basilar artery [12]. The basilar artery courses through the basilar sulcus along the midline of the anterior pons [12]. This vascular arrangement creates a specific vulnerability; occlusion or hemorrhage of the paramedian branches of the basilar artery typically leads to bilateral ventral pontine infarction, the most common cause of locked-in syndrome [9] [14].

Pathophysiology of Locked-in Syndrome

Neural Circuit Disruption

Locked-in syndrome results from precisely localized damage that disrupts specific neural pathways while sparing others. The primary pathology involves interruption of corticospinal tracts (responsible for voluntary motor control), corticobulbar tracts (controlling face, head, and neck muscles), and corticopontine tracts [9] [14]. This disruption occurs at the level of the ventral pons, which contains these descending fiber bundles as they travel from the cerebral cortex to the spinal cord and cerebellum.

Consciousness preservation in LIS is explained by the sparing of the reticular activating system (RAS) located in the dorsal brainstem tegmentum [9]. The RAS remains intact because it lies dorsal to the typical lesion site in the ventral pons, allowing maintained arousal, wakefulness, and cognitive function despite profound motor paralysis.

Figure 1: Neural Pathway Disruption in Locked-in Syndrome. The diagram contrasts intact motor pathways with the disrupted state in LIS, where ventral pons damage interrupts corticospinal tracts while sparing consciousness networks.

Etiological Spectrum

Multiple pathological processes can produce the ventral pontine damage characteristic of locked-in syndrome:

Vascular causes represent the most common etiology, accounting for approximately 86% of cases in one large survey [9]. These include basilar artery thrombosis leading to pontine infarction or hemorrhage. Patients often have comorbid hypertension, atherosclerotic disease, or atrial fibrillation [9].

Non-vascular causes encompass traumatic brain injury (producing pontine contusion or vertebrobasilar dissection), masses (pontine glioma, metastasis), infections (pontine abscess, meningitis), and demyelinating disorders (central pontine myelinolysis, multiple sclerosis, Guillain-Barré syndrome) [9] [15].

Clinical Presentation and Diagnostic Assessment

Diagnostic Criteria and Classification

Locked-in syndrome manifests through a distinctive clinical profile characterized by three primary forms:

Table 2: Clinical Variants of Locked-in Syndrome

Form	Motor Function	Eye Movement	Communication Capacity	Sensation
Classical	Total immobility	Vertical eye movements and blinking preserved	Yes (via eye coding)	Typically lost
Incomplete	Some preserved motor function beyond eyes	Vertical eye movements and blinking preserved	Yes (via eye coding and possibly other movements)	May be partially preserved
Total Immobility	Complete body paralysis	No eye movement	Only via EEG/BCI	Completely lost

The classical form presents with total immobility except for preserved vertical eye movement and blinking, with intact cognition and consciousness [9] [14]. The incomplete form shares features with the classical form but demonstrates additional preserved motor functions. The total immobility form involves complete paralysis including ocular motility, requiring electroencephalography (EEG) to detect conscious awareness [9].

Experimental Protocols for Assessment

Precise diagnostic protocols are essential to distinguish LIS from other disorders of consciousness such as coma or vegetative state. The following methodologies represent standardized approaches for clinical assessment and research:

Behavioral Assessment Protocol:

• Consciousness Evaluation: Administer Coma Recovery Scale-Revised (CRS-R) repeatedly to detect subtle signs of awareness [9]
• Motor Response Mapping: Systematically test for any residual voluntary movement, beginning with eye tracking and progressing to other muscle groups
• Communication Testing: Establish reliable yes/no communication codes using vertical eye movements or blinking (e.g., upward gaze for "yes," downward for "no")
• Sensory Examination: Map residual sensation to light touch, pain, and temperature throughout the body
• Cranial Nerve Assessment: Specifically evaluate facial sensation, corneal reflexes, and gag reflex

Neurophysiological Testing Protocol:

• Electroencephalography (EEG): Record during resting state and during command-following tasks to detect covert consciousness [14]
• Evoked Potentials: Assess brainstem auditory evoked responses (BAER) and somatosensory evoked potentials (SSEP) to evaluate pathway integrity [14]
• Sleep Architecture Analysis: Document preserved sleep-wake cycles on polysomnography [15]

Neuroimaging Protocol:

• Structural MRI: Obtain high-resolution T1 and T2-weighted images to precisely localize pontine lesions [14]
• Diffusion Tensor Imaging (DTI): Reconstruct corticospinal and corticobulbar tracts to visualize disruption patterns
• Functional MRI (fMRI): Assess task-based responses to motor imagery and language comprehension commands

Figure 2: Comprehensive Assessment Protocol for Locked-in Syndrome. The workflow outlines multimodal evaluation strategies to correctly diagnose LIS and distinguish it from other disorders of consciousness.

Brain-Computer Interfaces for Communication Restoration

BCI Modalities and Applications

Brain-computer interfaces represent a promising technological approach to restore communication for LIS patients by bypassing damaged motor pathways. These systems detect conscious brain signals and translate them into commands for external devices [10].

Electroencephalography (EEG)-based BCIs utilize non-invasive electrodes to measure electrical brain activity. Common paradigms include:

• P300 Speller: Detection of event-related potentials during character selection
• Motor Imagery: Interpretation of sensorimotor rhythms during imagined movements
• Steady-State Visual Evoked Potentials (SSVEP): Response to flickering visual stimuli

Invasive BCIs involve implanted electrodes that provide higher signal resolution:

• Electrocorticography (ECoG): Grid electrodes placed on the cortical surface
• Intracortical microelectrodes: Penetrating electrodes that record individual neurons

Functional Near-Infrared Spectroscopy (fNIRS) measures hemodynamic responses associated with neural activity, providing an alternative modality particularly suitable for patients with limited eye movement [15].

BCI Experimental Protocol

Implementing BCI communication systems for LIS patients requires systematic methodology:

Patient Preparation and Screening:

• Conduct comprehensive assessment to establish cognitive capacity and sensory capabilities
• Determine optimal BCI modality based on residual function and practical constraints
• Obtain informed consent through established eye-coding communication channels

System Calibration:

• Position sensors according to international standards (10-20 system for EEG)
• Record baseline brain activity during resting state and various cognitive tasks
• Identify reliable neural markers for communication (e.g., P300 responses, sensorimotor rhythms)

Training Protocol:

• Implement graded training tasks beginning with simple binary choices
• Progress to more complex communication paradigms as proficiency develops
• Incorporate individualized vocabulary based on patient needs and preferences

Validation and Optimization:

• Measure communication accuracy across multiple sessions
• Adjust parameters to maximize information transfer rate
• Implement adaptive algorithms to accommodate neural plasticity

Research Reagent Solutions

Table 3: Essential Research Tools for LIS and BCI Investigations

Category	Specific Tools/Reagents	Research Function	Application in LIS
Neuroimaging	3T MRI Scanner, DTI sequences, fMRI tasks	Structural and functional lesion mapping	Localize ventral pontine damage, assess preserved networks
Neurophysiology	High-density EEG systems, Evoked potential amplifiers	Detect covert consciousness, monitor brain function	Establish communication channels, diagnose LIS
BCI Hardware	Active EEG electrodes, fNIRS systems, ECoG grids	Acquire brain signals for communication	Restore expressive abilities, enable environmental control
BCI Software	OpenVibe, BCI2000, custom signal processing algorithms	Process neural data, implement communication paradigms	Translate brain activity into commands, optimize information transfer
Assessment Tools	CRS-R, WHIM, LIS-specific QoL scales	Measure clinical status, functional outcomes	Track recovery, evaluate intervention efficacy

Therapeutic Approaches and Future Directions

Current Management Strategies

Supportive care forms the cornerstone of LIS management, focusing on preventing complications and maintaining quality of life. Essential components include:

Respiratory Support: Most LIS patients require tracheotomy and mechanical ventilation due to impaired voluntary breathing control [14]. Regular pulmonary hygiene and ventilator management are crucial.

Nutritional Support: Gastrostomy tube placement ensures adequate nutrition and hydration while preventing aspiration [14].

Preventive Care: Comprehensive programs to avoid complications of immobility including pressure ulcers, contractures, deep vein thrombosis, and urinary tract infections [14].

Communication Rehabilitation: Intensive training to optimize use of preserved eye movements and implementation of appropriate assistive technologies [14].

Emerging Research and Neurotechnology

Recent advances in neuroscience offer promising directions for LIS research and treatment:

Deep Brain Stimulation (DBS): Emerging evidence suggests that DBS targeting thalamic nuclei such as the centromedian-parafascicular complex may modulate consciousness networks [16]. Electric field modeling indicates optimal stimulation sites in the inferior parafascicular nucleus and adjacent ventral tegmental tract [16].

Advanced BCIs: Next-generation interfaces are focusing on flexible neural implants, closed-loop neurostimulation, and artificial intelligence integration to improve the speed and reliability of communication [10].

Personalized Therapeutic Platforms: Development of customized digital prescription systems that deliver tailored therapeutic strategies based on individual patient characteristics and residual capacities [10].

The ongoing BRAIN Initiative aims to accelerate these technological developments through interdisciplinary collaboration, with specific focus on mapping neural circuits, developing novel recording technologies, and advancing human neuroscience [17].

The ventral pons serves as a critical anatomical locus whose integrity is essential for translating conscious intention into voluntary action. Lesions in this specific region produce the distinctive clinical profile of locked-in syndrome, characterized by preserved awareness amidst profound paralysis. Understanding the precise neuroanatomical correlates of this condition provides not only diagnostic clarity but also foundational knowledge for developing targeted interventions.

Future research directions should prioritize the refinement of brain-computer interface technologies, exploration of neuromodulation approaches, and development of comprehensive care models that address both the physiological and psychosocial needs of individuals living with locked-in syndrome. The integration of advanced neuroimaging, neurotechnology, and personalized rehabilitation holds promise for significantly improving quality of life and restoring communicative autonomy for this unique population.

Locked-in syndrome (LIS) presents one of the most challenging conditions in neurological care, characterized by complete paralysis of voluntary muscles while cognitive function remains preserved. The primary etiologies leading to LIS include stroke, traumatic brain injury (TBI), and progressive neurological diseases such as amyotrophic lateral sclerosis (ALS). Research into brain-computer interface (BCI) communication systems offers promising avenues for restoring interaction and improving quality of life for LIS patients. Understanding the epidemiological patterns and underlying causes of these precursor conditions is fundamental for identifying at-risk populations, guiding therapeutic development, and contextualizing BCI research within the broader landscape of neurological disease burden. This whitepaper synthesizes the most current global epidemiological data on stroke, ALS, and TBI, with particular emphasis on their relationship to LIS and the experimental methodologies driving innovation in communication neuroprosthetics.

Epidemiological Landscape of Key Etiologies

Global Burden of Stroke

Stroke remains a leading cause of mortality and long-term disability worldwide, representing a predominant cause of LIS. According to the most recent Global Burden of Disease (GBD) 2021 data, stroke is the second leading cause of death and the third leading cause of death and disability combined globally [18].

Table 1: Global Stroke Epidemiology (2021)

Metric	Value	Details
Annual Global Incidence	11.9 million	151 per 100,000 [18]
Global Prevalence	93.8 million	People living after stroke [18]
Annual Global Deaths	Approximately 7 million	Second leading cause of death globally [18]
Global DALYs	Over 160 million	Disability-Adjusted Life Years [18]
Economic Burden	> US $890 billion	0.66% of global GDP [18]
Stroke Types	Ischemic (65.3%), ICH (28.8%), SAH (5.8%)	ICH=Intracerebral Hemorrhage; SAH=Subarachnoid Hemorrhage [18]
Demographics	53% male, 47% female; 53% occur in people <70 years	[18]

Key risk factors for stroke include metabolic risks (attributable to 68.8% of all strokes), environmental risks (36.7%), and behavioral risks (35.2%) [18]. The GBD 2021 study identified high systolic blood pressure as a predominant risk factor, alongside air pollution, high body mass index, and high fasting plasma glucose. Notably, the number of stroke incidents has risen substantially between 1990 and 2021, with a 70.0% increase in incident strokes, 44.0% increase in deaths from stroke, and 86.0% increase in prevalent strokes [18].

Amyotrophic Lateral Sclerosis (ALS) Epidemiology

ALS is a progressive neurodegenerative disease that leads to loss of voluntary motor control and, in its advanced stages, can result in a completely locked-in state (CLIS). A 2025 population-based study in Catalonia, Spain, provides updated epidemiological insights [19].

Table 2: ALS Epidemiology from a Spanish Cohort (2015-2020)

Metric	Value
Incidence	2.39 per 100,000 person-years
Prevalence	7.98 cases per 100,000 persons
Median Age at Diagnosis	68 years
Gender Distribution	50.4% female
Common Comorbidities	Dementia (6.8% before diagnosis), Depression/Anxiety (45.7%)
Median Survival	2.19 years from diagnosis
Mortality Predictors	Older age, alcohol abuse, history of stroke, dementia

Globally, ALS is considered a rare disease, with annual incidence ranging from 1.7 to 2.2 per 100,000 people [20]. The disease is categorized primarily into sporadic (90-95% of cases) and familial (5-10%) forms. Key genetic mutations associated with ALS include C9orf72, SOD1, FUS, and TARDBP, with ongoing research investigating additional genetic risk factors such as ARPP21 and NEK1 [20].

Traumatic Brain Injury (TBI) Burden

TBI represents a significant cause of acquired brain injury that can lead to LIS, particularly when injuries affect the brainstem. According to 2021 GBD data, TBI continues to impose a substantial global burden [21] [22].

Table 3: Global Traumatic Brain Injury Epidemiology (2021)

Metric	Value
Annual Global Incidence	20.8 million new cases
Age-Standardized Incidence Rate	259 per 100,000 population
Severity Distribution	56.63% moderate/severe TBI
Trend (1990-2021)	Declining age-standardized incidence (EAPC: -0.11%)
Gender Distribution	Higher incidence in males across all age groups
Leading Cause	Falls (across most age groups and regions)
Regional Variation	Highest rates in Central & Eastern Europe; lowest in Sub-Saharan Africa

The study found significant regional variations, with the highest age-standardized incidence rates observed in Eastern Europe (522 per 100,000), Australasia (479), and Central Europe (479) [21]. The economic burden of TBI is substantial, with acute phase costs averaging approximately $22,000 per case and accumulating long-term costs comparable to those of cardiovascular diseases, cancer, and diabetes [21].

Experimental Protocols in BCI Research for LIS

Hybrid Eye-Tracking and BCI Protocol for Assistive Communication

A 2025 pilot study by Pinto et al. developed a protocol for a hybrid BCI framework combining eye-tracking and EEG to support patients transitioning from LIS to CLIS [23].

Objective: To create a seamless communication system that maintains functionality as oculomotor control declines in progressive conditions like ALS.

Participants: Five healthy participants (as a preliminary proof-of-concept).

Apparatus and Setup:

• Eye-Tracking System: Commercial eye-tracker to monitor gaze patterns and detect residual oculomotor function.
• EEG Acquisition: Multi-channel EEG system with electrodes placed according to the international 10-20 system, focusing on areas relevant to the P300 potential (e.g., Pz, Cz, Fz).
• Stimulus Presentation: A visual speller paradigm (e.g., a matrix of letters or symbols) was displayed on a screen. Rows and columns flashed in a random sequence to elicit P300 responses.
• Real-Time Data Processing Unit: A computer running integrated software for simultaneous ET and EEG data acquisition, processing, and fusion.

Procedure:

• Calibration: Individual calibration of the eye-tracker and the P300 speller for each participant.
• Task: Participants were instructed to focus attention on a target character in the speller matrix.
• Data Acquisition: Simultaneous recording of (a) gaze data from the eye-tracker and (b) continuous EEG data.
• Data Fusion and Intention Detection:
  • The eye-tracker data provided preliminary, continuous information about the user's likely target.
  • The EEG data was analyzed in real-time to detect the P300 event-related potential (ERP) time-locked to the flashing of the target character.
  • A fusion algorithm integrated the probabilities from both modalities to make a final, robust determination of user intention.

Outcome Measures: System accuracy (%) and information transfer rate (bits per minute) for character selection.

Protocol for Streaming Brain-to-Voice Neuroprosthesis

A groundbreaking study funded by the NIH, published in 2025, detailed a protocol for a BCI that restores natural speech by translating brain activity into audible words with minimal delay [24].

Objective: To develop a BCI that enables near-synchronous, fluent speech synthesis for individuals with paralysis-induced loss of speech.

Participant: A 47-year-old woman with paralysis and anarthria (inability to speak) for 18 years following a stroke.

Apparatus and Setup:

• Electrode Implantation: A high-density electrode array was surgically implanted over the cortical area responsible for speech encoding.
• Neural Signal Acquisition: High-fidelity system for recording neural signals from the implanted array.
• Stimulus Presentation: A monitor for displaying text prompts.
• Computing System: A powerful computer running a custom deep learning model for real-time decoding and speech synthesis.

Procedure:

• Training Data Collection:
  • The participant was presented with over 12,000 sentences (comprising >1,000 unique words) on a screen.
  • She was instructed to silently attempt to say each sentence.
  • Neural activity was recorded during these 23,000+ silent attempt trials to create a comprehensive dataset linking brain patterns to specific speech sounds and words.
• Model Training: A deep learning system was trained on the collected data to learn the mapping between the participant's neural activity and the intended phonetic and linguistic elements.
• Speech Synthesis: The participant's own voice, retrieved from a pre-injury recording, was used to create a personalized vocal synthesizer.
• Real-Time Streaming Decoding:
  • The trained system operated on a streaming basis, processing neural data in increments of 80 milliseconds.
  • As the participant silently attempted to speak, the system decoded the neural patterns and synthesized the corresponding speech in near real-time.

Outcome Measures: Decoding speed (words per minute), vocabulary size, accuracy (success rate), and latency (delay between brain activity and speech output).

Protocol for Assessing Consciousness in LIS Using EEG

For effective BCI-based communication, it is crucial to determine the optimal time to interact with a patient, especially those in CLIS. A 2025 study presented an EEG-based protocol for assessing consciousness levels in LIS patients [25].

Objective: To estimate normalized consciousness levels (NCL) in LIS patients to determine states conducive to communication.

Participants: Four LIS patients with EEG data recorded over several years.

Apparatus:

• EEG System: Multi-channel EEG recording system.
• Processing Software: Software for advanced signal processing and feature extraction (e.g., MATLAB, Python with MNE, etc.).

Procedure:

• Data Acquisition: Resting-state EEG was recorded from the patients.
• Pre-processing: Standard EEG pre-processing was applied, including filtering, artifact removal (e.g., ocular, muscular), and bad channel interpolation.
• Feature Extraction: Multiple features were extracted from the pre-processed EEG signals to capture different aspects of brain activity associated with conscious states:
  • Frequency Measures: Power in standard frequency bands (delta, theta, alpha, beta, gamma).
  • Complexity Measures: Lempel-Ziv Complexity (LZC) to quantify the complexity of the neural signal.
  • Connectivity Measures: Functional connectivity between different brain regions (e.g., using coherence, phase-locking value).
• Consciousness Level Estimation: The extracted features were integrated into a model to compute a Normalized Consciousness Level (NCL) score, ranging from 0 (not conscious) to 1 (fully conscious). This model was designed to maximize the probability of correctly determining the patient's state in the absence of ground truth.
• Validation: The derived NCL scores were compared with the patients' subsequent performance on communication tasks (when possible) to evaluate the method's validity.

Outcome Measures: NCL score and its correlation with BCI communication performance accuracy.

Table 4: Essential Research Materials for BCI and LIS Research

Item	Function / Application
High-Density ECoG/EEG Arrays	Record electrical activity from the cortex with high spatial and temporal resolution. Critical for decoding speech and cognitive states [24] [25].
Eye-Tracking Systems	Monitor oculomotor function and serve as a primary or hybrid communication channel in LIS before CLIS transition [23].
P300 & SSVEP Stimulation Software	Present visual/auditory stimuli to elicit event-related potentials used for BCI control and consciousness assessment [23] [25].
Deep Learning Models (e.g., RNNs, CNNs)	Decode complex neural signals into intended commands or speech in real-time. Core to modern streaming neuroprostheses [24].
Signal Processing Toolboxes (e.g., EEGLAB, MNE-Python)	Pre-process neural data (filtering, artifact removal) and extract features for analysis and model training [25].
Validated pTDP-43 Antibodies	Investigate TDP-43 proteinopathy, a pathological hallmark in ~97% of ALS cases, for biomarker and mechanistic studies [20].
Antisense Oligonucleotides (e.g., Tofersen)	Investigate as a therapeutic strategy for ALS patients with specific genetic mutations (e.g., SOD1) [20].
Seed Amplification Assays (SAAs)	Detect protein aggregates (e.g., TDP-43) for early diagnosis and monitoring of ALS [20].

Visualizing Workflows and Pathways

Workflow for Hybrid ET-BCI Communication System

The following diagram illustrates the integrated workflow for a hybrid eye-tracking and BCI system, designed to support continuous communication for patients as they transition into locked-in states.

Consciousness Assessment Pathway for LIS/CLIS

This diagram outlines the analytical pathway for assessing consciousness levels in non-communicative patients using resting-state EEG data, which is vital for determining the optimal time to initiate BCI communication.

Stroke, ALS, and TBI represent significant neurological challenges that can result in locked-in syndrome, creating an urgent need for advanced communication solutions. The epidemiological data underscores the substantial and growing global burden of these conditions. Current BCI research is rising to this challenge, with experimental protocols evolving from simple spellers to sophisticated, streaming brain-to-voice neuroprostheses that approach natural communication speeds. The integration of multi-modal data, such as eye-tracking with EEG, and the development of methods to assess consciousness levels, are critical for creating effective interventions tailored to the individual patient's clinical state. Continued research into the epidemiology, pathophysiology, and neurotechnology for these conditions is paramount for developing targeted prevention strategies and effective restorative solutions to improve the lives of those affected by LIS.

Locked-in Syndrome (LIS) represents one of the most profound neurological conditions, characterized by complete paralysis of nearly all voluntary muscles while cognitive function and consciousness remain fully intact. For these individuals, the loss of communication is not merely an inconvenience but a fundamental barrier to human interaction, medical care, and quality of life. Within this context, Brain-Computer Interface (BCI) technology has emerged as a transformative solution, directly linking neural activity to external communication devices. This whitepaper examines the critical imperative for restoring communication capabilities in LIS patients, exploring the significant impacts on both quality of life and clinical care, with a specific focus on the technical advances driving this revolutionary field forward. The restoration of communication is not just a technical challenge but a fundamental human right and clinical necessity for this population.

Technical Approaches to BCI Communication

BCI systems for communication operate by detecting and interpreting specific neural signals associated with user intent. These systems can be broadly categorized into non-invasive approaches that use external sensors and invasive approaches that require surgical implantation of recording devices.

Signal Modalities and Processing Pathways

Different BCI paradigms exploit distinct neurophysiological signals and require specialized processing approaches. The table below summarizes the primary signal modalities used in communication BCIs.

Table 1: Primary Neural Signal Modalities for BCI Communication

Signal Modality	Origin/Type	Recording Method	Typical Applications	Key Characteristics
P300 Event-Related Potential	Endogenous cognitive response to rare stimuli	Non-invasive (EEG) or invasive (ECoG)	Matrix spellers, RSVP spellers	Positive deflection ~300ms post-stimulus; requires attention to "oddball" stimuli [26]
Steady-State Visual Evoked Potential (SSVEP)	Neural oscillations entrained to visual stimulation frequency	Primarily non-invasive (EEG)	Frequency-coded spellers, control interfaces	Periodic response to visual flicker; requires gaze control; high information transfer rates [27]
Local Field Potentials (LFPs)	Population-level synaptic activity	Invasive (intracortical microelectrodes)	Communication spellers, cursor control	Summed activity of neuronal populations; more stable than single-unit activity over time [28]
Action Potentials (Spikes)	Firing of individual neurons	Invasive (intracortical microelectrodes)	High-performance typing, cursor control	High-frequency signals (~300-5000 Hz); rich information content but prone to instability [28]
Slow Cortical Potentials (SCPs)	Low-frequency shifts in cortical polarization	Non-invasive (EEG)	Early spelling devices (e.g., Thought Translation Device)	Requires extensive training; very slow communication rates (~0.5-2 characters/minute) [25]

Hybrid BCI Systems for Enhanced Reliability

Recent research has focused on hybrid BCI systems that combine multiple modalities to create more robust communication channels. A promising approach integrates eye-tracking with P300-based BCIs to support gradual transition between modalities as patients progress from LIS to complete LIS (CLIS), where even eye movement control is lost [23]. This fusion approach processes gaze and EEG data in real-time, using sophisticated algorithms to enhance the detection of user intention when oculomotor function begins to decline. The hybrid framework represents a crucial advancement for maintaining communication continuity throughout disease progression in conditions like ALS [23].

Quantitative Performance of Communication BCIs

The performance of communication BCIs has improved dramatically in recent years, with some systems now approaching practically useful speeds for daily communication.

Performance Metrics Across BCI Paradigms

Table 2: Performance Comparison of Modern Communication BCIs

BCI Type	Patient Population	Performance Metric	Reported Values	Key Advances
Intracortical BCI (BrainGate)	ALS, tetraplegia	Typing rate (copy typing)	3.07-6.88 correct characters/min [28]	LFP-based signals enabled stable performance for 76-138 days without recalibration [28]
Intracortical BCI (ReFIT Kalman Filter + HMM)	ALS, spinal cord injury	Typing rate (free typing)	24.4 ± 3.3 correct characters/min (ALS participant) [29]	Point-and-click control with optimized keyboard layout (OPTI-II); highest reported performance [29]
Speech Neuroprosthesis	Stroke-induced anarthria	Word decoding rate	90.9 words/min (50-word vocabulary); 47.5 words/min (1,000+ word vocabulary) [24]	Deep learning system translating brain activity to audible speech in <80ms; uses patient's own voice [24]
P300 Matrix Speller	ALS, LIS	Communication accuracy	~80% at 7.8 characters/min [26]	Multiple presentation sequences with bootstrapping and trial averaging [26]
Hybrid ET-BCI Framework	Healthy controls (LIS model)	Classification accuracy	Maintained high accuracy during modality transition [23]	Real-time fusion of gaze and EEG data; enables gradual transition as motor function declines [23]

Experimental Protocols and Methodologies

P300 Speller Protocol with bigP3BCI Dataset

The bigP3BCI dataset provides a standardized framework for P300-based BCI research, featuring data from both able-bodied individuals and those with ALS tested under various conditions [30]. The experimental protocol follows these key stages:

• Setup and Calibration: EEG signals are collected non-invasively at 256 Hz using passive gel-based or active dry electrodes connected to biosignal amplifiers. Electrode impedance checks ensure low impedance prior to recording. For hybrid systems, eye gaze position is simultaneously recorded using infrared eye trackers calibrated for each participant [30].

• Copy-Spelling Task: Participants perform copy-spelling of predefined tokens using a P300 speller application. The user interface typically presents a grid of characters (commonly 6×6 or 9×8). The system intensifies subsets of characters in pseudorandom order while the user focuses attention on the target character [30].

• Signal Processing Pipeline: EEG data time-locked to each stimulus event is processed using the following workflow: bandpass filtering → feature extraction → P300 detection using classifiers like Linear Discriminant Analysis → character decoding based on detected P300 responses [30] [26].

• Validation Approach: Experiments typically include both calibration phases (without BCI feedback to collect labeled training data) and test phases (with BCI feedback to evaluate algorithm performance) [30].

Figure 1: P300 BCI Experimental Workflow

Intracortical BCI Protocol for Long-Term Stable Communication

The groundbreaking work on intracortical BCIs has established protocols for long-term stable communication in LIS:

• Surgical Implementation: Participants receive 96-channel intracortical microelectrode arrays implanted in the hand area of the dominant precentral gyrus under FDA Investigational Device Exemption and institutional review board approval [28].

• Signal Selection for Stability: Unlike earlier approaches relying on sorted spikes, stable long-term communication utilizes local field potentials (LFPs), which demonstrate greater stability over time compared to neuronal action potentials. This enables decoder use for extended periods (76-138 days) without recalibration or performance degradation [28].

• Spelling Interface: Participants use applications like FlashSpeller, where options are presented for 1.5-2 seconds with brief intervals between presentations. The system identifies discrete selection events based on neuronal activity to select characters and functions (backspace, space, word completion, text-to-speech, email) [28].

• Performance Assessment: Both "free typing" (responding to conversational prompts) and "copy typing" (typing predetermined sentences) paradigms are used to evaluate performance in realistic and standardized conditions, respectively [29].

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Research Tools for BCI Communication Development

Tool/Category	Specific Examples	Function/Application	Key Features
Data Acquisition Platforms	BCI2000 [30]	Open-source software platform for BCI research	Supported by NIH; enables synchronized EEG and eye tracker data collection [30]
Signal Acquisition Hardware	g.tec biosignal amplifiers [30]	EEG signal recording and preliminary filtering	Compatible with gel-based and dry electrodes; integrated impedance checking [30]
Eye Tracking Systems	Tobii Pro X2-30 [30]	Gaze position tracking for hybrid BCI systems	Infrared eye tracking; synchronizable with EEG via BCI2000 EyeTrackerLogger [30]
Public Datasets	bigP3BCI Dataset [30]	Algorithm development and validation	Standardized EDF+ format; includes EEG, eye tracker, and clinical data; diverse participant population [30]
Invasive Recording Arrays	Blackrock Microsystems arrays [28]	Intracortical signal recording for invasive BCIs	96-channel microelectrode arrays; record both spikes and LFPs [28]
Decoding Algorithms	ReFIT Kalman Filter [29]	Continuous cursor control from neural signals	High-performance 2D control for point-and-click interfaces [29]
State Classification	Hidden Markov Models (HMM) [29]	Discrete selection ("click") detection	Robust classification of intentional selection commands [29]
Stimulus Presentation	Matrix Speller [26], RSVP [26]	Visual paradigms for ERP elicitation	Configurable row/column or rapid serial visual presentation [26]

Consciousness Assessment and Communication Timing

A critical advancement in LIS care involves determining the optimal timing for communication attempts through objective assessment of consciousness states. Researchers have developed methodologies to analyze EEG data from LIS patients to assess consciousness levels, referred to as normalized consciousness levels (NCL) [25]. This approach extracts multiple features based on frequency, complexity, and connectivity measures to determine the probability of a patient being fully conscious, addressing the fundamental challenge of identifying when to initiate communication with non-communicative patients [25].

Figure 2: Consciousness Assessment Workflow

Future Directions and Clinical Translation

The future of BCI communication technology points toward several promising directions. Integration with artificial intelligence is expected to enhance signal analysis accuracy and enable personalized BCI systems that adapt to individual users [31]. Advances in neurotechnology hardware, particularly in implantable electrodes and non-invasive sensing techniques, will improve both functionality and long-term safety [31]. Most importantly, multidisciplinary approaches that combine expertise from researchers, clinicians, and ethicists are essential for developing effective and ethical BCI solutions that can successfully transition from laboratory demonstrations to practical clinical tools [31].

Critical challenges remain in achieving widespread clinical adoption. Current systems face technical hurdles related to signal stability and system robustness [31], ethical concerns regarding privacy and data security of neural data [31], and practical barriers including high costs and need for technical support [31]. Overcoming these challenges requires continued collaboration across disciplines to ensure that the profound benefits of restored communication can reach the patients who need them most.

The restoration of communication for individuals with Locked-in Syndrome represents a critical imperative that transcends mere technical achievement, fundamentally impacting quality of life, personhood, and clinical care. Research has demonstrated that BCI technology can successfully provide communication channels through various approaches, from non-invasive EEG-based systems to sophisticated intracortical interfaces. Recent advances in hybrid systems, stable long-term implantation, and rapid speech decoding have dramatically improved performance to practically useful levels. As the field progresses, the integration of sophisticated signal processing, artificial intelligence, and conscious assessment methodologies promises to further enhance the reliability and accessibility of these transformative technologies. The ongoing translation of BCI communication systems from research laboratories to clinical practice offers hope for restoring the fundamental human right of communication to those with severe motor impairments, ultimately enabling greater autonomy, improved medical care, and enhanced quality of life.

Locked-in Syndrome (LIS) presents a unique paradox in neurology: preserved consciousness and cognition in the context of profound motor paralysis. This condition, most often resulting from vascular injury to the ventral pons or midbrain, leaves patients tetraplegic and anarthric while largely sparing cerebral function and awareness [32] [33]. The established clinical profile of LIS emphasizes the preservation of consciousness and vertical eye movements, but emerging research reveals a more nuanced and variable cognitive picture that critically impacts communication rehabilitation strategies, particularly with brain-computer interfaces (BCIs) [32].

Understanding this variable cognitive profile is essential for developing effective BCI-based communication tools. While patients retain consciousness, specific cognitive domains—especially executive functions and complex comprehension—may be significantly impacted, creating unexpected barriers to BCI use [32] [34]. This technical review synthesizes current evidence on the cognitive profile of LIS, provides detailed experimental methodologies for its assessment, and establishes a framework for aligning BCI communication protocols with individual cognitive capabilities.

Neuropathological Basis of Cognitive Variability

The pathophysiological basis of LIS begins with lesions in the ventral pons or caudal ventral midbrain, typically from ischemic or hemorrhagic infarction affecting the vertebrobasilar artery system [32] [33]. While this spares the cerebrum and reticular activating system (preserving consciousness), the cognitive variability observed in LIS suggests that the functional isolation of the cerebrum is not complete.

As illustrated in Table 1, the brainstem structures affected in LIS create a disconnection between intact cerebral function and motor output. However, the observed impairments in higher-order cognition suggest possible secondary mechanisms including:

• Cerebellar involvement in cases where lesions extend beyond the ventral pons
• Chronic cerebral hypoperfusion resulting from the primary brainstem injury
• Environmental factors including reduced mental and physical stimulation that impact neuroplasticity [32]

Table 1: Brainstem Structures and Their Status in Classical LIS

Structure(s)	Function	Status in Classical LIS
CN 3, 4	Vertical eye movement, pupillary reflex, eyelid control	Intact
CN 6, paramedian pontine reticular formation	Bilateral horizontal gaze	Injured
Corticobulbar tracts (CN 5, 7, 9, 10, 11, 12)	Facial, oropharyngeal, and respiratory function	Variable
Corticospinal tracts	Limb and truncal motor functions	Injured
Medial lemniscus and spinothalamic pathways	Sensation	Intact
Reticular activating system	Arousal, consciousness, awareness	Intact

These neuropathological mechanisms explain why cognitive function is most impaired immediately post-injury, with common acute deficits in attention, memory, and cognitive endurance [32]. While many functions recover, residual deficits often persist in specific domains that directly impact communication capacity.

The Variable Cognitive Profile in LIS: Quantitative Analysis

Research indicates a distinct pattern of cognitive sparing and impairment in chronic LIS. Observational studies suggest recovery of multiple core functions, but persistent deficits in higher-order processing [32]. The table below synthesizes quantitative findings across cognitive domains:

Table 2: Cognitive Domain Profiles in Chronic LIS

Cognitive Domain	Typical Status in Chronic LIS	Functional Implications for Communication
Reading Comprehension	Preserved for basic material [32]	Foundation for text-based BCI systems
Oral Comprehension	Preserved for simple commands [32]	Enables following BCI task instructions
Visual Recognition	Generally intact [32]	Supports visually-based BCI paradigms
Short-term Memory	Typically recovered [32]	Essential for maintaining task context
Intellectual Functioning	Generally preserved [32]	Maintains capacity for learning BCI control
Complex Sentence Comprehension	Often impaired [32]	Limits use of complex instructions
Mental Calculation	Frequently impaired [32]	Impacts performance in certain BCI tasks
Problem Solving	Often deficient [32]	Affects troubleshooting of BCI systems
Working Memory	Commonly impaired [32]	Reduces capacity for multi-step commands
Mental Flexibility	Typically impaired [32]	Limits adaptation to changing BCI paradigms
Executive Functioning	Generally impaired [32]	Impacts overall BCI learning and performance

This profile reveals a critical pattern: while basic cognitive capacities remain sufficient for communication, the specific executive functions necessary for learning and operating BCIs may be compromised. This explains why some LIS patients struggle with BCI protocols that healthy subjects find manageable [34].

Additionally, environmental and psychological factors compound these cognitive challenges. Patients often experience chronic low mental stimulation and social isolation, which can exacerbate cognitive deficits and predispose patients to depression, which may further manifest as attention and memory difficulties [32].

Assessment Protocols for Consciousness and Cognition

Accurate assessment of consciousness and cognitive function is prerequisite to developing appropriate communication strategies. Standardized behavioral assessment scales represent the first-line approach, though they present challenges when language impairments exist [35].

Behavioral Assessment Tools

The Coma Recovery Scale-Revised (CRS-R) is the most sensitive tool for differentiating between disorders of consciousness, examining arousal, auditory and visual perception, motor and oro-motor abilities, and communication skills [35]. The Simplified Evaluation of CONsciousness Disorders (SECONDs) provides a rapid alternative with substantial agreement with CRS-R (κ = 0.78-0.85), including the five most frequent signs of conscious awareness plus arousal and communication items [35].

EEG-Based Consciousness Assessment

Advanced EEG analysis provides objective measures of consciousness levels when behavioral assessment is limited by motor paralysis. The methodology below has been validated for assessing consciousness in LIS patients [25] [36]:

Table 3: Experimental Protocol for EEG-Based Consciousness Assessment

Protocol Component	Specifications	Application in LIS
EEG Acquisition	16-channel g.Nautilus PRO headset, 24-bit resolution [37]	High-quality signal acquisition at bedside
Feature Extraction	Frequency, complexity (Lempel-Ziv), connectivity measures [25] [36]	Multi-dimensional consciousness assessment
Paradigms	Auditory Evoked Potentials (AEP), Vibro-tactile Evoked Potentials (VEP), Motor Imagery [37]	Flexibility based on patient capabilities
Stimulus Parameters	AEP: high/low frequency tones; VEP: vibrations on left/right body parts [37]	Elicitation of measurable brain responses
Analysis Method	Normalized Consciousness Level (NCL) computation (0-1 scale) [25] [36]	Standardized metric for consciousness likelihood
Validation	Comparison with behavioral measures and clinical outcome [25] [37]	Verification of assessment accuracy

Diagram 1: LIS Cognitive and BCI Assessment Workflow. This flowchart illustrates the integrated process for assessing consciousness, evaluating cognitive domains, and selecting appropriate BCI communication protocols based on individual patient profiles.

BCI Communication Protocols and Cognitive Alignment

Brain-Computer Interfaces represent the most promising technology for restoring communication in LIS, but their effectiveness depends on proper alignment with the patient's cognitive profile. Both non-invasive and invasive approaches have demonstrated potential, with varying requirements for cognitive capacity.

Non-Invasive BCI Approaches

Auditory P300 BCI Protocol:

• Stimuli: Spoken words "yes" and "no" delivered alternately to right and left ears with standard (100ms) and deviant (150ms) durations [34]
• Task Instruction: Patients focus attention on the relevant stream ("yes" or "no") based on their intended response
• EEG Analysis: Detection of P300 event-related potentials and attentional modulations in response to target stimuli
• Performance: Healthy subjects achieve 86% accuracy (50 questions), but most LIS patients perform at chance level, with only 2/7 patients with severe motor disability achieving 100% accuracy [34]

Vibro-tactile P300 BCI Protocol:

• Stimuli: Vibrators placed on left and right hands delivering tactile stimuli [37]
• Task Instruction: Patients count vibrations on designated hand (left for "yes," right for "no")
• EEG Analysis: Classification of attentional modulations in sensorimotor cortex
• Performance: 5 out of 6 LIS patients successfully elicited P300 responses, demonstrating higher feasibility than auditory paradigms for some patients [25]

Invasive BCI Approaches

Intracortical BCI with Local Field Potentials (LFPs):

• Implant: 96-channel intracortical multielectrode array in dominant precentral gyrus [28]
• Signal Extraction: Local field potentials (LFPs) rather than sorted spikes for enhanced stability
• Task: FlashSpeller application with characters presented for 1.5-2 seconds each
• Performance: Stable spelling rates of 3.07-6.88 correct characters/minute maintained for 76-138 days without recalibration [28]

The significant performance gap between healthy subjects and LIS patients in auditory BCI paradigms [34] underscores the critical influence of LIS-related cognitive factors on BCI usability. This supports the need for individualized cognitive assessment and protocol selection.

Research Reagents and Experimental Solutions

The following table details essential research tools and methodologies for investigating cognitive function and developing communication solutions in LIS:

Table 4: Essential Research Solutions for LIS Cognitive and Communication Research

Research Tool	Specifications	Experimental Application
mindBEAGLE System	16-channel g.Nautilus PRO EEG, auditory/vibro-tactile stimulation, transcranial electrical stimulation capability [37]	Integrated assessment of consciousness, command-following, and basic communication
g.BSanalyze Software	Offline EEG analysis package for resting-state data [37]	Advanced analysis of consciousness markers and cognitive processing
Auditory Oddball Paradigm	Spoken words "yes"/"no" (100ms/150ms duration), 250ms SOA, lateralized presentation [34]	Assessment of auditory attention and P300 response capability
Vibro-tactile Stimulation System	Vibrators for left/right hand placement, programmable stimulation patterns [37]	Assessment of tactile attention and somatosensory processing
Intracortical Microelectrode Array	96-channel Blackrock Microsystems array, LFP and spiking activity recording [28]	Long-term stable signal acquisition for invasive BCI communication
FlashSpeller Application	Character presentation 1.5-2s duration, 0.1-0.3s inter-stimulus interval [28]	Text entry interface for BCI communication systems

Diagram 2: BCI Communication Protocol Decision Framework. This diagram illustrates the selection process for BCI communication protocols based on patient sensory capabilities and cognitive profiles, with performance outcomes observed in LIS populations.

The cognitive profile of Locked-in Syndrome presents a complex picture of preserved basic cognition with variable executive function and comprehension deficits. This profile directly impacts the effectiveness of BCI communication systems, which often place significant demands on precisely those cognitive domains that are most compromised in LIS.

Successful communication restoration requires multidimensional assessment of both consciousness and specific cognitive capacities, followed by individualized protocol selection that matches BCI paradigms to preserved cognitive strengths. The significant performance gap between healthy subjects and LIS patients in certain BCI paradigms underscores that technical feasibility in healthy populations does not guarantee success with the target LIS population.

Future research must prioritize the development of cognitive-friendly BCI protocols that accommodate the specific executive function and working memory challenges common in LIS. Furthermore, longitudinal studies tracking cognitive recovery patterns could identify optimal timing for introducing increasingly complex communication tools. By aligning BCI development with the nuanced cognitive reality of LIS, researchers can create more effective communication solutions that truly restore connection and quality of life for this unique population.

BCI Modalities and Applications: Bridging the Brain to the Outside World

Locked-in Syndrome (LIS) is a profound neurological condition characterized by complete paralysis of nearly all voluntary muscles while cognitive function and consciousness remain largely intact. Typically resulting from ventral pons lesions in the brainstem due to stroke, traumatic brain injury, or progressive neurological diseases like Amyotrophic Lateral Sclerosis (ALS), LIS leaves patients conscious but unable to speak or move, with communication often limited to vertical eye movements or blinking [38] [32]. The estimated prevalence is approximately 1 in 20,000 individuals, though this may be underestimated due to frequent misdiagnosis [38]. For these patients, the establishment of functional communication channels is not merely a convenience but a fundamental imperative for quality of life, autonomy, and psychological well-being.

Non-invasive Brain-Computer Interface (BCI) technology represents a revolutionary approach to restoring communication for LIS patients by translating brain signals directly into commands without requiring muscular control. Electroencephalography (EEG)-based systems have emerged as particularly promising due to their safety, portability, and relatively low cost compared to invasive alternatives [39]. Among EEG-based approaches, three paradigms have demonstrated significant potential: the P300 speller, Steady-State Visual Evoked Potentials (SSVEP), and Motor Imagery (MI). These technologies offer a crucial lifeline to the external world, potentially mitigating the "unbearable condition of complete isolation" that LIS patients can experience [40]. This review provides an in-depth technical examination of these non-invasive frontiers, their experimental protocols, performance characteristics, and implementation frameworks within LIS communication research.

P300 Speller Systems: Foundations and Advancements

Fundamental Mechanisms and Neurophysiological Basis

The P300 speller is one of the most extensively researched and implemented BCI paradigms for communication. It is based on the P300 event-related potential (ERP), a positive deflection in the EEG signal occurring approximately 300ms after the presentation of a rare or significant stimulus within a series of common stimuli [40] [41]. This neurophysiological response is elicited through the "oddball" paradigm, where users focus attention on infrequent target stimuli interspersed among frequent non-target stimuli.

The classical P300 speller implementation, first introduced by Farwell and Donchin, employs a 6×6 matrix containing letters, numbers, and symbols [40]. Rows and columns of this matrix flash in random sequence, and when the desired character flashes, it constitutes the rare target stimulus that elicits the P300 response. By detecting which row and which column produce this characteristic signal, the system can identify the specific character the user intends to select [40] [41]. The neural generators of the P300 are primarily located in the parietal cortex, making this signal particularly suitable for BCI applications due to its consistent topography and temporal characteristics [40].

Experimental Protocols and Implementation

Successful implementation of a P300 speller requires careful experimental design and signal processing. The standard protocol involves:

• Stimulus Presentation: Visual stimuli (rows/columns) are intensified for approximately 100ms with inter-stimulus intervals typically ranging from 70-150ms [40]. The stimulation sequence is randomized to ensure that each row and column flashes once per trial.

• EEG Acquisition: EEG data is recorded from multiple electrodes, with focus on parietal sites (P3, Pz, P4, etc.) where the P300 response is most prominent. The recorded signals are typically sampled at 256Hz or higher and bandpass filtered between 0.1-30Hz [41].

• Signal Processing: Advanced algorithms are employed for P300 detection. Common approaches include:

  • Wavelet transformations combined with Support Vector Machines (SVM) [41]
  • Stepwise Linear Discriminant Analysis (SWLDA)
  • Spatial filtering techniques such as xDAWN

• Classification: The processed features are fed into a classifier that determines whether a P300 response was present for each flash event, enabling identification of the target row and column.

Sellers et al. demonstrated the successful application of a P300 speller in a 68-year-old male with LIS following brainstem stroke [40]. Their protocol, initiated 6 months post-stroke, involved 62 sessions across 56 weeks. The researchers employed both copy spelling (for system calibration) and free spelling modes. When initial calibration accuracy exceeded 70%, the system switched to free spelling mode, allowing the participant to autonomously compose messages [40].

Table 1: Performance Metrics of P300 Speller Variations in LIS Communication

Speller Type	Sessions	Mean Accuracy	Application Context	Key Findings
Standard 6×6 Speller	7	32.3%	Initial testing	Limited efficacy in LIS
Four-Choice Speller (Y/N/P/E)	7	94.7%	Binary questions	High accuracy for basic communication
Seven-Choice Toggle Speller	4	85.6%	Letter selection	Effective but required navigation through sub-menus
3×6 Speller	19	81.5%	Free communication	Enabled expression of complex needs and desires

Adaptive Paradigms for LIS Applications

A critical finding in P300 speller research for LIS populations is the need for paradigm adaptation. Sellers et al. discovered that the standard 6×6 matrix achieved only 32.3% accuracy in copy spelling mode with their LIS participant [40]. However, when they implemented simplified versions—starting with a four-choice speller (Yes, No, Pass, End) that achieved 94.7% accuracy—they established a foundation for more complex communication. This evolved into a seven-choice toggle speller (85.6% accuracy) and eventually a 3×6 speller (81.5% accuracy) that enabled free communication, including personal expressions like the desire to buy presents for family members [40].

The comparison between the 3×6 speller and a traditional letter board revealed the BCI's advantage: the speller took 24 minutes with correct selection every 1.33 minutes to complete a sentence, compared to 29 minutes with correct selection every 3.22 minutes for the letter board [40]. This demonstrates not only the functional utility but also the efficiency potential of well-calibrated P300 systems for LIS communication.

SSVEP-Based BCIs: Visual Stimulation Paradigms

Fundamental Principles and Neural Mechanisms

Steady-State Visual Evoked Potentials (SSVEP) represent another prominent approach to non-invasive BCI. SSVEPs are oscillatory brain responses elicited in the visual cortex when a user focuses attention on a visual stimulus flickering at a fixed frequency [42]. These responses manifest as increased EEG power at the fundamental frequency of the visual stimulus and its harmonics, primarily observed over occipital and parieto-occipital brain regions [42].

The SSVEP response mechanism involves the synchronization of neural populations in the primary and secondary visual cortices to the rhythmic external stimulation. Unlike the transient P300 response, SSVEP provides a continuous, frequency-tagged signal that can be reliably detected through frequency analysis methods. This characteristic makes SSVEP-based BCIs particularly valuable for achieving high information transfer rates (ITR) in communication applications [42].

Experimental Protocols and Advanced Paradigms

Standard SSVEP-BCI protocols involve:

• Stimulus Design: Visual stimuli typically flicker at frequencies between 6-40 Hz, with the most robust responses often observed in the 8-15 Hz range [42]. Stimuli can be presented on traditional LCD/LED displays or through emerging technologies like head-mounted augmented reality (AR) devices.

• EEG Acquisition: Signals are recorded from occipital and parieto-occipital electrodes (O1, O2, Oz, POz, PO3, PO4, etc.) with sampling rates ≥256Hz. Electrode impedance is maintained below 10kΩ to ensure signal quality [42].

• Signal Processing: Common processing methods include:

  • Canonical Correlation Analysis (CCA)
  • Ensemble Task-Related Component Analysis (TRCA)
  • Power Spectral Density Analysis

• Target Identification: The system identifies the attended target based on which stimulus frequency demonstrates the strongest neural response.

Recent innovations in SSVEP-BCIs include binocularly incongruent paradigms utilizing AR headsets. Unlike traditional approaches that present the same stimulus to both eyes, these novel systems can deliver different flicker frequencies or phases to each eye simultaneously [42]. This binocular dual-frequency coding strategy has demonstrated superior performance compared to conventional congruent stimulation, achieving higher target separability and information transfer rates [42].

Table 2: SSVEP-BCI Performance Across Different Implementation Platforms

Platform	Targets	Accuracy	ITR (bits/min)	Key Features
Traditional LCD	40-160	>90%	Up to 200 [42]	High performance but limited portability
Head-Mounted AR	6-8	87.5-93.3%	36.84-45.57 [42]	Wearable, integrates virtual and real environments
Binocular Incongruent AR	8	Improved over congruent	Higher than congruent [42]	Enhanced target separability, reduced UIHC

Hybrid SSVEP Applications in Rehabilitation

Beyond communication, SSVEP-BCIs have shown promise in motor rehabilitation through hybrid paradigms. Recent research has integrated SSVEP with action observation (AO) and motor imagery (MI) to create robust systems for motor recovery [43]. In one study, participants observed alternating flickering hand pictures (AO condition), performed MI while focusing on flickering hand images (MI condition), or combined both (AO+MI condition) [43].

The results demonstrated that AO combined with MI more effectively activated the motor cortex compared to AO alone. The hybrid classification approach achieved impressive accuracy rates: 86.42% ± 8.42% for AO, 88.54% ± 10.31% for MI, and 88.91% ± 9.61% for AO+MI conditions [43]. This suggests that SSVEP-based hybrid BCIs offer a promising avenue for developing more robust and effective neurorehabilitation systems, potentially benefiting LIS patients with residual motor recovery potential.

Motor Imagery-Based BCIs: Cognitive Movement Simulation

Neural Correlates and Physiological Basis

Motor Imagery (MI)-based BCIs utilize the volitional imagination of movement without actual motor execution. This cognitive process activates similar neural networks to those involved in physical movement preparation and execution, primarily engaging the sensorimotor cortex [44]. The key electrophysiological phenomena associated with MI are Event-Related Desynchronization (ERD) and Event-Related Synchronization (ERS).

ERD represents a decrease in oscillatory power in the mu (8-12 Hz) and beta (13-30 Hz) frequency bands over sensorimotor areas during movement imagination, reflecting cortical activation and increased neural firing [44]. Conversely, ERS typically follows movement termination and manifests as increased power in these bands, representing cortical idling or inhibition. These oscillatory changes provide the control signals for MI-based BCIs, allowing users to modulate their brain activity to communicate or control external devices [44].

Experimental Protocols and Rehabilitation Applications

Standard protocols for MI-BCI implementation include:

• Task Design: Participants imagine specific motor acts (e.g., hand grasping, finger tapping) in response to visual or auditory cues, with typical trial durations of 4-8 seconds including rest periods.

• EEG Acquisition: Signals are recorded from sensorimotor electrodes (C3, Cz, C4, and surrounding locations) with focus on mu and beta frequency bands.

• Signal Processing: Common approaches include:

  • Common Spatial Patterns (CSP) for feature extraction
  • Filter Bank CSP (FBCSP)
  • Riemannian geometry-based classification

• Feedback Provision: Real-time visual or auditory feedback is provided to facilitate learning and improve BCI control.

In a recent pilot study with ischemic stroke patients, MI-based BCI training combined with robotic hand assistance demonstrated significant improvements in motor function across all participants [44]. EEG analysis revealed ERD in the high-alpha band power at motor cortex locations, though with considerable individual differences in both the frequency and power of neural activity [44]. This highlights the importance of personalized approaches in MI-BCI implementation, particularly for patients with neurological conditions.

Auditory MI-BCI for Complete Locked-In State

For patients in the completely locked-in state (CLIS) who may lack reliable visual function, auditory MI paradigms offer a promising alternative. In a groundbreaking study, a CLIS patient with ALS was able to operate an EEG-based BCI using volitional modulation of alpha and beta band power guided by real-time auditory feedback [45]. The system enabled "Yes"/"No" communication by mapping increases in band power to affirmative responses and baseline activity to negative responses.

Over multiple online sessions, the patient achieved above-chance performance in answering both general knowledge and personally relevant assistive questions, achieving perfect scores in the final session for assistive needs communication [45]. This demonstrates that non-invasive EEG-BCIs can potentially restore basic communication even in CLIS, addressing one of the most challenging scenarios in neurological care.

Hybrid BCI Systems: Integrating Multiple Modalities

Paradigm Fusion and Implementation

Hybrid BCI systems that combine multiple signal modalities have emerged as a powerful approach to enhance performance, robustness, and versatility. The integration of P300 and SSVEP paradigms has been particularly successful, leveraging the complementary strengths of both approaches [41].

One innovative implementation is the Frequency Enhanced Row and Column (FERC) paradigm, which incorporates frequency coding into the traditional P300 speller matrix [41]. In this design, each row and column flashes not only to elicit P300 responses but also at a specific frequency to simultaneously evoke SSVEP responses. This dual-stimulation approach provides two independent information channels for target identification, significantly improving accuracy and speed compared to single-modality systems [41].

Performance Advantages of Hybrid Systems

Research demonstrates clear advantages of hybrid P300-SSVEP systems over single-modality approaches. In offline tests, the hybrid FERC paradigm achieved 96.86% accuracy, substantially higher than P300-only (75.29%) or SSVEP-only (89.13%) configurations [41]. Online performance remained strong with 94.29% accuracy and an information transfer rate of 28.64 bits/min averaged across subjects [41].

Despite initial concerns about potential interference between simultaneous P300 and SSVEP stimuli, studies indicate that while some signal reduction may occur, the extracted features remain sufficiently discriminative for accurate classification [41]. This robustness makes hybrid systems particularly valuable for clinical applications where reliability is paramount.

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Research Materials and Equipment for EEG-Based BCI Research

Item	Function	Example Specifications	Application Notes
EEG Amplifier	Signal acquisition and digitization	32+ channels, sampling rate ≥512 Hz, impedance monitoring [45] [42]	Portable systems (e.g., eego, Neuroscan Grael) enable home-based studies
Electrode Caps	Standardized sensor placement	10/20 or 10/10 system, wet or dry electrodes [45]	Consistent positioning critical for reproducible results
Visual Stimulation Hardware	Eliciting evoked potentials	LCD/LED displays (60Hz+ refresh rate) or AR headsets (HoloLens 2) [42]	AR enables portable, immersive environments
Signal Processing Toolboxes	Algorithm implementation	Open-source platforms (EEGLAB, BCILAB, OpenBCI) [39]	Facilitates reproducible analysis and method comparison
Classification Algorithms	Pattern recognition and intent decoding	SVM, TRCA, CCA, Deep Learning models [43] [41]	Ensemble methods often outperform single algorithms
Robotic Assistive Devices	Providing haptic feedback	Exoskeleton hands (e.g., RxHEAL system) [44]	Closed-loop systems enhance neuroplasticity in rehabilitation

Implementation Workflows and Signaling Pathways

The experimental workflow for implementing EEG-based BCIs follows a systematic sequence from signal acquisition to application control. The process begins with proper electrode placement according to international 10-20 or 10-10 systems, ensuring optimal signal capture from relevant brain regions. Following signal acquisition and preprocessing to remove artifacts, feature extraction identifies the characteristic neural patterns associated with each BCI modality. Classification algorithms then translate these features into control commands, which drive the output applications—whether communication spellers, neurorehabilitation tools, or assistive devices. This entire process operates in a closed-loop system, with real-time feedback provided to the user to facilitate learning and improve control precision [44].

Diagram 1: EEG-Based BCI Implementation Workflow. The process begins with signal acquisition, progresses through processing and decoding stages, and culminates in application control, with continuous user feedback enabling closed-loop system optimization.

The neural signaling pathways differ across BCI paradigms, engaging distinct but sometimes overlapping brain networks. P300 responses primarily involve parietal regions for context updating and attention allocation. SSVEP signals originate in visual cortices with synchronization to external rhythmic stimuli. Motor imagery engages sensorimotor networks, producing characteristic ERD/ERS patterns in mu and beta rhythms. Hybrid systems simultaneously engage multiple of these pathways, providing complementary information streams that enhance overall system performance and robustness.

Diagram 2: Neural Signaling Pathways in Hybrid BCI Systems. Different stimulus modalities engage distinct neural pathways producing characteristic signatures, which are detected through specialized methods and integrated to generate robust BCI control signals.

EEG-based P300 spellers, SSVEP, and motor imagery represent three foundational pillars of non-invasive BCI research for LIS communication. Each approach offers distinct advantages: P300 systems provide intuitive matrix-based spelling, SSVEP enables high-speed communication, and motor imagery offers more endogenous control without requiring external stimulation. The emerging trend toward hybrid systems that combine multiple modalities promises enhanced robustness and performance, potentially overcoming limitations of individual approaches.

Future research directions should focus on several critical areas. First, improving individual adaptation and calibration procedures is essential, particularly given the significant inter-subject variability observed across studies [44]. Second, developing more sophisticated signal processing algorithms, including deep learning approaches, may enhance detection accuracy and reduce calibration requirements. Third, advancing wearable BCI technology through AR/VR integration and dry electrode systems will increase practicality for daily use [42]. Finally, establishing standardized evaluation metrics and protocols will facilitate more meaningful comparisons across studies and accelerate clinical translation.

For LIS patients, these technological advances represent more than academic exercises—they offer the potential to restore fundamental human connections and autonomy. As BCI technology continues to evolve, the prospect of seamless, reliable communication for even the most severely paralyzed individuals appears increasingly achievable, promising to transform quality of life for those living with locked-in syndrome.

For patients with locked-in syndrome (LIS), a severe neurological disorder characterized by complete paralysis of nearly all voluntary muscles while cognitive function remains intact, brain-computer interfaces (BCIs) represent a critical pathway to restore communication [25]. The efficacy of these communication systems depends fundamentally on the quality of neural signals acquired from the brain, making the choice of neural interfacing technology paramount. Two primary invasive approaches have emerged for high-fidelity signal acquisition: electrocorticography (ECoG), which records from the surface of the brain, and intracortical microelectrodes, which penetrate into brain tissue to record from individual neurons [46]. This technical guide provides an in-depth analysis of both technologies, their performance characteristics, experimental methodologies, and their specific applications to LIS BCI communication research, framing this discussion within the broader context of restoring natural communication capabilities to completely non-verbal patients.

Electrocorticography (ECoG)

ECoG involves the surgical placement of electrode grids or strips directly onto the exposed cortical surface of the brain. This technique has a well-established history in clinical neurology, particularly for epilepsy focus localization and intraoperative cortical mapping of motor, sensory, and language areas [46]. A significant advantage of ECoG is that electrodes can be easily placed, removed, and repositioned during surgery, while capturing signals across large areas of the brain surface [46]. However, this approach requires a larger craniotomy, which increases surgical invasiveness [46].

A key advancement in ECoG technology is the development of minimally invasive approaches. Recent innovations include origami-inspired soft fluidic actuators that enable large-area ECoG coverage through small burr-hole craniotomies [47]. These devices can be folded into a compressed state for implantation, then expanded on the cortical surface to cover large areas, significantly reducing surgical invasiveness while maintaining recording capability [47].

At the physiological level, ECoG electrodes capture local field potentials (LFPs), representing the summed activity of hundreds of thousands of neurons across several millimeters of cortex [46]. Research has demonstrated that despite the macroscopic nature of the recording, ECoG is surprisingly local, with a spatial spread diameter of approximately 3 mm in the primary visual cortex of awake monkeys [48]. ECoG signals primarily reflect coordinated activity across cortical layers, capturing broader brain rhythms but missing the detailed spiking activity of individual neurons [46].

Intracortical Microelectrodes

Intracortical microelectrodes penetrate 1-2 mm into the brain tissue, placing recording contacts in close proximity to individual neurons [46]. This proximity enables the detection of action potentials (spikes) from single neurons or small neural populations, providing access to the richest neural signals with the highest information content [46]. The fundamental advantage lies in the electrode's location: since electrical signals fall off quickly with distance, capturing single-neuron activity requires electrodes immediately adjacent to neurons [46].

Recent engineering advances have addressed chronic stability challenges through flexible designs. Ultraflexible stim-nanoelectronic threads (StimNETs) with shank thicknesses of just 1-1.3 μm have demonstrated seamless integration with nervous tissue, eliciting focal neuronal activation at low currents (2 μA) and maintaining stable behavioral responses for over 8 months with minimal tissue response [49]. These tissue-integrated electrodes provide a path for robust, long-lasting neural interfacing by reducing the foreign body response that typically compromises signal quality over time [49].

Table 1: Technical Comparison of ECoG and Intracortical Electrodes

Parameter	ECoG	Intracortical Electrodes
Spatial Resolution	Mesoscale (mm range); spatial spread ~3 mm diameter [48]	Microscale (μm range); single neuron resolution [46]
Signal Type	Local field potentials (LFPs); summed population activity [46]	Single-unit & multi-unit activity; action potentials [46]
Temporal Resolution	High (ms range), suitable for frequency analysis [50]	Very high (sub-ms), captures precise spike timing [46]
Invasiveness	Lower (surface recording); minimally invasive approaches available [47]	Higher (penetrates brain tissue); flexible designs reduce tissue damage [49]
Chronic Stability	Clinically established for weeks; limited by fibrous encapsulation [46]	Months to years with flexible electrodes; tissue integration key [49]
Information Rate	Limited by signal averaging [46]	At least 10x higher than ECoG [46]
Decoding Delay	Multi-second delays typical [46]	~100-200 milliseconds [46]
Surgical Approach	Large craniotomy or minimally invasive burr holes with expansion [47]	Small craniotomy with penetration into cortical tissue [46]

Performance Metrics for BCI Communication

The performance differential between ECoG and intracortical approaches becomes particularly evident in communication BCIs, where metrics such as vocabulary size, words per minute, and decoding latency directly impact functional utility. For patients with LIS, these parameters determine whether a communication system approaches natural conversational fluency or remains severely constrained.

Table 2: BCI Communication Performance Comparison

Performance Metric	ECoG Systems	Intracortical Systems	Natural Human Speech
Vocabulary Size	~50-1,000 words [46]	Up to 125,000 words [46]	Virtually unlimited
Communication Rate	Up to 78 words/minute (with 25% error rate) [46]	62 words/minute (23.8% error); 32 words/minute (2.5% error) [46]	~130 words/minute [24]
Decoding Latency	Multi-second delays [46]	~80-100 milliseconds [46] [24]	Near-instantaneous
Accuracy	Varies widely; up to 97% for limited gestures [46]	Up to 97.5% with large vocabulary [46]	~98-99% [46]
Training Requirements	Extensive for specific tasks/words	Rapid adaptation demonstrated [46]	N/A

For LIS patients, these performance differences have profound implications. While ECoG-based systems have enabled basic communication, intracortical approaches have demonstrated the ability to approach the vocabulary size and speed of natural human conversation. A recent breakthrough intracortical BCI achieved a 97.5% accuracy with a 125,000-word vocabulary, becoming the participant's preferred communication method and approaching the error rate of able-bodied speakers (1-2%) [46]. Another streaming BCI demonstrated the ability to translate brain activity into audible words in increments of 80 milliseconds, approaching synchronous communication [24].

Figure 1: Technology Decision Framework for LIS BCI Communication Research

Experimental Protocols and Methodologies

ECoG Signal Acquisition and Processing

High-density ECoG recording requires specific methodologies to maximize signal quality while minimizing invasiveness. A refined technique for chronic ECoG recording involves combining circumscribed trepanations with high-density electrode arrays at specific sites of interest, significantly reducing surgical risks and infection likelihood while preserving high spatial resolution [51].

Electrode Placement and Surgical Protocol:

• Pre-surgical planning using MRI/CT imaging to identify target regions
• Limited craniotomy optimized for electrode placement
• Subdural insertion of electrode grids/strips with direct cortical contact
• For minimally invasive approaches: origami-inspired folding and implantation through burr-hole craniotomy followed by fluidic actuation for expansion [47]
• Secure fixation of electrode leads and connector

Data Acquisition Parameters:

• Sampling rate: 512 Hz to 2,048 Hz depending on amplifier system [50] [52]
• Band-pass filtering: 0.16-250 Hz for clinical amplifiers [50]
• Referencing: Common average reference or two-contact subdural strip reference [50]
• Notch filtering at 60 Hz and harmonics to remove line noise [50]

Signal Processing Pipeline:

• Preprocessing: Remove bad electrodes, despike and interpolate high-amplitude artifacts, common average re-referencing [50]
• Spectral Analysis: Band-pass filtering into standard frequency bands (delta, theta, alpha, beta, gamma)
• High-Gamma Extraction: Apply Butterworth band-pass infinite impulse response filter at 70-200 Hz, extract broadband power using Hilbert transform envelope [50]
• Feature Extraction: Temporal correlations between electrodes, spectral power, phase relationships [53]

For natural language decoding studies, participants typically listen to or produce speech while ECoG signals are recorded. The "Podcast" ECoG dataset methodology involves presenting a 30-minute audio story while recording from 1,330 electrodes across 9 participants, providing a rich dataset for modeling neural activity during natural language comprehension [50].

Intracortical Signal Acquisition and Processing

Intracortical recording requires specialized microelectrodes and signal processing approaches to resolve single-neuron activity. The Connexus device from Paradromics, for example, features electrodes designed to reach 1.5 mm below the cortical surface where they can measure strong spiking activity of cortical processes [46].

Electrode Implantation Protocol:

• Small craniotomy over target region
• Precise insertion of microelectrode arrays using stereotactic guidance
• For flexible electrodes: careful insertion to avoid buckling while maintaining trajectory
• Secure anchoring to skull with dental acrylic or specialized platforms
• Connector placement and wound closure

Chronic Intracortical Recording Methodology:

• Ultra-flexible StimNETs (1 μm thickness) for minimal tissue disruption [49]
• 32 individually addressed microcontacts with 60 μm center-to-center spacing [49]
• In vivo two-photon imaging to verify tissue integration and record neuronal activation [49]
• Continuous stimulation at 50 Hz during simultaneous calcium imaging to quantify volumetric population activation [49]

Signal Processing for Single-Unit Activity:

• Spike Detection: Band-pass filtering (300-3000 Hz), threshold crossing detection
• Spike Sorting: Principal component analysis, clustering algorithms to isolate single units
• Feature Extraction: Firing rates, interspike intervals, burst patterns, population vectors
• Decoding Models: Kalman filters, deep learning networks for intention decoding

Validation Methods:

• Two-photon calcium imaging during electrical stimulation to map activation patterns [49]
• Behavioral tasks to correlate neural activity with output
• Histological analysis post-mortem to assess tissue response [49]

Figure 2: Experimental Workflow for LIS BCI Communication Development

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Materials for High-Fidelity Neural Signal Acquisition

Category	Specific Products/Technologies	Function & Application
ECoG Electrodes	Ad-Tech Medical Instrument grids/strips (2.3 mm diameter contacts) [50] [48]	Standard clinical ECoG recording; epilepsy monitoring & functional mapping
High-Density ECoG	FG64C-SP05X-000 grids (64 electrodes, 5 mm spacing) [52]	High-resolution cortical mapping; research-grade ECoG with improved spatial sampling
Minimally Invasive ECoG	Origami-inspired fluidic actuators [47]	Large-area coverage through small craniotomies; reduced surgical invasiveness
Intracortical Arrays	Paradromics Connexus Dots [46]	High-channel-count intracortical recording; designed for chronic implantation
Flexible Intracortical	StimNETs (1 μm thickness) [49]	Ultraflexible intracortical electrodes; minimal tissue response, chronic stability
Hybrid Arrays	Custom microelectrode + ECoG combinations [48]	Simultaneous recording at multiple spatial scales; signal comparison & validation
Data Acquisition	NicoletOne C64 (512 Hz), NeuroWorks Quantum (2048 Hz) [50] [52]	Signal amplification, filtering, and digitization; clinical and research applications
Signal Processing	Custom MATLAB/Python pipelines [50]	Spike sorting, LFP analysis, feature extraction, decoding model implementation

The choice between ECoG and intracortical approaches for LIS BCI communication involves fundamental trade-offs between invasiveness and performance. ECoG provides a less invasive option with established clinical utility, suitable for basic communication needs and large-scale brain monitoring. However, intracortical microelectrodes currently offer superior performance for restoring naturalistic communication, with higher information transfer rates, lower latencies, and larger vocabulary sizes that begin to approach natural human conversation. For LIS patients transitioning to complete locked-in state (CLIS), where no residual muscle movement remains and consciousness assessment becomes challenging [25], the higher-performance capabilities of intracortical approaches may provide the critical difference between basic communication and truly fluent interaction. As both technologies continue to advance—with ECoG becoming less invasive through soft robotics approaches [47] and intracortical electrodes achieving greater chronic stability through flexible designs [49]—the potential for restoring natural communication to LIS patients continues to improve, offering hope for enhanced quality of life and social connection for this vulnerable population.

Locked-In Syndrome (LIS) and its most severe form, Complete Locked-In Syndrome (CLIS), represent the ultimate communication barrier in neurodegenerative disease. Patients with Amyotrophic Lateral Sclerosis (ALS) progressively lose voluntary motor control, often culminating in a state of near-total physical isolation while cognitive function remains preserved [23] [25]. In early LIS stages, eye-tracking (ET) systems serve as vital communication tools, allowing interaction through residual eye movements and blinks. However, as oculomotor function declines with disease progression toward CLIS, these systems become unreliable, creating a devastating communication gap precisely when patients need it most [23] [54].

Electroencephalography (EEG)-based Brain-Computer Interfaces (BCIs) offer a non-muscular communication alternative, but delayed adoption presents a critical problem. The "extinction of goal-directed thinking" hypothesis suggests that prolonged periods without successful intentional interaction can diminish the very cognitive processes necessary to operate BCIs effectively [23] [54]. This creates a therapeutic catch-22: by the time patients transition to CLIS, they may have lost the goal-directed thinking required to use the BCI that could restore their communication.

The hybrid ET-BCI framework addresses this challenge by enabling a gradual transition between modalities. This approach allows early BCI familiarization while patients still retain reliable eye movement, potentially preserving cognitive engagement and communication continuity throughout disease progression [23] [54]. This technical guide explores the experimental foundations, methodologies, and future directions of this integrated approach within the broader context of LIS communication research.

Technical Foundations: Neural and Ocular Signaling Pathways

Neural Correlates of Consciousness and Communication

Consciousness assessment in LIS/CLIS patients presents significant challenges due to the absence of behavioral markers. Research indicates that LIS patients retain both consciousness and cognitive functions despite their inability to perform voluntary muscle movements [25]. Assessing this consciousness is vital for determining optimal communication timing. Studies analyze EEG data using features based on frequency, complexity, and connectivity measures to derive normalized consciousness levels (NCL) representing the likelihood of a patient being fully conscious [25].

The P300 event-related potential, a positive deflection in EEG signals occurring approximately 300ms after a rare or significant stimulus, serves as a robust neural marker for BCI communication. This auditory or visual evoked potential indicates conscious stimulus processing and can be voluntarily modulated by patients following commands, thereby providing a communication channel [25]. In CLIS patients, traditional visual P300 paradigms become ineffective, leading to the development of vibro-tactile P300 and motor imagery approaches that have achieved communication accuracies up to 90% [25].

Oculometric Indicators of Cognitive State

Beyond gaze coordination, eye-tracking parameters provide valuable indicators of user state during interface operation. Pupil dilation correlates with cognitive workload and attentional focus, while blink rate and fixation behavior offer additional metrics of user engagement and task complexity [54]. These ocular metrics can be integrated into hybrid BCI-ET systems to assess user state and potentially optimize system responsiveness dynamically, creating a more adaptive communication interface [54].

Experimental Protocols: Methodology for Hybrid System Development

Pilot Study Design and Participant Recruitment

The foundational research for ET-BCI integration involved five healthy female volunteers (mean age 22.8 ± 1.8 years) with normal vision [54]. While healthy participants cannot fully replicate the LIS/CLIS condition, initial validation with able-bodied individuals establishes protocol safety and feasibility before clinical translation. All participants provided informed consent, and the study implemented standard experimental controls including consistent viewing distance (60cm) and environment [54].

Table: Key Research Reagent Solutions for Hybrid ET-BCI Systems

Component	Specific Product/Model	Function in Experimental Setup
EEG Acquisition System	g.USBamp (g.tec medical engineering GmbH)	Records neural signals at 256 Hz sample rate from 16 electrodes [54]
EEG Electrodes	g.Ladybird electrodes	placed according to extended international 10-20 system [54]
Eye Tracker	Tobii Pro Spark ET	Records gaze data and pupil diameter at 60 Hz [54]
ET Software Toolbox	Titta toolbox	Provides wrap-around of Tobii Pro SDK for real-time processing [54]
Visual Presentation	Psychophysics toolbox	Controls stimulus presentation and timing for P300 oddball paradigm [54]
Stimulus Paradigm	Modified P300-based Visual Standard Face (Vsf)	Presents words in grid layout with flashing neutral face for P300 elicitation [54]

Multimodal Data Acquisition and Signal Processing

The hybrid system employs synchronized data acquisition from neural and ocular pathways:

EEG Acquisition Protocol: Neural signals were recorded from 16 g.Ladybird electrodes placed in Fz, Cz, C3, C4, Cz, CPz, Pz, P3, P4, PO7, PO8, POz, Oz, FPz, FCz, FC1, and FC2 according to the extended international 10-20 system. The ground electrode was positioned at AFz with reference at the right earlobe. Signals underwent real-time processing with a band-pass filter (0.1-30 Hz) and notch filter (50 Hz) to eliminate powerline interference [54].

ET Acquisition Protocol: Gaze data and pupil diameter were recorded simultaneously using the Tobii Pro Spark ET at 60 Hz sampling rate. The system was positioned below the acquisition laptop screen with participants seated approximately 60 cm from the screen. Head position and viewing distance were standardized before each session to ensure measurement consistency [54].

ET-BCI Fusion Algorithm and Experimental Paradigm

The visual BCI interface adapted from the P300-based Visual Standard Face paradigm presented a two-row by four-column grid with seven Portuguese words ('Sim', 'Não', 'Tosse', 'Ajuda', 'Stop', 'TV', and 'Sono') overlaid with a neural face during each stimulus event [54]. Participants were instructed to focus on a target word and mentally count the number of times the neutral face flashed over that word while ignoring non-targets—a classic oddball paradigm for P300 elicitation.

The core innovation lies in the ET-BCI fusion algorithm that combines confidence scores from both modalities:

• ET Confidence Scores: For each Area of Interest (AOI) k, the confidence score CET(k) was calculated as the ratio of gaze points falling within the AOI to the total gaze points recorded: CET(k) = nk/Ntotal [54].

• BCI Confidence Scores: EEG data were decoded using a Gaussian Naïve Bayes classifier, computing Bayesian scores by combining prior probability and class-conditional likelihood: Scoreik = Prioriᵀ × pdfGaussian(x∣μi,Σi) [54].

Figure 1: Hybrid ET-BCI System Workflow illustrating the parallel processing of neural and ocular signals culminating in a fused decision output.

Quantitative Results: Performance Metrics and Comparative Analysis

The hybrid approach validation with healthy participants demonstrated that combining both modalities maintains high accuracy while offering insights for improving communication continuity. The complementary nature of ET and BCI signals allows for graceful degradation as oculomotor function declines, with the BCI component assuming greater responsibility for communication detection [23] [54].

Table: Performance Comparison of Communication Modalities in LIS/CLIS

Communication Modality	Reported Performance Metrics	Advantages	Limitations
Eye-Tracking (ET) Systems	High accuracy in early LIS [54]	Fast, intuitive, minimal setup [54]	Becomes unreliable as oculomotor function declines [23] [54]
P300-based BCI	Good accuracy with LIS patients [54]	Non-muscular communication pathway [23]	Delayed adoption reduces performance; requires goal-directed thinking [23] [54]
Hybrid ET-BCI Framework	Maintains high accuracy during transition [23] [54]	Enables gradual modality transition, preserves communication continuity [23] [54]	Limited long-term data on CLIS patients [23]
Fully Implanted BCI	Up to 99% word accuracy, 56 words/minute in controlled tests [55]	Enables communication in CLIS [56]	Invasive surgical procedure required [56]

Figure 2: Transition Pathway from LIS to CLIS showing the shifting modality emphasis from ET-dominant to BCI-dominant communication.

Future Research Directions and Clinical Translation

Artificial Intelligence and Adaptive Learning

The integration of artificial intelligence and machine learning represents the most promising direction for enhancing hybrid BCI systems. Recent systematic reviews identify transfer learning, support vector machines, and convolutional neural networks as key ML techniques that improve signal classification, feature extraction, and real-time adaptability in BCI systems [57]. These approaches address the critical challenge of neural signal variability between users and across time in the same user, potentially reducing lengthy calibration sessions that pose particular difficulties for CLIS patients [57].

Invasive BCIs for Complete Locked-In State

While this whitepaper focuses primarily on non-invasive approaches, recent breakthroughs in implanted BCIs demonstrate the feasibility of communication even in CLIS. Researchers at the Wyss Center for Bio and Neuroengineering enabled a completely locked-in individual to communicate via an intracortical microelectrode array implanted in the motor cortex [56]. The participant learned to generate brain activity by attempting different movements, with signals decoded by a machine learning model to indicate 'yes' or 'no' responses, ultimately forming words and sentences through an auditory speller program [56]. This technology, used successfully at home for over two years, represents a significant advancement for the CLIS population where non-invasive methods may fail [56].

Personalized Digital Prescription Systems

Future BCI development trends point toward personalized digital prescription systems that deliver customized therapeutic strategies via digital platforms [10]. These systems would account for individual patterns of disease progression, residual motor function, and cognitive preservation to optimize the transition timing between ET and BCI modalities. Combined with closed-loop neurostimulation and flexible neural interfaces, this personalized approach could dramatically improve quality of life for LIS/CLIS patients [10].

The hybrid ET-BCI framework represents a paradigm shift in assistive communication technology for patients progressing from LIS to CLIS. By enabling seamless transition between communication modalities, this approach addresses the critical gap that has historically left patients isolated as oculomotor function declines. The integrated methodology leverages the complementary strengths of both technologies—the speed and intuitiveness of eye-tracking with the non-muscular communication pathway of BCIs.

While technical challenges remain in signal processing, user adaptation, and clinical implementation, the foundation established by current research points toward a future where communication continuity can be maintained throughout the devastating progression of ALS and other neurodegenerative diseases. Through continued innovation in AI-driven signal processing, minimally invasive technologies, and personalized adaptive systems, the goal of eliminating the communication isolation of locked-in syndromes appears increasingly achievable.

The success of future systems will depend not only on technological advancements but also on addressing ethical considerations, cost barriers, and implementation challenges through clear regulatory frameworks and multidisciplinary collaboration between engineers, clinicians, and end-users.

Locked-In Syndrome (LIS) represents one of the most challenging neurological conditions, characterized by preserved cognitive function amidst nearly complete voluntary muscle paralysis, including speech. Patients typically retain consciousness and awareness but lose the ability to communicate, relegated to communicating through limited means such as eye movements or blinking [25]. The progression of conditions like Amyotrophic Lateral Sclerosis (ALS) can lead to a Completely Locked-In State (CLIS), where even these residual communication channels are lost, creating profound isolation [23]. Traditional assistive technologies, including eye-tracking systems, become ineffective as oculomotor function declines, creating an urgent need for alternative communication pathways [23].

Brain-computer interfaces have emerged as a transformative technology for restoring communication in LIS. Recent breakthroughs in AI-powered speech neuroprostheses have achieved what was once considered science fiction: decoding neural signals into intelligible, naturalistic speech in near-real-time. These systems represent a paradigm shift from earlier text-based communication toward direct speech synthesis, potentially restoring not just communication but vocal identity and spontaneous conversation [58] [24] [59].

This technical guide examines the core principles, experimental methodologies, and performance metrics of contemporary brain-to-speech neuroprostheses, with particular focus on their application within LIS and CLIS research contexts.

Core Technological Framework

Neural Signal Acquisition Modalities

Brain-to-speech neuroprostheses employ various modalities for recording neural activity, each with distinct advantages and implementation considerations:

• Electrocorticography (ECoG) Arrays: These high-density electrode arrays are placed directly on the brain surface, providing high spatial and temporal resolution signals from the speech motor cortex. The clinical trial led by UC Berkeley and UCSF utilized this approach, recording from regions critical for speech production [58] [24].

• Microelectrode Arrays (MEAs): In the UC Davis BrainGate2 trial, researchers implanted four microelectrode arrays that penetrate the brain surface to record neuronal firing patterns with exceptional precision [59].

• Non-Invasive Alternatives: Research continues into non-invasive approaches, including EEG-based systems that measure electrical activity from the scalp [23] [25]. Hybrid frameworks that combine modalities, such as eye-tracking with EEG-based BCIs, offer potential pathways for gradual transition as patients progress from LIS to CLIS [23].

AI-Driven Speech Decoding Architecture

The core innovation enabling recent advances lies in artificial intelligence architectures that translate neural signals into audible speech:

• Real-Time Streaming Algorithms: Modern systems employ deep learning models that sample neural data in windows as brief as 80 milliseconds (0.08 seconds), enabling near-synchronous voice output [58] [24]. This represents a radical improvement over previous systems that experienced multi-second delays.

• Neural Feature Extraction: Algorithms identify and process relevant features from neural signals, including frequency patterns, complexity measures, and connectivity markers [25]. These features capture the neural correlates of speech motor commands.

• Acoustic Synthesis: The decoded representations are mapped to acoustic parameters using pretrained speech synthesis models. Notably, systems can incorporate a patient's pre-injury voice, preserving vocal identity [58] [24].

Table 1: Quantitative Performance Metrics of Recent Speech Neuroprostheses

Performance Metric	UC Berkeley/UCSF System	UC Davis BrainGate2 System	Traditional Assistive Technologies
Decoding Rate	47.5 WPM (full vocabulary)90.9 WPM (50-word vocabulary) [24]	Real-time streaming(25ms latency) [59]	Varies significantlyTypically <10-20 WPM
Latency	<80ms from speech attempt to audio output [24]	25ms delay [59]	Often several seconds
Vocabulary Size	>1,000 words [24]	Not specified	Limited
Intelligibility	High intelligibility>99% success rate [24]	~60% word accuracy [59]	Dependent on modality
Novel Word Production	Capable of decoding untrained words and sentences [58]	Able to produce new words not in training set [59]	Generally limited to predefined sets

Experimental Protocols and Methodologies

Participant Selection and Surgical Implementation

Current speech neuroprosthesis research focuses on individuals with severe paralysis resulting from conditions such as brainstem stroke or ALS. The pioneering UC Berkeley/UCSF study involved a 47-year-old woman who had been unable to speak for 18 years following a stroke [24]. Similarly, the UC Davis trial enrolled a participant with ALS [59]. These individuals represent the target population for this technology—cognitively intact but completely unable to vocalize.

Surgical procedures involve the implantation of electrode arrays over critical speech regions of the motor cortex. The UCSF team utilized high-density ECoG arrays placed on the brain surface [58], while the BrainGate2 trial implemented penetrating microelectrode arrays [59]. Both approaches require precise localization of speech-related cortical areas to optimize signal acquisition.

Training Data Acquisition and Algorithm Development

The foundation of effective speech decoding lies in comprehensive training datasets collected through specific experimental paradigms:

• Silent Speech Attempt Protocol: Participants are shown text prompts on a screen and attempt to speak them without vocalizing. In the UCSF study, their subject Ann silently attempted over 23,000 sentences containing more than 1,000 unique words [24]. This process creates a mapping between neural activation patterns and linguistic targets.

• Neural Data Alignment: Researchers align recorded neural signals with the intended speech output. The UC Davis team developed algorithms that "map neural activity to intended sounds at each moment of time" [59], capturing the precise temporal dynamics of speech production.

• Synthetic Target Generation: When actual audio output is unavailable (due to paralysis), researchers use pretrained text-to-speech models to generate target audio. The UCSF team reconstructed the participant's pre-injury voice using historical recordings [58] [24].

The following diagram illustrates the complete experimental workflow from surgical implantation to real-time synthesis:

Performance Validation and Testing

Rigorous validation methodologies ensure the efficacy and generalizability of speech neuroprostheses:

• Closed-Vocabulary Testing: Researchers assess performance using words and sentences included in the training set, measuring accuracy and latency metrics [24].

• Open-Vocabulary Assessment: Systems are tested with novel words not encountered during training. The UCSF team used 26 rare words from the NATO phonetic alphabet ("Alpha," "Bravo," "Charlie") to verify generalization capability [58].

• Intelligibility Metrics: Independent listeners evaluate synthesized speech outputs, reporting word accuracy rates. The UC Davis system achieved approximately 60% intelligibility [59], while the UCSF system exceeded 99% accuracy for trained vocabulary [24].

• Real-Time Conversation Assessment: Researchers observe and document the system's performance during unstructured conversation attempts, noting the potential for interruption, questioning intonation, and continuous dialogue [59].

Table 2: Key Research Reagents and Experimental Materials

Resource Category	Specific Examples	Research Function
Neural Recording Arrays	High-density ECoG gridsPenetrating microelectrode arrays (MEAs)	Capture neural signals from speech motor cortex with high spatial and temporal resolution [58] [59]
Signal Processing Tools	Custom deep learning algorithmsReal-time processing software	Translate neural activity into intended speech commands with minimal latency [58] [24]
Data Collection Paradigms	Visual presentation softwareSilent speech protocol frameworks	Generate training data mapping neural patterns to speech targets [24] [59]
Speech Synthesis Models	Pre-trained text-to-speech systemsVoice banking resources	Convert decoded neural patterns into audible speech, potentially using participant's pre-injury voice [58] [24]
Validation Metrics	Intelligibility assessment toolsLatency measurement systems	Quantify system performance and benchmark against communication thresholds [24] [59]

Current Limitations and Future Research Directions

Despite remarkable progress, significant challenges remain in the development of optimal speech neuroprostheses:

• Participant Scalability: Current studies involve limited participants (often single cases), necessitating expansion to diverse populations with varying etiologies of paralysis [59].

• Expressivity Enhancement: While intelligible, synthesized speech often lacks the emotional nuance of natural conversation. Future systems aim to decode paralinguistic features such as tone, pitch, and volume to convey emotional state [58] [59].

• Hardware Optimization: Fully implantable, wireless systems represent the next frontier, reducing infection risk and improving quality of life [24].

• CLIS-Specific Challenges: Reaching patients in Completely Locked-In States requires novel approaches, as they cannot participate in initial training protocols. Research into consciousness assessment using EEG features may help identify optimal communication windows [25].

The following diagram illustrates the neural signal processing pathway from acquisition to synthesis:

AI-powered brain-to-speech neuroprostheses represent a transformative advancement in assistive communication technology for individuals with locked-in syndrome. By leveraging sophisticated neural interfaces and deep learning algorithms, these systems can now decode and synthesize intelligible speech in near-real-time, dramatically improving communication rates over previous technologies. The restoration of naturalistic speech through these neuroprostheses not only enhances practical communication capabilities but also reaffirms personal identity and social connection for those affected by severe paralysis.

While challenges remain in scaling these technologies and enhancing their expressivity, the current state of the field demonstrates unprecedented potential for restoring authentic communication. As research progresses, these systems will likely become more sophisticated, eventually offering fully implantable, wireless solutions that seamlessly integrate into daily life, ultimately transforming the lived experience of locked-in syndrome.

Complete Locked-In Syndrome (CLIS) represents the most severe form of motor paralysis, where individuals lose all voluntary muscle control, including eye movements and blinking, thereby eliminating all standard routes of communication [60]. This state typically results from progressive neurological disorders such as advanced Amyotrophic Lateral Sclerosis (ALS) or brainstem stroke [61]. The clinical definition of CLIS requires the absence of consistent and reliable eye movement control, which renders eye-tracking communication systems useless [60]. For researchers and clinicians, this condition presents a profound challenge: how to detect consciousness and restore communication when an individual cannot produce any voluntary muscular responses. Before the advent of advanced Brain-Computer Interfaces (BCIs), no assistive technology could provide voluntary communication for people in CLIS [60]. This whitepaper examines two promising approaches—auditory neurofeedback training and yes/no communication paradigms—that have demonstrated potential to bypass the motor system and establish functional spelling interfaces for this population, framed within the broader context of locked-in syndrome BCI communication research.

Invasive Intracortical Approaches: Auditory Neurofeedback Training

Core Methodology and Experimental Protocol

The auditory neurofeedback approach for CLIS communication utilizes implanted microelectrode arrays to record neural signals, which are then transformed into an auditory interface. The foundational study demonstrating the viability of this method involved implanting two 64-microelectrode arrays in the supplementary motor cortex and primary motor cortex of a patient with CLIS resulting from ALS [60]. The methodology follows a structured protocol:

• Post-implantation baseline recording (10 minutes): The patient is instructed to rest while experimenters measure baseline firing rates across individual neural channels [60].
• Channel selection procedure: Software algorithms identify channels with differential modulation potential for high and low target tones through iterative testing [60].
• Neurofeedback training without reward: Patients attempt to match a feedback tone frequency to successive target tones by modulating their neural firing rates [60].
• Neurofeedback training with reward: When patients maintain the feedback tone within a predefined target range for a configurable number of interactions (each 250ms), a reward sound (250ms) reinforces successful performance [60].
• Performance thresholding: Patients must achieve ≥80% accuracy in neurofeedback tasks before advancing to spelling interfaces [60].
• Speller implementation: Successful neural modulation is translated to letter selection using custom software, enabling word and phrase formation [60].

This protocol established that a patient in CLIS could voluntarily modulate neural firing rates based on auditory feedback, achieving 86.6% accuracy across 5,700 trials over 356 days post-training [60].

Quantitative Performance Data

Table 1: Performance Metrics for Intracortical Auditory Neurofeedback in CLIS Communication

Performance Metric	Result	Context/Measurement Period
Overall Neurofeedback Accuracy	86.6%	4,936 correct trials out of 5,700 total trials [60]
Error Rate - 'Up' Trials	13.2%	Fraction where modulated tone didn't match high-frequency target [60]
Error Rate - 'Down' Trials	12.2%	Fraction where modulated tone didn't match low-frequency target [60]
Pre-Speller Neurofeedback Accuracy	90.0% (median)	Last feedback sessions before speller sessions [60]
Days with ≥90% Accuracy	52.6% of days	During at least one feedback trial block [60]
Communication Rate	~1 letter/minute	During productive spelling sessions [60]

Research Reagent Solutions for Intracortical Approaches

Table 2: Essential Research Materials for Intracortical BCI Spelling Interfaces

Research Reagent/Equipment	Function/Specification	Implementation Example
Microelectrode Arrays	Neural signal acquisition	Two 64-microelectrode arrays implanted in supplementary and primary motor cortex [60]
Spike Rate Metric (SRM)	Neural activity quantification	Real-time calculation from single or multiple channels for neurofeedback [60]
Auditory Feedback Tone	Real-time neurofeedback	Frequency mapped to spike rate metric; provides continuous performance feedback [60]
Custom Spelling Software	Letter selection interface	Translates binary "yes/no" neural responses to letter selection and word formation [60]
Signal Processing Pipeline	Neural data interpretation	Converts raw neural signals to classification outputs for communication commands [60]

Figure 1: Experimental workflow for intracortical auditory neurofeedback training in CLIS patients

Non-Invasive Auditory BCI: Yes/No Communication Paradigms

Experimental Design and Oddball Paradigm Methodology

Non-invasive auditory BCIs represent a less invasive alternative for establishing communication with individuals in CLIS. These systems typically employ an "oddball" paradigm using spoken words or sounds to elicit event-related potentials (ERPs) that can be classified as "yes" or "no" responses [61]. The standard methodology includes:

• Stimulus design: Using spoken words "yes" and "no" pronounced by a synthesized voice, with standard (100ms) and deviant (150ms) durations to create distinct neural responses [61].
• Lateralized presentation: "Yes" stimuli presented to the right ear, "no" stimuli to the left ear to create spatially distinct neural activation patterns [61].
• Stimulus Onset Asynchrony (SOA) adjustment: Setting SOA to 250ms for healthy subjects but customizing for individual patients based on responsiveness and fatigue levels [61].
• Instruction paradigm: Patients are instructed to selectively attend to either the "yes" or "no" stream based on their intended response while ignoring the other stream [61].
• EEG acquisition: Using standard EEG caps with multiple electrodes (varied by system), typically focusing on parietal and central regions for P300 detection [61].
• Classifier training: Employing machine learning algorithms (often linear discriminant analysis or support vector machines) to distinguish between attention to "yes" versus "no" streams based on ERP components including P300 and N200 [61].

This approach leverages the P300 event-related potential—a positive deflection in the EEG signal occurring approximately 300ms after a rare or significant stimulus—when the patient attends to their intended response choice [61].

Performance Data Across Populations

Table 3: Performance Metrics for Non-Invasive Auditory BCI Across Subject Groups

Subject Group	Sample Size	Performance Outcome	Notable Findings
Healthy Controls	18 subjects	86% average accuracy (50 questions)	10/18 subjects >90% accuracy; only 1 subject at chance level [61]
ALS Patients	4 patients	2 patients reached 100% accuracy	Significant performance variability across patients [61]
Classical LIS Patients	3 patients	Majority at chance level	Limited success with standard auditory paradigm [61]
All Patients Combined	7 patients	Mixed results, generally below healthy controls	Highlights challenge of translating to target population [61]

Research Reagent Solutions for Non-Invasive Approaches

Table 4: Essential Research Materials for Non-Invasive Auditory BCI

Research Reagent/Equipment	Function/Specification	Implementation Details
EEG Acquisition System	Neural signal recording	Vamp or similar systems; number of electrodes varies by setup [61]
Auditory Stimulus Set	Eliciting ERPs	Spoken words "yes"/"no" with standard (100ms) and deviant (150ms) durations [61]
Stimulus Presentation Software	Paradigm implementation	Controls SOA (typically 250ms), randomization, and lateralized presentation [61]
Signal Processing Pipeline	ERP analysis and classification	Extracts P3b and N200 components; implements classification algorithms [61]
Calibration Interface	System individualization	Adjusts parameters based on individual patient performance and responsiveness [61]

Figure 2: Signaling pathway for auditory oddball paradigm in yes/no BCI communication

Comparative Analysis and Research Implications

Performance and Practical Considerations

When evaluating both intracortical and auditory BCI approaches for CLIS communication, distinct advantages and limitations emerge. The intracortical auditory neurofeedback approach demonstrates higher overall performance (86.6% accuracy) and has proven effective in genuine CLIS cases [60]. However, this approach requires invasive surgery with associated risks of infection and hemorrhage, limiting its broad applicability [61]. Additionally, the implementation is highly resource-intensive, requiring specialized surgical expertise, custom software, and ongoing technical support [60].

In contrast, non-invasive auditory BCI approaches eliminate surgical risks and could potentially be deployed more widely [61]. However, current implementations show inconsistent results with the target CLIS population, with most patients unable to achieve control above chance level [61]. The translation gap between healthy subjects and impaired patients remains a significant challenge, possibly due to cognitive impairments, sensory deficits, or attention limitations in the clinical population [61].

Neurophysiological Mechanisms

The success of these approaches relies on distinct but complementary neural mechanisms. Intracortical approaches leverage voluntary modulation of motor cortex firing rates, demonstrating that despite complete motor paralysis, the intention to communicate remains intact and can be harnessed through neurofeedback [60]. The supplementary motor area appears particularly important for this voluntary control, as it's associated with movement intention and planning rather than execution.

Non-invasive auditory approaches instead rely on attention-modulated event-related potentials, particularly the P300 component and N200 modulation [61]. The P300 (P3b) reflects conscious attention to rare, task-relevant stimuli, while the N200 modulation in responses to standard sounds reflects sustained attention to the relevant stimulus stream. These findings confirm that despite motor paralysis, CLIS patients may retain the ability to direct selective attention, which can be harnessed for communication.

Future Research Directions and Clinical Translation

The development of spelling interfaces for CLIS represents a rapidly advancing frontier in BCI research. Future directions should focus on hybrid approaches that combine the reliability of intracortical signals with the practicality of non-invasive systems [62]. Research should also explore multimodal feedback systems that integrate auditory, tactile, and potentially residual visual pathways to enhance usability [63]. Additionally, standardized evaluation metrics specific to BCI usability and communication efficacy need development to enable cross-study comparisons [63].

From a clinical translation perspective, addressing the high costs associated with BCI technology (approximately $60,000 per unit for some systems) remains crucial for widespread adoption [64]. Furthermore, developing compact, wearable BCI systems that can be used outside research settings represents a critical step toward functional independence for CLIS individuals [64]. The recently reported development of a wearable BCI cap with machine learning that interprets brain activity for computer tasks using a "one-size-fits-all" approach to minimize training time shows promise in this direction [64].

As research progresses, spelling interfaces for CLIS have the potential to transform from experimental paradigms to standard clinical tools, finally providing a voice to those who have been trapped in silence. The continued collaboration between neuroscientists, engineers, clinicians, and individuals with lived experience of CLIS will be essential to realizing this potential and unlocking communication for all.

Overcoming Implementation Hurdles: Technical, Clinical, and Ethical Challenges

For individuals with locked-in syndrome (LIS), a condition characterized by complete paralysis of nearly all voluntary muscles while cognitive functions remain intact, brain-computer interfaces (BCIs) represent a critical pathway for restoring communication [33]. The successful implementation of these systems hinges on the reliable acquisition and interpretation of neural signals, making signal quality and stability paramount concerns in BCI research [65] [28]. This technical guide examines the core challenges and solutions associated with combating noise and ensuring stability in both non-invasive and invasive neural recording methodologies within the context of LIS communication research.

The challenge is particularly acute for non-invasive electroencephalography (EEG)-based systems, where signals are susceptible to significant degradation from physiological, environmental, and motion artifacts [66] [67]. Conversely, invasive intracortical interfaces, while offering superior signal resolution, face challenges related to long-term signal instability, often necessitating frequent recalibration and technical support—a significant barrier to independent home use [28] [4]. This review synthesizes current strategies to overcome these hurdles, detailing noise mitigation techniques for EEG and highlighting advances in stable signal acquisition from invasive local field potentials (LFPs). By framing this discussion within the urgent clinical need for reliable LIS communication tools, this guide aims to equip researchers and developers with the methodologies to build more robust and clinically viable BCI systems.

Noise and Instability: Fundamental Challenges in BCI for LIS

The core challenge in developing BCIs for LIS is establishing a reliable communication channel that is resilient to the various sources of noise and instability inherent in neural recordings. For non-invasive BCIs, particularly EEG, the primary issue is signal degradation. EEG recordings are contaminated by diverse noises and artifacts, which can be categorized as:

• Physiological Noise: Generated by the user's body, including cardiac signals (ECG), muscle contractions (EMG), ocular movements (EOG), and blinking [66] [68].
• Environmental Noise: Arising from external electromagnetic fields emitted by AC power lines, room lighting, and other electronic equipment [66].
• Motion Artifacts: Caused by physical movement of the electrode cables or the interface itself relative to the scalp [66].

These artifacts can obscure the neural signals of interest, complicating their accurate analysis and classification [68]. The signal-to-noise ratio (SNR) is a critical metric, and maximizing it is a central goal of both experimental design and post-processing algorithms [66].

In contrast, invasive BCIs, which involve implanting electrode arrays directly into the cerebral cortex, face a different fundamental challenge: long-term signal instability. While they provide high-resolution signals, the recorded neuronal action potentials (spikes) can be highly unstable across periods of days to months [28]. This instability can result from microscopic movements of the electrode array relative to the brain tissue, biological responses to the implant, or degradation of the electrode materials [28]. This has historically required frequent recalibration of the BCI's decoding algorithms by skilled engineers, preventing independent use and posing a major obstacle for a practical communication system for individuals with LIS [28] [4].

Combatting Noise in Non-Invasive EEG Recordings

Enhancing the SNR in EEG requires a multi-pronged approach, encompassing strategies applied before, during, and after data recording. The following table summarizes the primary sources of noise and their standard countermeasures.

Table 1: Common EEG Noise Sources and Mitigation Strategies

Noise Category	Specific Examples	Mitigation Strategies
Physiological	Ocular (EOG), Cardiac (ECG), Muscle (EMG)	Independent Component Analysis (ICA), Artifact Subspace Reconstruction (ASR), experimental design minimizing movement [66] [68].
Environmental	AC Power Lines, Lighting, Computer Equipment	Use of Faraday cages, shortening electrode cables, replacing AC equipment with DC where possible [66].
Motion Artifacts	Cable movement, unstable electrode-skin contact	Secure cables to cap, ensure low electrode impedances, use dual-layer electrode setups to detect and subtract motion [66].

Advanced Signal Processing Algorithms

Post-recording, mathematical techniques are essential for isolating the neural signal. Several advanced algorithms have been developed:

• Artifact Subspace Reconstruction (ASR): An online, component-based method for removing transient or large-amplitude artifacts. ASR uses statistical anomaly detection in multi-channel data to identify and remove components that introduce a large amount of variance not representative of brain signals [66].
• Independent Component Analysis (ICA): A blind source separation technique that decomposes multi-channel EEG data into statistically independent components. These can be manually or automatically classified as belonging to either neural sources or artifacts (e.g., eye blinks, heartbeats), allowing the artifact components to be removed before signal reconstruction [66] [68].
• Canonical Correlations Analysis (CCA): This method utilizes the autocorrelation within a time series to distinguish brain signals from noise, as brain signals typically have higher autocorrelation values. CCA is particularly useful for building linear models of stimulus-response relationships in event-related potential research and can be implemented in real-time [66].
• Singular Spectrum Analysis (SSA): A non-parametric technique that does not assume linearity or stationarity of the data. SSA decomposes the signal and uses the distribution of eigenvalues of a scaled Hankel matrix to separate the signal subspace from the noise subspace, enabling effective filtering and denoising [68].

Exploiting Noise: Cross-Modal Stochastic Resonance

Intriguingly, recent research explores not only removing noise but also leveraging it to enhance BCI performance through a phenomenon known as stochastic resonance (SR). SR is a counter-intuitive concept where the addition of an optimal level of noise to a non-linear system can enhance the detection of a weak signal [69] [70].

A groundbreaking application is cross-modal SR, where noise in one sensory modality enhances signal processing in another. For example, presenting auditory noise (e.g., Gaussian white noise at -10 dBW) has been shown to significantly enhance the steady-state motion visual evoked potential (SSMVEP) response to visual stimuli [69]. The proposed mechanism involves enhanced functional connectivity and phase synchronization between the auditory and visual cortices, leading to improved classification accuracy and shorter required time windows for BCI control [69]. This approach demonstrates that external auditory noise can be a powerful tool for improving the performance of visual BCIs.

The diagram below illustrates the experimental workflow and proposed neural mechanism for cross-modal stochastic resonance in a BCI paradigm.

Ensuring Long-Term Stability in Invasive Recordings

For invasive BCIs, the key to long-term stability lies in the choice of neural signal features. While action potentials (spikes) offer high spatial and temporal resolution, their instability has driven research into more robust alternatives.

Local Field Potentials (LFPs) for Stable Communication

A significant breakthrough has been the use of local field potentials (LFPs), which represent the summed low-frequency electrical activity from populations of neurons near the recording electrode. LFPs are inherently more stable than spiking activity from individual neurons over long periods [28].

A landmark study demonstrated this stability in two individuals with tetraplegia (one with LIS from brainstem stroke and one with ALS). The participants used an LFP-based BCI for 76 and 138 days, respectively, without a single recalibration of the decoding algorithm [28] [4]. The system enabled them to type messages and write emails at rates of 3.07 and 6.88 correct characters per minute, providing a slow but highly reliable communication channel without requiring continuous technical intervention [28]. This proves that stable, long-term BCI communication is feasible for individuals with LIS.

Comparative Analysis of Neural Signal Features

The table below compares the characteristics of different signal types used in invasive BCIs, highlighting the trade-offs between performance and stability.

Table 2: Comparison of Neural Signal Features for Invasive BCIs

Signal Feature	Description	Advantages	Disadvantages	Stability for Long-Term Use
Single/Multi-Unit Spikes	Action potentials from individual neurons or small groups.	High spatial & temporal resolution; precise control.	Highly unstable over time; requires frequent recalibration.	Low
Local Field Potentials (LFPs)	Low-frequency signals from neuronal populations.	More stable than spikes; suitable for long-term decoding.	Lower spatial resolution; less precise for complex control.	High
Electrocorticography (ECoG)	Signals recorded from the surface of the brain.	Good balance of signal quality and stability; higher resolution than EEG.	Requires craniotomy; coverage limited by grid size.	Medium-High

The Scientist's Toolkit: Key Reagents & Materials

This section details essential components and methodologies referenced in the cited LIS BCI communication research.

Table 3: Research Reagent Solutions for BCI Signal Processing

Item / Technique	Function / Description	Application Context
g.USBamp System (g.tec)	A multi-channel biosignal amplification and data acquisition system.	Used in research for high-quality EEG signal recording [69].
Independent Component Analysis (ICA)	A computational method for separating multivariate signals into additive, statistically independent subcomponents.	Critical for identifying and removing physiological artifacts (EOG, EMG) from EEG data [66] [68].
Artifact Subspace Reconstruction (ASR)	A statistical, component-based algorithm for real-time removal of large-amplitude artifacts from multi-channel EEG.	Useful for online cleaning of EEG data during BCI operation, improving signal quality [66].
96-Channel Intracortical Microelectrode Array (Blackrock Microsystems)	A grid of microelectrodes implanted into the cerebral cortex to record neural activity.	The invasive sensor used in clinical trials (e.g., BrainGate) to record spiking activity and LFPs [28].
Local Field Potential (LFP) Decoder	A translation algorithm that maps low-frequency neural signals to control commands.	Enables stable, long-term BCI communication without frequent recalibration [28] [4].
Stochastic Resonance (SR) Stimulator	A device to deliver controlled levels of auditory or visual noise.	Used to enhance weak neural responses via cross-modal stochastic resonance, improving BCI performance [69] [70].

The path to reliable BCI communication for individuals with locked-in syndrome is being paved by concerted efforts to combat signal noise and ensure long-term stability. In the non-invasive domain, this involves rigorous experimental design coupled with sophisticated signal processing techniques like ICA and ASR, and even the strategic use of cross-modal noise to enhance signals via stochastic resonance. For invasive systems, a paradigm shift towards leveraging stable signal features like local field potentials has demonstrated that independent, long-term communication is a tangible reality. Future research must continue to bridge the gap between technical performance and practical clinical application, focusing on standardized protocols, adaptive algorithms, and robust hardware that can function reliably in the home environment. By prioritizing signal quality and stability, researchers can unlock the full transformative potential of BCIs, finally restoring the fundamental human need to communicate for those living with LIS.

Effective communication for individuals with Locked-In Syndrome (LIS) using Brain-Computer Interfaces (BCIs) depends critically on accurately assessing the user's state of consciousness and cognitive readiness. This technical guide synthesizes current research on electrophysiological monitoring techniques to evaluate consciousness levels and cognitive states in LIS and Complete Locked-In Syndrome (CLIS) patients. We present standardized frameworks for quantifying consciousness through multi-modal EEG analysis, detail experimental protocols for assessing communication readiness, and provide benchmarking standards for BCI performance evaluation. Within the broader thesis of LIS BCI communication research, this work establishes that pre-communication assessment of user state is fundamental to achieving reliable interaction, particularly given the absence of behavioral ground truth in CLIS patients. The protocols and metrics detailed herein aim to enable researchers to determine optimal communication windows and maximize BCI efficacy.

Locked-In Syndrome (LIS) is a severe neurological condition characterized by preserved consciousness and cognitive capacity alongside complete paralysis of nearly all voluntary muscles, typically resulting from brainstem lesions [25]. Patients in a Complete Locked-In State (CLIS) lack even residual eye movement control, eliminating all voluntary communication channels and making it impossible to behaviorally ascertain their state of consciousness at any given moment [25]. This creates a fundamental challenge for BCI-mediated communication: without knowledge of whether a patient is conscious and cognitively ready, communication attempts may fail unpredictably.

The broader thesis of LIS BCI research must therefore encompass not only the development of communication modalities but also methods to pre-assess the user's state. Research indicates that LIS patients retain consciousness and cognitive functions, but their ability to engage with BCIs fluctuates [25]. The ability to interact with their environment significantly enhances wellbeing and quality of life, making reliable communication paramount [25]. This paper establishes a technical framework for assessing consciousness levels and cognitive readiness to identify optimal communication windows, thereby addressing a critical prerequisite for effective BCI interaction.

Quantitative Framework for Consciousness Assessment

Consciousness assessment in LIS requires a multi-dimensional approach, as no single metric reliably captures its complex nature. The following quantitative framework integrates complementary electrophysiological measures to maximize the probability of correctly determining a patient's state.

Normalized Consciousness Level (NCL) Framework

A recent approach analyzes EEG data from LIS patients to derive a Normalized Consciousness Level (NCL), representing a scale from 0 to 1, where 1 indicates a high likelihood of full consciousness and 0 indicates its likely absence [25]. This framework acknowledges the absence of ground truth in CLIS and integrates multiple EEG features to approximate consciousness probability.

Table 1: Primary Quantitative Metrics for Consciousness Assessment in LIS

Metric Category	Specific Metrics	Conscious State Indicators	Application in LIS/CLIS
Spectral Analysis	Power across frequency bands (Delta, Theta, Alpha, Beta, Gamma)	Increased alpha/theta power, specific beta/gamma patterns	Differentiates states across patients; shows inter-patient variability [25]
Complexity Measures	Lempel-Ziv Complexity (LZC), Perturbational Complexity Index (PCI)	Higher complexity values comparable to healthy controls	Distinguishes conscious from unconscious states; LIS patients show PCI values similar to healthy controls [25]
Connectivity Measures	Spectral coherence, functional connectivity networks	Strong, integrated fronto-parietal connectivity	Reveals preserved network connectivity in conscious patients misdiagnosed as unresponsive [25]
Event-Related Potentials (ERPs)	P300 amplitude and latency, MMN (Mismatch Negativity)	Robust P300 responses	Present in all LIS patients; predicts recovery; used in command-following paradigms [25]

Information Transfer Rate as a Performance Benchmark

Beyond consciousness assessment, cognitive readiness for communication can be quantified using information transfer rates. Recent benchmarking establishes that high-performing BCIs should achieve information transfer rates exceeding 200 bits per second with minimal latency (<56ms) to support fluid communication [71]. These engineering metrics correlate with practical usability, as rates above ~40 bps (the approximate rate of transcribed human speech) enable realistic conversation-paced interfaces [71].

Table 2: BCI Performance Benchmarks for Communication Readiness

BCI System Type	Reported Information Transfer Rate	Typical Latency	Sufficiency for Real-Time Communication
High-Performance Intracortical (Connexus BCI)	200+ bps [71]	56 ms	Excellent (exceeds speech rate)
Standard Intracortical (e.g., Neuralink, Utah Array)	~10 bps [71]	Variable	Limited (slower than speech)
Endovascular (e.g., Synchron)	~1-2 bps [71]	Variable	Poor (requires significant compromises)
Non-Invasive (EEG-based P300)	Not explicitly reported, but typically <1 bps inferred	High	Very Poor (requires structured interfaces)

Experimental Protocols for Assessing Cognitive Readiness

Standardized experimental protocols are essential for reliably evaluating a patient's readiness to engage with a BCI. The following methodologies are validated in LIS and related disorders of consciousness.

Resting-State EEG Protocol

Objective: To assess baseline consciousness levels without requiring task performance, making it suitable for CLIS patients.

Methodology:

• Recording Parameters: Record EEG from a minimum of 16 channels (international 10-20 system) for at least 5 minutes in a quiet, dimly lit room [25].
• Patient State: Ensure the patient is in a resting but awake state, confirmed via monitoring of physiological signals if possible.
• Data Extraction: Calculate the following features from pre-processed (artifact-removed) EEG signals:
  • Spectral Power: Compute power spectral density in standard frequency bands.
  • Complexity: Calculate Lempel-Ziv Complexity (LZC) to measure signal entropy and irregularity.
  • Connectivity: Determine functional connectivity between frontal, parietal, and occipital regions using coherence or phase-based measures [25].
• Analysis: Integrate features into a composite NCL score using machine learning classifiers (e.g., support vector machines) trained on data from healthy controls and patients with established diagnoses [25].

Active Paradigm Protocols

Objective: To engage residual cognitive functions and detect command-following, providing direct evidence of consciousness.

P300 Auditory Oddball Protocol:

• Stimuli: Present a sequence of frequent standard tones and rare target tones (85/15 ratio) via headphones.
• Instruction: Inform the patient to mentally count the target tones.
• Recording: Time-lock EEG recording to stimulus onset.
• Analysis: Average responses to target versus standard tones; look for P300 component (positive deflection ~300ms post-stimulus) at parietal electrodes [25].

Motor Imagery Protocol:

• Instruction: Ask the patient to imagine specific motor acts (e.g., right hand versus left hand movement) in response to cues.
• Recording: Monitor EEG during imagery periods.
• Analysis: Perform spectral analysis to detect event-related desynchronization (ERD) in the mu/beta rhythms over sensorimotor cortex contralateral to the imagined movement [25].

Visual Imagery Protocol (Emerging):

• Cueing: Present an auditory cue (e.g., "imagine a face") to prompt spontaneous visual imagery from long-term memory, avoiding visual working memory confounds [72].
• Recording: Monitor EEG with emphasis on occipital and frontal electrodes.
• Analysis: Employ pattern classification to distinguish between brain activity patterns associated with different imagery categories; focus on gamma band (30-100 Hz) activity and frontal electrode contributions [72].

Hybrid BCI-ET Transition Readiness Protocol

Objective: To support patients transitioning from LIS (with residual eye control) to CLIS by combining Eye-Tracking (ET) and BCI modalities.

Methodology:

• Setup: Simultaneously record EEG and gaze data.
• Task: Implement a modified P300-based BCI where items are selected both by gaze (ET) and brain response (BCI).
• Fusion Algorithm: Process gaze and EEG data in real-time using a fusion algorithm that weights the reliability of each modality.
• Output: Use the combined signal to enhance detection of user intention, maintaining high accuracy as oculomotor function declines [23].

Technical Implementation and Benchmarking

Implementation of user-state monitoring requires rigorous attention to signal acquisition, processing, and performance benchmarking to ensure reliable results.

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Materials and Tools for User-State Monitoring Research

Item / Tool	Function / Purpose	Implementation Example
High-Density EEG System	Recording electrophysiological activity with sufficient spatial resolution.	64+ channel systems with active electrodes for improved signal quality in resting-state and ERP studies [25].
EMG/EOG Monitoring	Detecting and eliminating artifacts from residual muscle activity or eye movements.	Simultaneous recording during EEG to identify and reject contaminated epochs [25].
Auditory Stimulator	Presenting controlled auditory stimuli for P300 and other ERP paradigms.	Precision audio delivery system with calibrated timing for event-related potential studies [25].
Visual Presentation System	Displaying visual cues for imagery tasks or visual ERP paradigms.	High-refresh-rate monitor with millisecond-accurate timing for stimulus presentation [72].
SONIC Benchmarking Suite	Standardized assessment of BCI information transfer rate and latency.	Open benchmarking standard to measure true information transfer with delay accounting; enables cross-platform performance comparison [71].
Lempel-Ziv Complexity Algorithm	Quantifying signal complexity as a proxy for conscious processing.	Algorithm implementation for calculating LZC from pre-processed EEG signals as part of NCL framework [25].
Connectivity Analysis Toolbox	Computing functional connectivity metrics between brain regions.	Software for calculating spectral coherence, phase locking value, and other connectivity measures from multi-channel EEG [25].

Signal Processing and Analysis Workflow

Raw electrophysiological data requires sophisticated processing to extract meaningful features related to consciousness and cognitive state.

SONIC Benchmarking Implementation

The Standard for Optimizing Neural Interface Capacity (SONIC) provides a rigorous framework for measuring BCI performance, which is critical for assessing communication capability.

Protocol:

• Stimulation: Present controlled sequences of sensory stimuli (e.g., five-tone sequences mapped to characters) to an animal model or human subject.
• Recording: Use the fully implanted BCI to record neural activity from relevant cortical areas (e.g., auditory cortex for sound stimuli).
• Decoding: Implement decoding algorithms to predict which stimuli were presented based solely on neural activity.
• Calculation: Compute the mutual information between presented and predicted stimuli to obtain a true measure of information transfer rate (bits per second) while accounting for latency [71].

Discussion and Future Directions

Integrating user-state monitoring into BCI communication systems represents a paradigm shift in assistive technology for LIS/CLIS patients. The multi-modal assessment framework presented here enables researchers to determine optimal communication windows, potentially increasing success rates and reducing user frustration.

Future research should focus on:

• Real-Time Implementation: Developing embedded systems that can continuously monitor consciousness levels and automatically initiate communication when optimal states are detected.
• Adaptive Hybrid BCIs: Creating systems that seamlessly transition between ET, ERP-based BCIs, and motor imagery based on real-time assessment of user capabilities [23].
• Standardized Benchmarking: Widespread adoption of rigorous benchmarks like SONIC to enable objective comparison across BCI platforms and accelerate performance improvements [71].
• Visual Imagery Refinement: Further investigation into spontaneous visual imagery from long-term memory as a more intuitive BCI control strategy, particularly by focusing on gamma-band activity and frontal electrode contributions [72].

The path forward requires close collaboration between clinical researchers, neural engineers, and signal processing experts to translate these assessment protocols into robust, clinically deployed systems that restore communication capacity to those most severely isolated by neurological conditions.

Complete Locked-In Syndrome (CLIS) represents the most profound form of motor paralysis, where individuals lose all voluntary muscle control, including eye movement, rendering them without any reliable means of communication despite often retaining intact cognitive and emotional processing [73] [74]. This condition typically results from progressive neurological diseases such as Amyotrophic Lateral Sclerosis (ALS) [75]. The central challenge in CLIS management and research lies in the paradoxical combination of apparent unconsciousness due to behavioral unresponsiveness with potentially preserved consciousness and cognitive function [76].

The development of Brain-Computer Interface (BCI) systems aims to bridge this communication gap by measuring and interpreting brain activity directly. However, establishing reliable communication has proven exceptionally difficult for patients in CLIS compared to those in classical Locked-In Syndrome (LIS), where some residual eye movement remains [74] [34]. This paper examines the theoretical framework of "goal-directed thought extinction" in CLIS, explores the effects of long-term paralysis on neuroplasticity, and evaluates technological innovations that show promise for restoring communication channels.

Theoretical Framework: The Challenge of Goal-Directed Cognition

The Extinction Hypothesis

A leading theoretical explanation for the difficulty in establishing BCI communication with CLIS patients proposes the "extinction of goal-directed cognition and thought" following complete paralysis [75]. This hypothesis suggests that the persistent inability to execute motor commands or observe their environmental consequences may fundamentally disrupt the neural mechanisms supporting voluntary, goal-directed thinking over time.

This theoretical account finds support in historical experimental psychology. Studies attempting to demonstrate instrumental learning in curarized (paralyzed) rats failed to achieve replication, suggesting that "intact or partially intact motor functions and somatic-motor system mediation of autonomic functions is a mandatory requirement for instrumental learning and control of physiological functions" [75]. The critical implication for CLIS is that BCIs requiring explicit voluntary control of brain signals—such as consciously modulating sensorimotor rhythms or slow cortical potentials—may be inherently unsuitable for patients who have transitioned into complete locked-in states.

Neuroplasticity and Language Attrition

The neural interdependence of language and motor systems presents an additional challenge. According to the sensorimotor approach to cognition, neural networks for linguistic tasks significantly overlap with those for movement and motor actions [76]. The "action perception theory" posits that during linguistic tasks, sensorimotor circuits become active, potentially lowering thresholds for subsequent activations [76].

In CLIS, the long-term "non-use" of motor semantic networks combined with absent proprioceptive input may increase excitation thresholds for semantic nodes related to action concepts. Research has demonstrated that BCI performance in CLIS patients can be affected by the semantic content of presented questions, with sentences containing action-related words or phrases showing different communication success rates compared to those focused on objects [76]. This suggests that complete immobility may lead to a form of language attrition specifically affecting action-related concepts, representing a significant neuroplastic adaptation to prolonged paralysis.

BCI Modalities and Performance in CLIS

Electroencephalography (EEG) Based Approaches

EEG-based BCIs have shown variable success in LIS but face significant challenges in CLIS:

Vibro-Tactile P300 Systems: The mindBEAGLE system utilizing vibro-tactile stimulation demonstrated promising results, with two out of three CLIS patients achieving 70% and 90% communication accuracy using a three-stimulator paradigm (VT3) placed on the left wrist, right wrist, and shoulder [73]. Patients mentally counted stimuli on the target hand to elicit a P300 response, allowing for binary communication. Healthy controls achieved 93% accuracy with the same paradigm, indicating the approach's validity [73].

Auditory P300 Systems: Studies using auditory oddball paradigms with spoken words "yes" and "no" delivered to different ears have shown limited success with CLIS patients [34]. While healthy subjects achieved average online accuracy of 86%, most patients could not perform above chance level, with only two ALS patients (not yet in CLIS) achieving 100% accuracy [34].

Motor Imagery Systems: BCIs based on imagining hand movements have shown poor efficacy in CLIS, with only 3 out of 12 LIS/CLIS patients achieving communication capability using this paradigm [73]. This supports the theoretical framework that explicitly volitional motor imagery may be compromised in complete paralysis.

Table 1: Performance Comparison of Non-Invasive BCI Modalities in CLIS

BCI Modality	Mechanism	CLIS Patient Performance	Healthy Control Performance	Key Studies
Vibro-Tactile P300	Counting tactile stimuli to elicit P300 response	2/3 patients achieved 70-90% accuracy	93% accuracy	Guger et al., 2017 [73]
Auditory P300	Attention to deviant auditory stimuli	Most patients at chance level; rare successes	86% accuracy	Combaz et al., 2024 [34]
Motor Imagery	ERD/ERS from imagined movements	Limited efficacy; only 3/12 LIS/CLIS patients could communicate	73% accuracy	Guger et al., 2017 [73]
fNIRS	Hemodynamic response to automatic answers	70% accuracy in binary questions	N/A	Chaudhary et al., 2017 [75]

Functional Near-Infrared Spectroscopy (fNIRS)

fNIRS measures hemodynamic responses associated with neuronal activity and has emerged as a promising alternative for CLIS communication. In a landmark study, four patients with advanced ALS—two in permanent CLIS and two entering CLIS—learned to answer personal questions using frontocentral oxygenation changes measured with fNIRS [75].

The approach exploited classical conditioning rather than voluntary control, using overlearned questions ("Berlin is the capital of France") and personal questions ("Your husband's name is Joachim") to elicit automatic "yes" or "no" responses. Online classification resulted in above-chance-level correct response rates over 70% [75]. This success with an implicit processing procedure aligns with the theoretical framework suggesting that classical conditioning might circumvent the volitional effort deficits in CLIS.

Invasive Approaches

Intracortical implants have demonstrated potential for CLIS communication, though with limitations. In one case study, a CLIS patient received an implanted electrode in the dominant left motor cortex [34]. Initially, when instructed to attempt or imagine movements, no cortical response could be detected. Reliable yes-no responses were only obtained three months after implantation using a neurofeedback protocol, eventually achieving 86.6% accuracy over 5700 trials [34]. This extended training period supports the concept that establishing communication in CLIS requires either alternative approaches (like implicit processing) or extensive retraining protocols.

Experimental Protocols and Methodologies

fNIRS-Based Communication Protocol

The successful fNIRS protocol involved several key methodological components [75]:

• Signal Acquisition: fNIRS systems measured oxygenation changes in frontocentral regions using near-infrared light, typically at two or more wavelengths to distinguish oxygenated and deoxygenated hemoglobin.

• Stimulus Presentation: Patients were presented with 10 questions with known "yes" answers and 10 questions with known "no" answers randomly during each session. Questions tapped into overlearned knowledge or personal information.

• Classification Approach: A linear Support Vector Machine (SVM) classifier was trained on the hemodynamic responses to known questions, then applied to classify answers to open questions where the answer wasn't known to the experimenters.

• Validation: Performance was validated through questions where answers could be verified (personal questions with known answers) before proceeding to open questions.

Figure 1: fNIRS-Based Communication Protocol Workflow

Vibro-Tactile P300 BCI Protocol

The mindBEAGLE system implemented a structured approach for CLIS assessment and communication [73]:

• Stimulator Placement: For VT3 mode, three vibro-tactile stimulators were placed on the left wrist, right wrist, and shoulder. The shoulder stimulator served as a distractor.

• Paradigm Explanation: Instructions were provided to patients to mentally count stimuli occurring on either the left or right wrist, corresponding to "yes" or "no" responses.

• Signal Processing: EEG was recorded from Fz, C3, Cz, C4, CP1, CPz, CP2, and Pz electrodes at 256 Hz sampling rate, filtered between 0.1 and 30 Hz.

• P300 Detection: The system detected P300 event-related potentials time-locked to the target stimuli, which occurred less frequently than non-target stimuli.

• Classification: Machine learning algorithms classified intended responses based on the presence or absence of P300 responses to specific stimulators.

• Validation: System accuracy was tested with known questions before proceeding to open communication.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 2: Essential Research Materials for CLIS BCI Research

Tool/Technology	Function/Application	Specific Examples/Models	Key Considerations
EEG Systems	Recording electrical brain activity	g.USBamp biosignal amplifier with 16 channels	Active electrodes (g.LADYbird); 0.1-30 Hz filtering [73]
fNIRS Systems	Measuring hemodynamic responses	Continuous wave fNIRS systems	Frontocentral placement; multiple wavelengths [75]
Tactile Stimulators	Delivering vibro-tactile stimuli	Compact vibration motors	Precise timing control; placement on wrists/shoulder [73]
Auditory Stimulation	Presenting auditory paradigms	In-ear headphones; synthesized speech	Sound duration manipulation (100ms vs 150ms) [34]
Classification Algorithms	Interpreting brain signals	Linear Support Vector Machine (SVM)	Individual calibration; adaptive learning [75]
Experimental Software	Paradigm design and data collection	mindBEAGLE software; custom BCI platforms	Flexible parameter adjustment; real-time feedback [73]

The challenge of addressing neuroplasticity and maintaining goal-directed thought in CLIS requires a multifaceted approach that acknowledges both the theoretical implications of long-term paralysis and the practical limitations of current BCI technologies. The evidence suggests that:

• Paradigm Shift Needed: Explicit volitional control paradigms (motor imagery, intentional P300 generation) show limited efficacy in CLIS, supporting the "goal-directed cognition extinction" hypothesis.

• Implicit Approaches Show Promise: fNIRS systems utilizing automatic processing of overlearned or personal questions demonstrate that communication may be possible through classical conditioning rather than voluntary control.

• Sensory Modality Matters: Vibro-tactile and auditory approaches offer alternatives for patients who may have visual impairments or difficulties with gaze-independent visual BCIs.

• Semantic Content Affects Performance: The semantic content of communication attempts influences success rates, suggesting domain-specific neuroplastic changes in CLIS that should inform future BCI design.

Future research should focus on developing increasingly implicit BCI approaches, optimizing paradigms based on semantic content analysis, and establishing standardized assessment protocols that can distinguish between different levels of cognitive preservation in CLIS. The ultimate goal remains the abolition of complete locked-in states through technological innovation that adapts to the unique neuroplastic challenges of prolonged complete paralysis.

For individuals with locked-in syndrome (LIS), a severe neurological condition characterized by complete paralysis of nearly all voluntary muscles while cognitive function remains intact, Brain-Computer Interfaces (BCIs) represent a potentially transformative communication solution [25]. The transition of BCI technology from controlled laboratory environments to unsupervised home use presents a complex set of engineering, clinical, and human-factors challenges. Successfully overcoming these hurdles is critical for restoring communication capabilities, which has been shown to significantly enhance the wellbeing and quality of life for LIS patients [25]. This technical guide examines the core requirements for viable home-use BCIs, including robust wireless operation, simplified user interfaces, and automated signal processing systems that can function reliably without technical supervision. The ultimate goal is to develop systems that are not only technologically sophisticated but also practical, accessible, and sustainable for long-term daily use by non-technical users in home environments.

Technical Hurdles in BCI Translation for LIS Communication

Signal Acquisition and Processing Constraints

The development of home-deployable BCIs requires balancing performance with practical constraints. Electroencephalography (EEG) remains the most common non-invasive technique, now enhanced by high-density arrays, dry electrodes, and artifact-reduction algorithms [77]. However, non-invasive neural signals are highly susceptible to various noise sources, including muscle activity, eye movements, and environmental electrical interference [77]. For LIS patients, who may progress to complete LIS (CLIS) with no residual muscle movement, the challenge is further compounded by the inability to ascertain consciousness at specific time points, creating uncertainty about optimal communication periods [25].

Advanced signal processing approaches must integrate multiple features based on frequency, complexity, and connectivity measures to maximize the probability of correctly determining patients' actual states given the absence of ground truth for consciousness in CLIS patients [25]. These systems must accomplish this with limited computational resources suitable for home deployment while maintaining sufficient temporal resolution to capture critical neural markers like the P300 event-related potential, which has been shown to be present in all LIS patients and may predict recovery outcomes [25].

Wireless Data Transmission and Power Management

Modern implanted BCI systems generate enormous data volumes, sampling from thousands of electrodes thousands of times per second [78]. This creates significant challenges for wireless transmission within the power constraints of implantable devices. Companies are developing various approaches to reduce or compress this data to accommodate existing wireless links [78]. The choice of transmission protocol must balance bandwidth requirements with power consumption, ensuring reasonable battery life while maintaining the fidelity of neural data. For non-invasive systems, wireless operation enables greater mobility and comfort but faces similar constraints in data throughput and battery longevity. Effective power management strategies, including inductive charging and low-power sleep modes, are essential components of any home-use BCI system.

Table: Technical Specifications of Current Commercial BCI Systems

Company/Device	Implantation Type	Electrode Count	Wireless Capability	Key Features
Neuralink [78]	Intracortical	1,000+	Yes	Fully implantable, wireless
Paradromics Connexus [79]	Intracortical	421	Yes	Fully wireless, focused on healthcare
Precision Layer 7 [79]	Surface (non-penetrating)	Multiple electrodes	Yes	Minimally invasive, safely removable
Synchron Stentrode [79]	Endovascular (via blood vessels)	Not specified	Yes	No open-brain surgery required

Usability and Accessibility Barriers

For LIS patients with no technical background or assistance, BCI systems must feature intuitive interfaces and minimal setup requirements. Research indicates that current systems often require substantial training for both patients and caregivers, creating significant adoption barriers [77]. The transition to CLIS presents additional challenges, as patients lose even the limited communication channels (like eye blinking) they previously relied upon [25]. Systems must therefore incorporate adaptive algorithms that can accommodate progressive neurological conditions and fluctuating cognitive states without requiring constant technical recalibration.

Current Wireless BCI Architectures and Home Deployment Strategies

Invasive and Minimally Invasive Approaches

Recent advances in fully implantable wireless BCI systems show significant promise for home deployment. Paradromics has developed the Connexus BCI, a dime-sized implant with 421 microelectrodes that features a patented on-chip processing system to record massive amounts of data for natural-speed communication [79]. This system is specifically engineered for long-term durability, with data showing no change in signal-to-noise ratio over two years—a critical consideration for implants intended to last at least a decade without replacement [79].

Precision Neuroscience has pursued a different architectural approach with its Layer 7 Cortical Interface, described as resembling "a piece of Scotch tape with tiny platinum electrodes" that conform gently to the brain's surface without penetrating it [79]. This minimally invasive design offers advantages in signal stability and reduced inflammation risk while recently achieving the significant milestone of FDA 510(k) clearance for implantation for up to 30 days [79]. The company has performed 38 human implants for short-term clinical and diagnostic applications, demonstrating a viable pathway toward home use [79].

Endovascular and Non-Invasive Alternatives

Synchron has developed a fundamentally different implantation approach through its stentrode system, which is delivered via blood vessels without open-brain surgery [79]. A permanently implanted stent-like device is inserted through the jugular vein into the brain's motor cortex, where the BrainOS platform collects signals from a receiver implanted in the user's chest [79]. This approach leverages the blood vessels as a stable biological environment for the electrodes, providing a natural barrier to the "saltwater bath of the brain" [79]. Such systems represent a compelling balance between signal quality and reduced surgical risk, potentially accelerating the transition to home use.

For non-invasive applications, Spanish company INBRAIN Neuroelectronics is pioneering the use of graphene—an exceptionally thin, flexible, and biocompatible material—to create a new breed of implantable devices [79]. Graphene's physical properties enable injection of 200 times more charge at very low impedance compared to traditional platinum electrodes, potentially overcoming the poor outcome reproducibility that has plagued other BCI systems [79].

Table: Comparison of BCI Signal Acquisition Methods for LIS Applications

Method	Spatial Resolution	Temporal Resolution	Key Applications in LIS	Limitations for Home Use
EEG [25] [77]	Low (scalp level)	High (milliseconds)	Consciousness assessment, P300 spellers, SSVEP systems	Susceptible to noise, limited spatial resolution
ECoG [25]	Intermediate (cortical surface)	High (milliseconds)	High-quality signal for communication systems	Requires surgical implantation
Intracortical Microelectrodes [25] [79]	High (single neuron)	High (milliseconds)	Direct neural control, speech decoding	Invasive, potential for signal drift over time
Endovascular Stentrode [79]	Intermediate	High	Motor command decoding, device control	Relatively new approach, limited long-term data

Experimental Protocols for BCI Evaluation in LIS

Consciousness Assessment Protocol for CLIS Patients

A critical challenge in communicating with complete LIS (CLIS) patients is determining the optimal time for interaction, as there are no behavioral indicators of consciousness [25]. Researchers have developed EEG-based protocols to assess consciousness levels through a multi-feature approach:

• Signal Acquisition: EEG data is recorded using standard clinical systems with appropriate electrode placements based on international standards [25].
• Feature Extraction: Multiple features are extracted from pre-processed EEG signals, including frequency bands (alpha, beta, theta, gamma), complexity measures (Lempel-Ziv complexity), and functional connectivity metrics between brain regions [25].
• Consciousness Level Calculation: Features are integrated to compute a normalized consciousness level (NCL) representing a scale from 0 to 1, where 1 indicates a high likelihood of full consciousness [25].
• Validation: While no ground truth exists for CLIS consciousness, results are compared with periodic communication attempts and any residual behavioral responses when possible [25].

This approach has demonstrated effectiveness in detecting neural markers of consciousness and differentiating between states across most patients, despite considerable inter-patient variability [25].

Speech Decoding and Communication Protocols

Recent advances in speech neuroprosthetics have demonstrated remarkable decoding accuracies. The UC Davis Health team developed a BCI that translates brain signals into speech with up to 97% accuracy—the most accurate system of its kind reported [80]. Their protocol involves:

• Surgical Implantation: Placement of electrode arrays in speech-related areas of the motor cortex [80].
• Training Data Collection: Recording neural activity as participants attempt to speak or imagine speaking words and sentences [80] [81].
• Algorithm Training: Using deep learning models to map neural patterns to intended speech sounds or words [24].
• Real-Time Decoding: Implementing optimized algorithms for low-latency speech synthesis during communication attempts [24].

In a notable study, researchers created a streaming brain-to-voice neuroprosthesis that decoded speech at rates of 47.5 words per minute for a 1,000-word vocabulary and 90.9 words per minute for a 50-word vocabulary, with less than 80 milliseconds of latency [24]. This near-synchronous voice streaming represents a significant advancement toward naturalistic conversation for LIS patients.

BCI Communication Workflow for LIS

Motor Imagery and Visual Interface Protocols

For LIS patients without speech capability, alternative communication protocols rely on motor imagery or visual evoked potentials:

• P300 Speller Systems: Utilizing the P300 event-related potential elicited when a desired item flashes among a matrix of choices [25] [77].
• Steady-State Visual Evoked Potential (SSVEP): Employing visual stimuli flickering at specific frequencies to enable communication through frequency-tagged EEG responses [25] [77].
• Motor Imagery Training: Training patients to modulate sensorimotor rhythms through imagined movements, which can be translated into control signals [25].

Each approach requires specific experimental protocols for calibration and validation, with considerations for patient fatigue, cognitive load, and individual variability in signal quality.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table: Essential Research Materials for BCI Development in LIS Applications

Item/Reagent	Specification/Type	Primary Function in BCI Research
EEG Systems [25]	High-density wet or dry electrode systems	Recording electrical brain activity for consciousness assessment and communication
Implantable Electrode Arrays [79]	Utah arrays, microelectrode grids, stent-based electrodes	High-resolution neural signal acquisition for invasive BCIs
Signal Processing Software [25]	MATLAB, Python with specialized toolboxes (EEGLAB, MNE)	Preprocessing, feature extraction, and decoding of neural signals
Deep Learning Frameworks [24]	TensorFlow, PyTorch with custom architectures	Training neural decoders for speech and command recognition
Wireless Data Transmission Modules [78]	Custom low-power RF or Bluetooth systems	Transmitting neural data from implanted or wearable systems
Stimulus Presentation Software [25]	Presentation, Psychtoolbox, custom applications	Delivering visual/auditory stimuli for evoked potential studies
Neurofeedback Interfaces [82]	Customizable visual feedback systems	Providing real-time feedback to users during BCI training

Implementation Framework and Future Directions

The "NEURO" framework provides a structured approach to guide the clinical translation of BCIs into practical home use [77]:

• N - Needs-focused: Address specific communication challenges faced by LIS patients
• E - Evidence-based: Rigorously validate systems through controlled trials
• U - User-centered: Design interfaces accessible to patients with severe motor limitations
• R - Regulatory-ready: Adhere to FDA and other regulatory requirements throughout development
• O - Outcome-oriented: Focus on meaningful functional communication outcomes

Future development priorities include creating fully closed-loop systems that automatically adapt to user state and signal quality, developing more efficient neural decoding algorithms with lower computational requirements, and establishing standardized validation protocols specifically for home-use BCI systems [77]. As these technologies mature, the focus must remain on creating systems that are not only technologically advanced but also practical, reliable, and accessible for daily use by LIS patients in home environments.

Wireless BCI System Architecture

Brain-Computer Interface (BCI) research for Locked-In Syndrome (LIS) represents one of the most technologically and ethically challenging frontiers in modern neuroscience. Patients with LIS, often resulting from Amyotrophic Lateral Sclerosis (ALS), brainstem stroke, or traumatic brain injury, retain full consciousness and cognitive function but lose virtually all voluntary muscle control, including the ability to speak or move [25]. The emergence of BCI technologies, both invasive and non-invasive, has created unprecedented opportunities to restore communication channels for this population. However, these advances bring complex ethical considerations regarding data privacy, brain data ownership, and informed consent that must be addressed within the specific clinical context of LIS [83] [84]. This whitepaper examines these ethical dimensions through the lens of current research, providing technical and methodological guidance for researchers and clinicians working at this intersection.

Technical Foundations of BCI for LIS Communication

BCI Modalities and Implementation Approaches

BCI systems for LIS communication employ various signal acquisition methods, each with distinct technical characteristics and implementation considerations. These systems can be broadly categorized into invasive and non-invasive approaches, with the selection dependent on the patient's specific condition, progression stage, and technical infrastructure availability.

Table: BCI Modalities for LIS Communication

Modality	Signal Source	Spatial Resolution	Temporal Resolution	Primary Applications in LIS	Key Limitations
Invasive (Intracortical)	Microelectrode arrays implanted in motor cortex	Very High (individual neurons)	Very High (spike timing)	Communication spelling (Yes/No paradigms) [56]	Surgical risk, signal stability over time, limited coverage
ECoG	Electrode grids on cortical surface	High (local field potentials)	High	Speech decoding, communication control [85]	Requires craniotomy, limited to cortical surface
EEG	Scalp electrodes	Low (smearing effect)	High	P300 speller, SSVEP, motor imagery [23] [25]	Susceptible to noise, limited bandwidth, requires preparation
Hybrid Systems	Combination of ET and EEG [23]	Variable	Variable	Transitional support as oculomotor function declines [23]	Increased system complexity, data fusion challenges

Core Experimental Protocols in LIS BCI Research

Research involving BCI communication for LIS patients follows specific methodological protocols tailored to address the unique challenges of this population. The following experimental workflows represent current best practices in the field.

Protocol 1: Intracortical BCI for Complete Locked-In State (CLIS)

Diagram: Intracortical BCI Communication Workflow for CLIS

This protocol, demonstrated successfully with a completely locked-in individual, enables communication through implanted microelectrode arrays that record neural signals from the motor cortex [56]. The participant learns to generate brain activity by attempting different movements, which are then decoded by machine learning models in real time. The system maps signals to 'yes' or 'no' responses, allowing letter-by-letter spelling through an auditory interface that reads letters aloud, with the participant confirming or rejecting each selection through neural responses. This approach has maintained communication capability over periods exceeding two years in home environments.

Protocol 2: Hybrid Eye-Tracking and BCI Transition System

Diagram: Hybrid BCI System for Progressive Transition

This approach addresses the challenging transition from LIS to Complete Locked-In State (CLIS) as oculomotor function declines [23]. The hybrid framework combines eye-tracking (ET) and EEG-based BCI modalities, processing both gaze and EEG data in real time. A specialized ET-BCI fusion algorithm enhances detection of user intention, maintaining communication accuracy even as users gradually shift between primary modalities. This is particularly valuable for ALS patients experiencing progressive loss of voluntary motor control, where eye-tracking systems become ineffective as the disease advances.

Essential Research Reagents and Materials

Table: Key Research Reagents and Materials for LIS BCI Research

Item/Category	Technical Specification	Research Function	Example Implementation
Microelectrode Arrays	3.2mm × 3.2mm, 64 needle-like electrodes [56]	Records neural signals from motor cortex	Intracortical BCI for CLIS patients
ECoG Grids	Surface electrodes, typically 64-256 contacts [85]	Records cortical surface potentials	Speech decoding paradigms
EEG Systems	16-256 channels, dry or wet electrodes	Non-invasive neural signal acquisition	P300 speller, SSVEP communication
Eye-Tracking Systems	Infrared video-oculography, >30Hz sampling	Gaze point detection for hybrid BCIs	Transitional interfaces for declining motor function
Signal Processing Libraries	Python (MNE, Scikit-learn), MATLAB	Preprocessing, feature extraction, classification	Real-time intention decoding
Machine Learning Models	SVM, LDA, Deep Learning networks	Classification of neural signals	Yes/No intent classification from motor imagery
Auditory Feedback Systems	Stereo headphones, text-to-speech engines	Provides neurofeedback for completely locked-in users	Auditory speller interfaces

Ethical Framework for LIS BCI Research

Brain Data Privacy and Security Considerations

The intimate nature of neural data raises significant privacy concerns, particularly for LIS patients who may lack alternative communication channels. Neurodata derived from BCI systems can reveal thoughts, intentions, emotional states, and potentially private information that individuals cannot otherwise express [83] [86]. For LIS patients, whose communication is exclusively mediated through BCI systems, the privacy stakes are exceptionally high.

Table: Brain Data Privacy Risks and Mitigation Strategies

Risk Category	Specific Manifestations in LIS Research	Potential Harm	Recommended Mitigations
Unauthorized Access	hacking of implanted devices, interception of neural signals [87]	manipulation of communication, extraction of private thoughts	strong encryption, secure authentication, regular security updates
Data Misuse	repurposing neural data for neuromarketing, employment decisions, or insurance assessments [83] [86]	discrimination, exploitation, loss of autonomy	data minimization, strict usage limitations, ethical review boards
Inference Risks	deriving sensitive information (cognitive decline, emotional states) from neural patterns [83]	psychological distress, stigmatization, altered care relationships	limited data retention, algorithmic transparency, participant education
Re-identification	combining "anonymized" neural data with other datasets to re-identify individuals [83]	loss of privacy, unintended disclosure of health status	robust de-identification, controlled data sharing, contractual limitations

Statistical insights into privacy attitudes reveal that approximately 62% of adults would be uncomfortable with neural data being used for advertising without explicit consent, while 78% support strict consent controls before neural data can be used for research or product development [83]. These findings underscore the importance of robust privacy protections in BCI research, particularly for vulnerable populations like LIS patients.

Brain Data Ownership and Governance

The question of who owns brain data generated through BCI systems remains legally ambiguous across most jurisdictions. For LIS patients, this ambiguity is particularly problematic as they may be unable to advocate for their ownership rights due to communication limitations. Current legal frameworks provide inadequate coverage for neural data, with regulations like GDPR and HIPAA offering partial protections but failing to address the unique characteristics of brain-derived information [83] [87].

Key principles emerging in brain data governance include:

• Consent: Ensuring explicit, informed consent for data collection and use
• Transparency: Clear disclosure of how neural data will be processed and shared
• Control: Maintaining individual agency over neural data throughout its lifecycle
• Accountability: Establishing clear responsibility for ethical data management
• Neurorights: Developing specific rights protections for neural data [88]

For LIS patients, practical implementation of these principles requires specialized approaches, such as advance directives for data governance, surrogate decision-makers for current consent, and ongoing monitoring of data usage throughout research participation.

Informed Consent in LIS BCI Research

Obtaining meaningful informed consent from LIS patients presents distinctive challenges due to their communication limitations and the complex, rapidly evolving nature of BCI technologies. Research indicates that implantable BCI research involves multiple risk domains that must be addressed during the consent process [84] [89].

Core Risk Domains in Implantable BCI Consent

• Safety Risks: Surgical complications, inflammatory responses, device failure [84]
• Cognitive and Communicative Effects: Potential impacts on thought processes or communication abilities [84]
• Identity and Agency: Changes to self-perception, personal identity, or sense of agency [84]
• Privacy and Security: Unauthorized access to neural data or manipulation of BCI systems [84] [89]
• Stigma and Discrimination: Social consequences of BCI use or neural data disclosure [84]
• Long-Term Dependence: Reliance on technology for basic communication and potential abandonment issues [89]

Consent Protocol Adaptations for LIS

Successful consent protocols for LIS patients incorporate several adaptations:

• Extended Timeline: Multiple shorter sessions to accommodate fatigue and cognitive load
• Multi-Modal Communication: Utilizing residual communication channels (eye tracking, EEG responses)
• Surrogate Involvement: Engaging family members or caregivers while prioritizing patient autonomy
• Ongoing Consent: Implementing continuous consent mechanisms throughout research participation
• Empirical Assessment: Evaluating comprehension through tailored assessment tools [84] [25]

Methodological Considerations for Ethical LIS BCI Research

Consciousness Assessment in Non-Communicative Patients

Determining the optimal timing for BCI communication implementation requires assessment of consciousness levels in LIS patients, particularly as they transition to CLIS. Recent methodological advances utilize EEG-based approaches to evaluate consciousness through multiple feature extraction, including frequency, complexity, and connectivity measures [25]. These approaches calculate normalized consciousness levels (NCL) representing the probability of a patient being fully conscious, helping researchers identify appropriate times to initiate communication protocols.

The technical methodology involves:

• Feature Extraction: Calculating power spectral density, Lempel-Ziv complexity, and functional connectivity metrics from resting-state EEG
• Model Integration: Combining multiple features to maximize accuracy of consciousness assessment
• Validation: Comparing results with behavioral measures when available and tracking longitudinal changes
• Clinical Application: Using consciousness assessments to determine appropriate timing for BCI communication attempts [25]

Ethical Implementation Framework

A comprehensive ethical framework for LIS BCI research should incorporate the following elements:

• Pre-Research Assessment

  • Evaluation of residual communication capacities
  • Determination of appropriate consent pathways
  • Establishment of data governance preferences in advance directives

• Privacy by Design Implementation

  • Data minimization strategies collecting only essential neural signals
  • On-device processing where feasible to limit external data transmission
  • Strong encryption for data in transit and at rest
  • Clear data retention and destruction policies [83] [86]

• Ongoing Monitoring and Adaptation

  • Regular reassessment of comprehension and continued consent
  • Monitoring for unexpected psychological impacts
  • Adaptation of communication systems as patient abilities change
  • Transparent reporting of research progress and challenges

• Stakeholder Engagement

  • Inclusion of patients, caregivers, and advocacy groups in research design
  • Interdisciplinary collaboration between clinicians, engineers, and ethicists
  • Participation in broader conversations about neurorights and neural data governance

BCI technologies offer tremendous potential for restoring communication to individuals with Locked-In Syndrome, but realizing this potential requires careful attention to the complex ethical landscape surrounding brain data privacy, ownership, and informed consent. Researchers in this field must implement robust technical safeguards for neural data, develop adapted consent protocols for non-traditional communication pathways, and advocate for clear governance frameworks that protect the rights and interests of vulnerable populations. As BCI technologies continue to advance, maintaining ethical integrity while pursuing technical innovation remains paramount for ensuring that these transformative technologies benefit those who need them most.

Evaluating Efficacy and Impact: Clinical Validation and Comparative Analysis of BCI Systems

Locked-in syndrome (LIS) is a severe neurological disorder characterized by complete paralysis of nearly all voluntary muscles, including those responsible for speech, while cognitive function and consciousness remain fully intact [25] [90]. Patients with complete LIS (CLIS) lack even residual eye movement, making communication impossible without technological assistance and often leading to misdiagnosis as disorders of consciousness [25]. For individuals who have spent years or even decades without a functional voice, brain-computer interface (BCI) technology represents a revolutionary frontier in neuroprosthetics, offering the potential to restore not just communication, but personhood.

This technical guide examines pioneering case studies demonstrating the restoration of audible speech through intracranial BCIs. Framed within the broader context of LIS communication research, we analyze the experimental protocols, quantitative outcomes, and underlying technological frameworks that have enabled these breakthroughs. The recent integration of artificial intelligence (AI) with high-density electrode arrays has transformed the field, moving from slow, discrete symbol selection to fluid, naturalistic speech synthesis in near-real time [24] [58]. These advances mark a critical shift from providing basic communicative tools to restoring a fundamental human capacity: the voice itself.

Case Studies of Speech Restoration

Case Study 1: Real-Time Speech Synthesis After 18-Year Silence

A landmark 2025 study detailed the case of a 47-year-old woman who had been unable to speak for 18 years following a stroke [24] [58]. Researchers from UC Berkeley and UC San Francisco implemented a brain-to-voice neuroprosthesis that achieved a critical breakthrough: streaming intelligible speech from brain signals with minimal latency.

Experimental Protocol: The participant, known as "Ann," underwent implantation of a high-density electrode array over the cortical speech centers. To train the deep learning system, researchers recorded neural activity as she silently attempted to speak sentences drawn from a vocabulary of over 1,000 words [24]. This extensive dataset comprised more than 23,000 silent speech attempts across 12,000 sentences, enabling robust model training. A pivotal innovation involved using a pre-trained text-to-speech model to generate audio targets, which were then mapped to the corresponding neural signals [58]. The output was synthesized using a recording of Ann's pre-injury voice, restoring not just communication but personal identity.

Table 1: Quantitative Performance Metrics for Real-Time Speech BCI

Performance Metric	Previous System (2023)	New Streaming System (2025)
Decoding Latency	~8 seconds per sentence	<80 milliseconds (near-synchronous) [24]
Vocabulary Size	50 words	1,000+ words [24]
Decoding Rate (Complex Vocabulary)	Not achieved	47.5 words per minute [24]
Decoding Rate (50-Word Vocabulary)	~15 words per minute	90.9 words per minute [24]
Novel Sentence Synthesis	Limited	Capable [58]
Output Voice	Synthetic	Participant's own voice [58]

Case Study 2: Intracortical BCI for Anarthria

Complementary research has focused on intracortical microelectrode arrays. In one case, an individual with ALS and locked-in syndrome used an implanted microelectrode to modulate neural firing rates to spell words [25]. This foundational work demonstrated that CLIS patients retain the cognitive capacity for intentional communication, a prerequisite for more complex speech restoration.

Another study involved a participant with anarthria following a brainstem stroke. Researchers used electrocorticography (ECoG) signals from the sensorimotor cortex to decode attempted speech with high accuracy [91]. This approach leveraged the brain's intact articulatory representations, even when the physical machinery of speech was paralyzed.

Methodological Framework

Experimental Workflow for Speech BCI Implementation

The restoration of functional communication via BCI follows a structured experimental pathway, from patient selection to system validation. The diagram below illustrates this multi-stage workflow.

Core Technical Architecture

The brain-to-voice neuroprosthesis relies on a sophisticated technical architecture that processes neural signals through a sequential pipeline. The following diagram outlines the core components and data flow that enable real-time speech synthesis.

The Scientist's Toolkit: Research Reagent Solutions

The advancement of speech-restoring BCIs depends on specialized materials and technologies. The following table details essential research reagents and their functions in developing and implementing these systems.

Table 2: Essential Research Reagents and Materials for Speech BCI Development

Research Reagent / Material	Function & Application
High-Density Electrode Arrays	Chronic neural signal recording from cortical speech centers (e.g., motor cortex, inferior frontal gyrus) [24] [58].
Microelectrode Arrays (MEAs)	Penetrating electrodes for high-resolution intracortical signal acquisition [58].
Biocompatible Substrates	Materials like carbon nanotube meshes and hydrogel interfaces that reduce immune response and improve signal longevity [92].
Deep Learning Models	AI algorithms that map neural activity patterns to speech components (phonemes, words) or articulatory movements [24] [58].
Pretrained Text-to-Speech Models	Provide target audio outputs for training when participant vocalization is impossible [58].
Digital Speech Biobanks	Repository of pre-injury voice recordings used to personalize synthesized speech output [58].
Wireless Neural Signal Processors	Transmit recorded brain data to external computing systems without physical tethers [92].

Technical and Ethical Future Directions

While current systems demonstrate unprecedented performance, several challenges remain. Future research aims to incorporate paralinguistic features such as emotional tone, pitch, and volume to bridge the gap to complete naturalism [58]. Engineering improvements focus on miniaturization and wireless operation, with developing technologies including biodegradable silk-based polymers and energy harvesting from body heat to create fully implantable, invisible interfaces [92].

These rapid advancements necessitate parallel development of ethical frameworks. The emergence of "neurorights" addresses concerns regarding neural data privacy, agency, and identity [92]. As BCI technology transitions from restorative medicine to potential consumer applications, establishing governance structures that prioritize user safety and autonomy is paramount. The same technology that restores voice to the voiceless could potentially be misused for neuro-monitoring or manipulation without appropriate safeguards [92].

The pioneering case studies detailed in this technical guide mark a historic convergence of neuroscience, engineering, and artificial intelligence. The ability to translate neural representations of speech directly into audible, fluent output in near-real-time represents a fundamental breakthrough in neuroprosthetics. For individuals with locked-in syndrome who have endured years or decades of silence, this technology offers more than communication—it restores a essential channel for human expression, identity, and connection. As research advances to incorporate emotional nuance and wireless convenience, brain-computer interfaces are poised to transform from assistive devices for the severely disabled into powerful tools that redefine the boundaries of human communication.

Locked-in syndrome (LIS) and its most severe form, the completely locked-in state (CLIS), represent the most challenging neurological conditions for restoring communication. Patients with fully functional cognition become trapped within non-functional bodies due to conditions like amyotrophic lateral sclerosis (ALS), brainstem stroke, or traumatic brain injury [25] [93]. Brain-computer interfaces have emerged as the only potential means for these individuals to reconnect with their environment, making the rigorous assessment of BCI performance metrics not merely technical exercises but essential evaluations of potential lifelines back to human interaction.

Quantifying communication performance through standardized metrics like speed, accuracy, and Information Transfer Rate (ITR) provides critical insights for comparing BCI modalities, optimizing paradigms, and ultimately deploying reliable communication systems for those who have lost all other means of interaction. This review synthesizes current performance data across the BCI spectrum, details experimental methodologies, and identifies persistent challenges in the quest to unlock communication for the most severely paralyzed populations.

Performance Metrics: Definitions and Computational Foundations

The comparison of BCI systems relies on three fundamental metrics, each capturing distinct aspects of performance:

• Accuracy: The percentage of correct selections or classifications in a communication task, typically calculated as (Number of Correct Selections / Total Selections) × 100. This metric is particularly crucial for binary "yes/no" systems where chance performance is 50%.

• Communication Speed: Generally measured in characters per minute (CPM) or words per minute (WPM). This metric reflects the practical utility of a system for real-time communication.

• Information Transfer Rate (ITR): Measured in bits per minute, ITR represents the amount of information conveyed per unit time, incorporating both speed and accuracy into a single metric. The standard formula for ITR in a BCI with N possible choices is: ITR = [log₂N + P log₂P + (1-P) log₂((1-P)/(N-1))] × (60/T) where P is classification accuracy, N is the number of possible commands, and T is the time per selection in seconds.

Comparative Performance Analysis Across BCI Paradigms

Table 1: Performance Metrics Across BCI Modalities for LIS/CLIS Communication

BCI Paradigm	Modality	Target Population	Reported Accuracy	Communication Speed	Information Transfer Rate	Key Studies
P300 Speller	Visual EEG	LIS with preserved gaze	Up to 100% (healthy); Variable in patients	Up to 11.20 bits/min (LIS)	46.07 bits/min (with language model)	[45]
SSVEP	Visual EEG	LIS with preserved gaze	Fluctuating in CLIS (below chance in 13/40 sessions)	Not specified	Not specified	[34]
Auditory P300	Auditory EEG	LIS/CLIS	60-100% (highly variable)	Slow due to long trial durations	Generally low (<10 bits/min)	[25] [34]
Motor Imagery	EEG	LIS	Successful in some LIS patients	Not specified	Not specified	[25]
Intracortical Implant	Invasive	CLIS	86.6% (over 5700 trials)	~1 letter/minute; ~5 bits/min	~5 bits/min (average)	[60]
Auditory Frequency Modulation	Non-invasive EEG	CLIS	Perfect in final session for assistive needs	Not specified	Not specified	[45]

Table 2: Detailed Performance Characteristics from Key CLIS Studies

Study	Paradigm	Modality	Number of Sessions	Accuracy Range	Best Performance	Limitations
Chaudhary et al. (2022) [60]	Intracortical neurofeedback	Invasive	135 days over 12 months	50-100% (daily variation)	86.6% average over 5700 trials	Required daily recalibration; 17.1% of sessions below 80% accuracy
Liu et al. (2025) [45]	Auditory frequency modulation	Non-invasive EEG	7 sessions (4 online)	Variable across sessions	100% on assistive questions in final session	General knowledge questions at chance in some sessions
Auditory Oddball (2024) [34]	Auditory attention	Non-invasive EEG	Multiple	Most patients at chance level	100% in 2/7 patients with ALS	Significant performance gap between healthy controls and target population

Experimental Protocols and Methodologies

Intracortical Neurofeedback for CLIS Communication

The landmark study by Chaudhary et al. (2022) established the first reliable communication protocol for a CLIS patient using implanted microelectrode arrays [60]. The methodology proceeded through several critical phases:

Implantation and Signal Acquisition: Two 64-microelectrode arrays were implanted in the supplementary motor area and primary motor cortex. Neural signals (multi-unit activity) were recorded and processed to compute a spike rate metric (SRM).

Neurofeedback Training: The core innovation involved mapping the SRM to the frequency of an auditory feedback tone. Patients were asked to match this tone to target frequencies through volitional modulation of neural firing rates.

• "Feedback without reward" sessions: Patients attempted to match feedback tone frequency to pseudo-randomly presented target tones.
• "Feedback with reward" sessions: Successful matching triggered a reward sound, reinforcing correct modulation.

Calibration Protocol: A critical daily 10-minute baseline recording established individual channel firing rates. Experimenters manually selected channels showing differential modulation for high and low target tones, with parameters updated iteratively throughout each session.

Speller Implementation: Once neurofeedback performance exceeded 80% accuracy threshold, patients progressed to a letter-by-letter spelling interface using the same modulation strategy to select characters.

Non-Invasive Auditory BCI with Frequency Band Modulation

Liu et al. (2025) developed an alternative non-invasive approach for CLIS communication using volitional modulation of frequency band power [45]:

Experimental Setup: 32-channel EEG recording with specific focus on alpha (8-13 Hz) and beta (13-30 Hz) band power at selected channels.

Paradigm Structure:

• Offline Sessions (3 sessions): Patients auditorily cued to either volitionally increase EEG amplitude in alpha/beta bands ("Modulated" class) or return to baseline ("Baseline" class). Each run contained 14-20 trials with visual feedback of band power.
• Online Sessions (4 sessions): Patients used learned modulation to answer "Yes"/"No" questions ("Modulated" = "Yes", "Baseline" = "No") with real-time auditory feedback reflecting BCI classification.

Classifier Training: Subject-specific binary classifiers were trained on offline session data, then deployed for online communication with continuous auditory feedback of classification confidence.

Auditory Oddball Paradigms for Binary Communication

The auditory oddball approach represents another non-invasive strategy tested with LIS patients [34]:

Stimulus Design: Spoken words "yes" and "no" delivered through separate ears with standard (100ms) and deviant (150ms) durations. Stimulus Onset Asynchrony (SOA) is adjusted per patient (250ms for healthy controls).

Attention Modulation: Patients instructed to attend selectively to either "yes" or "no" stream, eliciting P300 event-related potentials for deviant stimuli in the attended stream.

Classification Features: Utilizes both P300 responses to deviant stimuli and N200 modulation to standard sounds for binary classification.

The Research Toolkit: Essential Materials and Reagents

Table 3: Essential Research Materials for BCI Communication Studies

Category	Item	Specification/Function	Representative Use
Recording Hardware	EEG Systems (e.g., eego mylab, ANT Neuro)	32+ channels, 512+ Hz sampling rate, electrode impedance <20 kΩ	Scalp potential recording [45] [5]
	Intracortical Microelectrode Arrays	64-microelectrode arrays, bilateral implantation	Motor cortex signal acquisition in CLIS [60]
Electrode Configurations	10-20 International System	Standardized scalp electrode placement	Consistent EEG montage across studies [45]
	Tripolar Scalp Electrodes	Enhanced gamma activity detection	RSVP Keyboard spelling paradigm [94]
Stimulus Delivery	Auditory Stimulation Equipment	High-fidelity headphones, calibrated sound delivery	Auditory oddball paradigms [34]
	Visual Display Systems	Eye-tracker integrated displays	P300 speller implementations [45]
Software & Algorithms	Signal Processing Pipelines	Bandpass filtering (e.g., 7-30 Hz for alpha/beta), feature extraction	Real-time frequency band power analysis [45]
	Classification Algorithms	SVM, LDA, deep neural networks	Binary classification of intentional states [25] [5]
	BCI Software Platforms	BciPy, OpenVibe, custom solutions	Experimental paradigm implementation [94]
Experimental Materials	Language Models	N-gram, word prediction algorithms	Information transfer rate enhancement [45]
	Calibration Protocols	Subject-specific parameter optimization	Daily system recalibration [60]

Signaling Pathways and Neural Correlates of Successful BCI Control

The neurophysiological mechanisms underlying successful BCI control involve distributed networks and multiple frequency bands:

Frontal-Theta and Temporal-Gamma Dynamics: Learning to operate an imagined speech BCI involves broad EEG power increases in frontal theta (4-7 Hz) activity coupled with focal enhancement in temporal low-gamma (30-60 Hz) activity [5]. This pattern suggests that BCI skill acquisition engages both executive attention networks (frontal theta) and specialized language processing regions (temporal gamma).

Spatial and Frequency Tuning: Successful BCI operation requires users to adapt their neural activity through neurofeedback, progressively refining both spatial activation patterns and frequency band modulation [5]. This learning process demonstrates the bidirectional nature of BCI systems, where both the machine and human nervous system co-adapt.

Conscious State Signatures: For consciousness assessment in LIS patients, feature extraction based on frequency, complexity, and connectivity measures can help establish normalized consciousness levels (NCL) even in the absence of behavioral ground truth [25]. These electrophysiological signatures may help identify optimal communication windows.

Critical Challenges and Future Research Directions

Despite promising advances, BCI communication for LIS/CLIS patients faces several persistent challenges:

The Performance Gap Between Healthy Controls and Patients: Auditory BCI paradigms that achieve 86% accuracy in healthy subjects frequently fail with the target patient population, with most LIS patients performing at chance level [34]. This discrepancy underscores the limitation of developing systems primarily tested on healthy individuals.

Invasive vs. Non-Invasive Tradeoffs: While intracortical implants have demonstrated the first reliable CLIS communication [60], their surgical risks, ethical considerations, and technical complexity limit widespread adoption. Non-invasive alternatives show promise but with more variable performance [45].

Within-Subject Performance Variability: Both invasive and non-invasive systems exhibit significant day-to-day and even within-session performance fluctuations, necessitating frequent recalibration and complicating reliable deployment [60].

Visual vs. Auditory Modality Challenges: Visual BCIs typically offer higher ITRs but become unusable in CLIS due to oculomotor paralysis and visual impairment [34]. Auditory BCIs provide gaze-independent operation but suffer from lower communication speeds and increased cognitive load.

Future research priorities should include: hybrid multimodal approaches combining residual capabilities, adaptive classifiers that accommodate neural signal non-stationarity, standardized performance metrics for cross-study comparison, and user-centered design incorporating caregiver feedback into system development.

Brain-Computer Interfaces (BCIs) represent a transformative technology for restoring communication to individuals with Locked-In Syndrome (LIS), a condition characterized by complete paralysis of nearly all voluntary muscles while cognitive function remains intact. The central dilemma for researchers and clinicians lies in the fundamental trade-off between the high-resolution signals provided by invasive BCIs, which require surgical implantation, and the superior safety profile of non-invasive BCIs, which record from the scalp but yield lower-fidelity data. This whitepaper provides a comparative analysis of these approaches, detailing their technical specifications, associated risks, and clinical utility within the context of LIS communication research. Emerging solutions, such as endovascular implants and hybrid systems that combine electroencephalography (EEG) with other modalities like eye-tracking, are poised to mitigate these trade-offs, offering a path toward more accessible and robust communication neuroprosthetics [67] [23] [3].

Locked-In Syndrome (LIS), and its most severe form, the Completely Locked-In State (CLIS), result from neurological injuries such as brainstem stroke, traumatic brain injury, or progressive diseases like Amyotrophic Lateral Sclerosis (ALS). Patients retain consciousness and cognitive function but lose all voluntary motor control, including the ability to speak, gesture, or even move their eyes [25]. This profound isolation creates an urgent need for technologies that can bypass damaged motor pathways to restore communication. BCIs answer this need by directly translating neural activity into executable commands for external devices, such as speech synthesizers or computer cursors [95]. The choice of BCI modality—invasive or non-invasive—is therefore not merely a technical consideration but a critical decision that impacts patient risk, quality of life, and the very feasibility of communication as the disease progresses into CLIS, where even eye-tracking systems fail [23] [25].

Technical Comparative Analysis: Invasive versus Non-Invasive BCIs

The core distinction between invasive and non-invasive BCIs lies in the proximity of the recording electrodes to the neural tissue generating the signals. This proximity directly governs the signal quality and, consequently, the potential performance of the communication system.

Table 1: Core Technical and Practical Comparison

Feature	Invasive BCI	Non-Invasive BCI (EEG)
Signal Acquisition	Electrodes implanted on the cortical surface (ECoG) or within brain tissue (intracortical) [3] [95]	Electrodes placed on the scalp surface [67] [96]
Spatial Resolution	High (millimeter-scale) [3]	Low (centimeter-scale) [67]
Temporal Resolution	Very High (milliseconds) [3]	High (milliseconds) [67]
Signal Fidelity	High signal-to-noise ratio (SNR), records high-frequency neural activity [96] [95]	Low SNR, signals attenuated and blurred by skull and scalp [67] [96]
Primary Risks	Surgical risks (infection, bleeding, tissue damage), long-term biocompatibility concerns [3] [96]	Minimal physical risk; primarily skin irritation [67] [96]
Key Advantage	High bandwidth and precision for complex control and speech decoding [24] [95]	Safety, accessibility, and ease of use for basic communication [67] [96]
Key Limitation	Requires complex surgery with inherent risks [3] [96]	Susceptible to noise and low signal quality, limiting information transfer rate [67] [96]

Table 2: Clinical Application in LIS/CLIS Communication

Parameter	Invasive BCI	Non-Invasive BCI (EEG)
Target Patient Population	Patients with severe LIS/CLIS where non-invasive methods have failed; requires eligibility for surgery [23] [95]	Patients in early LIS with some residual control, or as a first-line communication tool [23] [25]
Communication Speed	High (e.g., >90 words per minute for speech decoding [24])	Slow to Moderate (e.g., up to 24 characters per minute for a speller [95])
Typical Control Paradigms	Direct speech decoding, imagined handwriting, complex cursor control [24] [95]	P300 spellers, Steady-State Visual Evoked Potentials (SSVEP), motor imagery [23] [25]
Usability in CLIS	Potentially high, as it does not rely on any peripheral motor function [95]	Challenging, requires sustained, goal-directed thinking which may diminish in CLIS [23] [25]

The Signal Quality Trade-off

The divergence in signal quality is the most significant technical differentiator. Invasive systems, such as the Utah Array from Blackrock Neurotech or the Stentrode from Synchron, place micro-electrodes directly on or in the cortex. This allows them to record the firing of individual neurons or local field potentials with high fidelity and rich spectral content [3] [95]. This high-bandwidth signal is a prerequisite for complex applications like real-time speech decoding, where a recent NIH-funded study achieved a remarkable rate of 47.5 words per minute with a 99% success rate by implanting an electrode array over the speech cortex [24].

In contrast, non-invasive EEG measures the summed electrical activity of millions of neurons through the skull and scalp, which acts as a physical low-pass filter. The resulting signal is blurred and susceptible to contamination from muscle movement, eye blinks, and environmental noise [67]. While advanced machine learning algorithms can extract usable commands for spellers or simple controls, the information transfer rate is fundamentally limited, and the ability to decode continuous, imagined speech remains a formidable challenge [67] [25].

Experimental Protocols in LIS Communication Research

Protocol 1: Intracortical BCI for Speech Restoration

This protocol is based on a landmark 2025 study that restored naturalistic, fluent speech to a patient with paralysis using an invasive BCI [24].

• Objective: To decode and synthesize audible speech in real-time from intracranial neural signals in a person with severe paralysis and anarthria.
• Participant Selection: An individual with intact cognitive function but long-term paralysis and inability to speak (e.g., from a brainstem stroke) [24].
• Methodology:
  • Surgical Implantation: A high-density electrode array (e.g., ECoG array) is surgically implanted over the cortical areas responsible for speech motor control (e.g., ventral sensorimotor cortex) [24].
  • Data Acquisition & Training Phase: Over multiple sessions, neural activity is recorded while the participant silently attempts to speak a large set of sentences (e.g., >12,000 sentences) [24].
  • Model Training: A deep learning model, often a recurrent neural network (RNN) or transformer, is trained to map the recorded neural patterns to the intended phonetic or articulatory features of the target words. The model is trained to output speech in streaming increments of 80 milliseconds [24].
  • Real-Time Synthesis & Output: During the communication phase, the trained model decodes the participant's attempted speech and drives a speech synthesizer. The output can be personalized to the user's pre-injury voice if a recording is available [24].
• Key Outcome Measures: Words per minute (WPM), decoding accuracy (%), and output latency (ms) [24].

The following workflow diagram illustrates this experimental protocol:

Protocol 2: Hybrid Eye-Tracking and EEG BCI for Communication Transition

This protocol addresses the challenge of patients transitioning from LIS to CLIS, where conventional eye-tracking becomes ineffective [23].

• Objective: To create a hybrid communication system that seamlessly transitions control from eye-tracking to an EEG-based BCI as oculomotor function declines.
• Participant Selection: Individuals with a progressive condition like ALS who are in the early LIS stage and still rely on eye-tracking for communication [23].
• Methodology:
  • Multi-Modal Data Recording: Gaze position (via an eye-tracker) and EEG data (via a cap with wet or dry electrodes) are recorded simultaneously from the participant while they perform a communication task, such as a P300 speller [23].
  • Fusion Algorithm Development: A software algorithm is developed to fuse the gaze and EEG data. Initially, the system may rely more heavily on the more robust eye-tracking signal [23].
  • Calibration for Transition: As the system detects a decline in oculomotor control (e.g., inconsistent gaze or fatigue), it automatically increases the weighting of the EEG-P300 signal to maintain accurate detection of the user's intention [23].
  • Validation: System accuracy and information transfer rate are measured and compared across modality (eye-tracking alone, EEG alone, and the hybrid mode) [23].
• Key Outcome Measures: Intention detection accuracy (%) in hybrid vs. single modality, and user experience/Workload NASA-TLX.

The logical structure of this hybrid system is shown below:

The Scientist's Toolkit: Key Research Reagents and Materials

Table 3: Essential Materials for BCI Research

Item	Function in Research	Specific Examples / Notes
Microelectrode Arrays	Record neural signals at high resolution in invasive BCIs.	Utah Array (Blackrock Neurotech) [3], Neuralace (Blackrock's flexible lattice) [3], Neuralink's chip [3].
ECoG Grids/Strips	Record from the cortical surface with high SNR, less risky than intracortical arrays.	Used in speech decoding studies [24]; Precision Neuroscience's "Layer 7" film [3].
Endovascular Stent Electrodes	Minimally invasive method to record cortical signals via blood vessels.	Synchron's Stentrode [3].
High-Density EEG Systems	Record brain activity from the scalp for non-invasive BCI.	Systems with 64-128 channels; research focuses on dry electrodes for usability [67] [62].
Signal Acquisition Systems	Amplify, filter, and digitize analog neural signals for processing.	Essential for both invasive and non-invasive setups; requires high sampling rates and resolution [62].
Deep Learning Software Frameworks	Decode complex neural patterns for high-level tasks like speech.	TensorFlow, PyTorch; crucial for translating neural data to text/speech in real-time [24].
Stimulus Presentation Software	Present visual/auditory paradigms to evoke brain responses (e.g., P300).	Packages like Psychtoolbox or Presentation; used for P300 spellers and SSVEP systems [23] [25].

The field of BCI for LIS communication is at a pivotal juncture. Invasive BCIs have demonstrated unprecedented performance in restoring rapid, naturalistic communication, particularly for speech, but their widespread adoption is gated by surgical risks and ethical considerations. Non-invasive BCIs offer an immediate, safe solution for many but hit a performance ceiling due to fundamental physiological constraints. The future lies in technologies that bridge this divide. Minimally invasive approaches, such as Synchron's Stentrode and Precision Neuroscience's Layer 7 array, aim to offer a favorable compromise [3]. Furthermore, the integration of multiple signal modalities, such as hybrid EEG and eye-tracking, promises to provide continuous, adaptive communication support for patients as their condition evolves [23]. For researchers, the path forward requires not only refining signal decoding algorithms but also rigorously addressing the human factors of usability, patient training, and the long-term reliability of these complex systems within the demanding context of complete paralysis.

Locked-in Syndrome (LIS) is a profound neurological condition characterized by complete paralysis of nearly all voluntary muscles while cognitive function remains intact. Patients with LIS face severe challenges in communication and interaction, leading to significant impairments in autonomy, well-being, and social reintegration [25]. Brain-Computer Interfaces (BCIs) have emerged as transformative technologies that can decode neural signals to restore communication channels for this population. This whitepaper provides a technical analysis of methodologies for quantitatively measuring the impact of BCIs on these critical quality-of-life domains within the context of LIS research. The framework presented enables researchers to standardize assessment protocols across clinical studies, facilitating direct comparison of intervention outcomes and advancing the therapeutic application of BCI technologies.

Quantitative Impact Assessment of BCI Interventions

The efficacy of BCI interventions must be evaluated through multidimensional metrics that capture performance data, functional outcomes, and subjective experiences. The tables below synthesize quantitative findings from recent studies and standardized assessment frameworks.

Table 1: Quantitative BCI Performance Metrics in Communication Restoration

Study/Technology	Patient Population	Communication Rate	Accuracy	Longevity/Stability
Intracortical LFP BCI [28]	LIS from brainstem stroke & ALS	3.07-6.88 characters/minute	Not specified	76-138 days without recalibration
Streaming Speech Neuroprosthesis [24]	Paralysis from stroke	47.5-90.9 words/minute	>99%	Continuous operation demonstrated
P300 Auditory BCI [34]	ALS patients (LIS)	Binary yes/no classification	Up to 100% in 2/7 patients	Single session testing
Hybrid ET-BCI System [23]	Healthy controls (LIS model)	Not specified	High accuracy maintained	Facilitates LIS-to-CLIS transition

Table 2: Standardized Metrics for Assessing Psychosocial Outcomes

Assessment Domain	Specific Metrics	Measurement Tools	Key Findings from Literature
Autonomy	Communication independence, Device control capability	FIM, task success rate, information transfer rate	BCI communication provides first-ever independence for complete LIS [28] [24]
Well-being	Quality of life, Emotional state, Frustration levels	QOLIBRI, visual analog scales, EEG emotional state detection	Ability to communicate significantly enhances wellbeing [25]; Mental state monitoring achieves 74.26% classification accuracy [97]
Social Reintegration	Social interaction frequency, Relationship satisfaction	Social network indices, custom questionnaires	Restoration of communication enables social reconnection and reduces caregiver dependence [28] [98]

Neural Mechanisms of BCI Operation

BCIs function by detecting and interpreting specific neural signals associated with conscious intention. The following diagram illustrates the core signal processing pathway shared across non-invasive BCI systems.

The fundamental neural signals leveraged in LIS communication BCIs include:

• Slow Cortical Potentials (SCPs): Low-frequency shifts in EEG signals that users can voluntarily regulate through biofeedback training. Early SCP-based systems enabled basic spelling but required extensive training and caused patient fatigue [25].
• P300 Event-Related Potentials: Positive deflections in EEG signals occurring approximately 300ms after a rare or significant stimulus within an oddball paradigm. These signals are particularly valuable for conscious attention detection in patients who cannot provide motor responses [25] [34].
• Steady-State Visual Evoked Potentials (SSVEPs): Oscillatory brain responses to visual stimuli flickering at specific frequencies. While effective, these require preserved gaze control, making them unsuitable for complete LIS (CLIS) [25] [34].
• Local Field Potentials (LFPs): Invasive signals representing aggregated neuronal activity near implanted electrodes. LFPs offer superior signal stability compared to single-unit recordings, enabling long-term BCI use without recalibration [28].

Experimental Protocols for BCI Assessment

Consciousness Assessment Protocol

Determining the optimal timing for communication attempts with non-responsive patients requires reliable consciousness assessment. The following protocol, adapted from recent research, enables consciousness detection without requiring motor responses [25].

Methodology Details:

• EEG Acquisition: Record resting-state EEG from multiple LIS patients over extended periods (years) using standard scalp electrodes [25].
• Feature Extraction: Compute multiple feature domains simultaneously:
  • Frequency Features: Band power analysis across delta, theta, alpha, beta, and gamma ranges
  • Complexity Features: Lempel-Ziv complexity (LZC) and Perturbational Complexity Index (PCI)
  • Connectivity Features: Spectral coherence and functional network connectivity measures [25]
• NCL Calculation: Integrate features into a Normalized Consciousness Level (NCL) score ranging from 0 (no consciousness) to 1 (full consciousness) using machine learning integration. This approach maximizes probability of correctly determining patient state despite absence of ground truth [25].
• Validation: Compare NCL scores with subsequent BCI communication performance accuracy to validate the consciousness assessment methodology [25].

Auditory BCI Communication Protocol

For patients with visual impairments or oculomotor paralysis, auditory BCIs provide a viable communication alternative. The following protocol details a validated auditory BCI implementation [34].

Experimental Design:

• Stimuli: Spoken words "yes" and "no" delivered via headphones ("yes" to right ear, "no" to left ear) using a synthesized male voice [34].
• Stimulus Parameters: Standard sounds (100ms duration) and deviant sounds (150ms duration) with Stimulus Onset Asynchrony (SOA) adjustable per patient (typically 250ms for healthy subjects) [34].
• Paradigm: Oddball sequence with intermixed streams, where "yes" stream begins 250ms before "no" stream to avoid simultaneous presentation [34].
• Task: Patients instructed to selectively attend to either "yes" or "no" stream based on their intended communication response.

Signal Processing Pipeline:

• EEG Recording: Multi-channel scalp recording during auditory presentation
• Preprocessing: Bandpass filtering, artifact removal, and epoching
• Feature Extraction: Temporal features from both standard and deviant stimuli, including N200 and P300 components
• Classification: Binary classification of attention to "yes" vs "no" streams using machine learning algorithms
• Online Feedback: Real-time classification results provided to enable actual communication

Performance Notes: While healthy subjects achieve high accuracy (average 86%), patients with severe motor disabilities show more variable performance, highlighting the need for patient-specific adaptations [34].

Research Reagent Solutions for BCI Studies

Table 3: Essential Research Materials and Tools for BCI Communication Studies

Tool/Category	Specific Examples	Research Application	Technical Specifications
Non-Invasive EEG Systems	EMOTIV EPOC+, Neuroelectrics, G.Tec Medical Engineering [8] [97]	Signal acquisition for communication and mental state classification	Multi-channel dry/wet electrodes, wireless connectivity, real-time processing
Invasive Implant Systems	Blackrock Microsystems arrays, Synchron Stentrode, Paradromics Connexus [28] [99] [98]	High-fidelity signal acquisition for complete LIS	96-channel microelectrode arrays, stent-based electrodes, biocompatible materials
Signal Processing Software	Custom MATLAB toolboxes, Python MNE, BCILAB	Feature extraction and machine learning	Support for EEG, ECoG, LFP data; real-time classification algorithms
Stimulation Hardware	Piezo-electric headphones, tactile stimulators	Auditory and vibro-tactile BCI paradigms	Precise timing control, adjustable intensity, patient-safe design
Mental State Classification	OvR-LR algorithms, powerband analysis tools [97]	Monitoring user affect and cognitive load	74.26% average accuracy for mental state prediction
Validation Tools	FIM, QOLIBRI, custom questionnaires [25] [97]	Assessing psychosocial outcomes	Standardized metrics for autonomy, well-being, and social reintegration

Discussion and Future Directions

The quantitative assessment of BCI impact on patient autonomy, well-being, and social reintegration requires continued methodological refinement. Several emerging trends are particularly noteworthy:

Hybrid System Integration: Research indicates that combining multiple modalities (e.g., eye-tracking with EEG) can support patients during the transition from LIS to complete LIS, maintaining communication continuity as motor function further declines [23].

Long-Term Stability Solutions: The demonstrated stability of Local Field Potential (LFP)-based decoding over several months without recalibration addresses a critical barrier to practical BCI implementation for everyday use [28].

Mental State Integration: Adaptive BCIs that monitor and respond to users' emotional states represent a promising direction for enhancing safety and usability, particularly for sensitive populations with severe physical limitations [97].

Standardized Metrics: Future research should prioritize the development and validation of standardized, quantifiable metrics specifically designed to capture psychosocial outcomes in the LIS population, enabling more rigorous comparison across studies and technologies.

As BCI technologies continue to evolve from laboratory demonstrations to clinical tools, the framework presented in this whitepaper provides researchers with methodologies to rigorously quantify their real-world impact, ultimately accelerating the development of more effective communication solutions for individuals with Locked-In Syndrome.

The translation of neurorehabilitation technologies from laboratory research to clinical practice remains a significant challenge, particularly for complex interventions such as Brain-Computer Interfaces (BCIs) for Locked-In Syndrome (LIS). This whitepaper presents the NEURO Framework, a comprehensive model designed to structure and accelerate the clinical translation and validation pathway. Developed within the context of LIS BCI communication research, the framework integrates validation checkpoints across five critical domains: Neurotechnology Readiness, Efficacy and Endpoints, User-Centered Design, Regulatory and Reimbursement Pathways, and Outcomes and Implementation. By providing a structured approach to technology development with clearly defined milestones, the NEURO Framework addresses the unique challenges of neurorehabilitation translation, with specific application to the development of communication technologies for individuals with severe paralysis.

Locked-In Syndrome (LIS) represents one of the most severe neurological conditions, characterized by preserved consciousness and cognitive function amid near-total paralysis of voluntary muscles, including those responsible for speech and movement [25]. Often resulting from brainstem stroke, traumatic brain injury, or progressive neurological diseases such as Amyotrophic Lateral Sclerosis (ALS), LIS presents profound communication challenges that significantly impact patients' quality of life and clinical outcomes [65] [25]. The progression to Complete Locked-In State (CLIS), where even residual eye movements are lost, creates what has been described as a "communication blackout," severing patients from interaction with their families, caregivers, and environment [23] [56].

Brain-Computer Interface technology has emerged as a promising solution to restore communication for individuals with LIS. BCIs establish a direct communication pathway between the brain and external devices, bypassing damaged neural pathways and paralyzed muscles [65]. Recent advances demonstrate remarkable progress, including intracortical microelectrode arrays that enable completely locked-in individuals to communicate [56] and speech neuroprostheses that translate brain activity into audible words at near-conversational speeds [24]. However, the translation of these technological breakthroughs from research demonstrations to clinically validated, widely available interventions has been hampered by numerous barriers, including methodological inconsistencies, limited long-term validation, and insufficient attention to implementation contexts.

The NEURO Framework addresses these challenges by providing a structured model for guiding clinical translation and validation in neurorehabilitation, with specific application to LIS BCI communication research. This framework synthesizes recent advances in BCI technology, validation methodologies, and implementation science to create an integrated pathway from concept to clinical adoption.

The NEURO Framework: Core Components

The NEURO Framework structures the translation pathway through five interdependent domains, each containing specific validation milestones and metrics. The framework's components are visualized below, illustrating their interconnected nature in the development pathway.

Neurotechnology Readiness

The Neurotechnology Readiness domain focuses on the technical development and validation of BCI systems, ensuring they meet the rigorous requirements for clinical use with LIS populations. This domain encompasses signal acquisition hardware, processing algorithms, and system integration.

Table 1: Neurotechnology Readiness Metrics for LIS BCI Systems

Component	Validation Metrics	Target Performance for Clinical Use	Assessment Methods
Signal Acquisition	Signal-to-noise ratio, spatial/temporal resolution, stability	>6 months stable recording [56], artifact rejection	Long-term stability testing, phantom testing
Decoding Algorithms	Classification accuracy, speed, adaptability	>90% accuracy for binary communication [56], <80ms latency for speech [24]	Cross-validation, online testing, adaptive learning assessment
System Integration	Reliability, ease of use, set-up time	<30-minute set-up by caregivers, >95% uptime [56]	Usability testing with caregivers, reliability monitoring

Recent advances in BCI platforms have enabled more robust signal processing and analysis. Frameworks like NeuronLab provide multi-platform, standalone environments that support the entire BCI lifecycle, from data acquisition to analysis, with specific functionality for sharing experiments between researchers and storing data in the cloud [100]. Similarly, the BCI-HIL framework enables real-time stimulus control and human-in-the-loop model training through a modular, hardware-independent design [101]. These technological developments create the infrastructure necessary for rigorous validation of BCI systems for LIS communication.

Efficacy and Endpoints

The Efficacy and Endpoints domain establishes methodological standards for evaluating BCI interventions in LIS populations, with particular attention to the selection of appropriate outcome measures and clinical trial designs.

Table 2: Efficacy Endpoints for LIS BCI Communication Research

Endpoint Category	Specific Measures	Validation Requirements	LIS-Specific Considerations
Primary Efficacy Endpoints	Communication accuracy rate, information transfer rate (ITR)	Superiority to established alternatives (e.g., eye-tracking), non-inferiority to baseline	Accounting for LIS/CLIS progression, cognitive fatigue
Functional Outcomes	Quality of life measures, participation in decision-making, social interaction	Minimal clinically important difference (MCID) established	Proxy reporting when necessary, environmental control metrics
Patient-Reported Outcomes	Satisfaction, ease of use, communication confidence	Reliability and validity in severe paralysis	Adaptive administration methods, binary response formats

The development of validated endpoints must consider the progressive nature of conditions leading to LIS, particularly ALS. As patients transition from LIS to CLIS, communication modalities may need to shift from eye-tracking to entirely BCI-based systems [23]. Hybrid BCI frameworks that combine multiple modalities (e.g., eye-tracking and EEG) can support this transition while maintaining communication continuity [23]. Efficacy testing must therefore account for different stages of disability progression and include appropriate biomarkers of consciousness and cognitive function to ensure that patients can effectively use the technology [25].

User-Centered Design

User-Centered Design emphasizes the importance of involving end users throughout the development process, particularly for LIS populations with severe physical limitations but preserved cognition. This domain addresses interface design, user experience, and integration with caregivers and clinical workflows.

The design of BCI systems for LIS must accommodate the unique needs and capabilities of this population. Research indicates that successful communication has been achieved with completely locked-in patients using auditory neurofeedback training, where a speller program reads letters aloud and the patient selects them through "yes/no" responses decoded from neural signals [56]. This approach demonstrates how interface design can adapt to the severe motor limitations of CLIS while maintaining communication capability.

Furthermore, the involvement of family members and caregivers is essential for successful implementation of BCI systems in home environments [56]. Studies have shown that with proper training and support, caregivers can operate BCI systems effectively, enabling long-term communication for individuals with severe paralysis outside hospital settings.

Regulatory and Reimbursement Pathways

The Regulatory and Reimbursement Pathways domain addresses the requirements for regulatory approval and healthcare coverage, ensuring that BCI technologies for LIS can successfully navigate the pathway from investigational devices to approved, reimbursed interventions.

Regulatory strategy must consider the classification of BCI systems as medical devices, which requires demonstration of safety and effectiveness through controlled trials. The recent development of BCI systems as medical-grade, implantable devices highlights the importance of robust clinical evidence generation [56]. For instance, the ABILITY system—a wireless implantable BCI device designed to flexibly connect to either microelectrode arrays or ECoG electrode grids—represents the type of technology that would require comprehensive regulatory review [56].

Reimbursement planning must begin early in the development process, with consideration of health economic outcomes and value propositions. As noted in recent research, "This technology, benefiting a patient and his family in their own environment, is a great example of how technological advances in the BCI field can be translated to create direct impact" [56]. Documenting this impact through rigorous health economic studies is essential for securing reimbursement from healthcare payers.

Outcomes and Implementation

The Outcomes and Implementation domain focuses on the long-term impact and real-world integration of BCI technologies for LIS communication, including health economic analysis, implementation planning, and sustainability.

Implementation planning must address the specific requirements for deploying BCI technologies in home and clinical settings. This includes technical support, caregiver training, and system maintenance. Recent studies have demonstrated that with appropriate support, BCI systems can be used successfully in home environments over extended periods [56]. This represents a significant advancement for the field, as it moves BCI technology from highly controlled laboratory settings to real-world environments where it can have meaningful impact on patients' lives.

Long-term outcomes assessment should include not only communication accuracy, but also broader impacts on quality of life, healthcare utilization, and caregiver burden. As noted in recent research, "The ability to interact with their environment has been shown to significantly enhance their wellbeing and quality of life" [25]. These broader outcomes are essential for demonstrating the value of BCI technologies beyond technical performance metrics.

Experimental Protocols and Methodologies

This section details specific experimental protocols relevant to the development and validation of BCI systems for LIS communication, providing methodological guidance for researchers working within the NEURO Framework.

Protocol for Consciousness Assessment in LIS/CLIS Populations

Background: Determining the optimal time for communication with LIS/CLIS patients is challenging due to fluctuations in consciousness levels and the absence of behavioral markers. This protocol outlines a method for assessing consciousness levels using EEG to identify appropriate times for BCI-based communication.

Materials:

• EEG acquisition system with appropriate electrode montage
• Signal processing software (e.g., EEGLAB, MNE-Python)
• Algorithms for feature extraction (spectral, complexity, connectivity measures)

Procedure:

• Acquire resting-state EEG data according to standardized protocols
• Preprocess data to remove artifacts and normalize signals
• Extract multiple feature types:
  • Frequency-based features (power spectral density across standard bands)
  • Complexity measures (Lempel-Ziv complexity, entropy measures)
  • Connectivity measures (coherence, phase-based connectivity)
• Integrate features to derive a normalized consciousness level (NCL) score ranging from 0 (not conscious) to 1 (fully conscious)
• Establish patient-specific baselines through repeated assessments
• Use NCL scores to determine optimal timing for communication sessions

Validation: Compare NCL scores with communication performance accuracy when possible. In completely locked-in patients where no behavioral validation is possible, use internal consistency measures and comparison with known physiological correlates of consciousness [25].

Protocol for Hybrid BCI Evaluation for LIS-to-CLIS Transition

Background: As patients with progressive conditions like ALS transition from LIS to CLIS, they often lose the ability to use eye-tracking systems, creating a communication gap. This protocol evaluates hybrid BCI systems that combine multiple modalities to support continuous communication during this transition.

Materials:

• Eye-tracking system
• EEG-based BCI system with appropriate paradigms (P300, motor imagery)
• Data fusion algorithms for integrating multiple input modalities
• Performance assessment metrics

Procedure:

• Recruit participants across the LIS-CLIS spectrum (including healthy controls for initial validation)
• Implement simultaneous eye-tracking and EEG recording during communication tasks
• Design communication interfaces that can utilize both input modalities adaptively
• Develop fusion algorithms that combine inputs from both modalities to enhance intention detection
• Test system performance in various scenarios:
  • Eye-tracking only (simulating early LIS)
  • Combined eye-tracking and EEG (simulating transition phase)
  • EEG only (simulating CLIS)
• Assess accuracy, information transfer rate, and user experience across conditions

Validation Metrics: Communication accuracy, information transfer rate, time to correct selection, user satisfaction measures [23].

The experimental workflow for developing and validating BCI systems within the NEURO Framework follows a structured pathway from concept to implementation, as illustrated below.

Protocol for At-Home BCI Deployment and Monitoring

Background: Long-term deployment of BCI systems in home environments presents unique challenges and opportunities for LIS communication. This protocol outlines methods for implementing and monitoring BCI use in home settings.

Materials:

• Portable BCI system suitable for home use
• Remote monitoring capabilities
• Caregiver training materials and support system
• Data collection infrastructure for long-term assessment

Procedure:

• Select appropriate BCI technology for home deployment based on reliability and ease of use
• Develop comprehensive training programs for caregivers and family members
• Implement remote monitoring and technical support systems
• Establish data collection protocols for:
  • System usage patterns and reliability
  • Communication accuracy and speed
  • Quality of life measures
  • Caregiver burden assessments
• Schedule regular follow-up assessments and system maintenance
• Create iterative feedback mechanisms for system improvements

Validation Metrics: System uptime, communication reliability, user and caregiver satisfaction, impact on quality of life measures [56].

The Scientist's Toolkit: Research Reagent Solutions

The development and validation of BCI systems for LIS communication requires specialized tools and frameworks. The table below details essential research reagents and their functions in advancing this research.

Table 3: Essential Research Reagents and Tools for LIS BCI Communication Research

Tool/Reagent	Function	Example Applications	Availability
NeuronLab Framework	Secure, multi-platform BCI framework covering entire lifecycle	P300 identification, limb movement detection, cloud-based data sharing [100]	Open-source
BCI-HIL Framework	Human-in-the-loop BCI research with real-time model training	Stimulus control, transfer learning, online EEG classification [101]	MIT License
DART-RT Assessment Tool	Predicts implementation potential of rehabilitation technologies	Evaluating cost, safety, effectiveness, clinical demand [102]	Research use
Intracortical Microelectrode Arrays	Record neural signals from motor cortex for communication	Yes/no selection for letter spelling in CLIS [56]	Investigational use
Auditory Neurofeedback Training	Enables communication through auditory feedback	Letter selection in spelling interface for CLIS [56]	Research protocol
Normalized Consciousness Level (NCL)	Assesses consciousness levels in LIS/CLIS patients	Determining optimal communication timing [25]	Analytical method

The NEURO Framework provides a comprehensive model for guiding the clinical translation and validation of neurorehabilitation technologies, with specific application to BCI communication systems for Locked-In Syndrome. By structuring the development pathway across five critical domains—Neurotechnology Readiness, Efficacy and Endpoints, User-Centered Design, Regulatory and Reimbursement Pathways, and Outcomes and Implementation—the framework addresses the complex challenges inherent in translating technological innovations into clinically meaningful interventions.

Recent advances in BCI technology, including the development of implantable systems that enable communication for completely locked-in individuals [56] and speech neuroprostheses that approach conversational speeds [24], demonstrate the tremendous potential of these approaches. However, realizing this potential requires systematic approaches to validation and implementation. The NEURO Framework provides this structure, enabling researchers, clinicians, and developers to navigate the complex translation pathway more efficiently and effectively.

As the field continues to advance, the NEURO Framework offers a adaptable structure for guiding development decisions, regulatory strategies, and implementation planning. By applying this framework to LIS BCI communication research, we can accelerate the translation of promising technologies from laboratory demonstrations to clinically available interventions that restore communication and improve quality of life for individuals with severe paralysis.

Conclusion

The field of BCI research for LIS communication is at a transformative juncture, demonstrated by successful clinical deployments that restore fundamental communication abilities. Key takeaways include the proven feasibility of both invasive and non-invasive systems, the critical role of AI and streaming decoding for real-time performance, and the importance of hybrid and user-centered design for practical application. Future directions must focus on developing robust, plug-and-play wireless systems for home use, establishing clear regulatory pathways, and conducting large-scale clinical trials to move from proof-of-concept to standard of care. For biomedical research, this necessitates intensified interdisciplinary collaboration to further refine signal decoding, create personalized digital avatars, and explore synergistic drug-device therapies that protect neural integrity and enhance BCI performance.

References

• 1. Boosting brain–computer interfaces with functional electrical stimulation... [https://jneuroengrehab.biomedcentral.com/articles/10.1186/s12984-023-01272-y]
• 2. Brain–Computer Interface–Based Communication in the Completely ... [https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1002593]
• 3. BCIs in 2025: Trials, Progress, and Challenges [https://andersenlab.com/blueprint/bci-challenges-and-opportunities]
• 4. Stable long-term BCI -enabled communication in ALS and locked - in ... [https://pubmed.ncbi.nlm.nih.gov/29694279/]
• 5. Learning to operate an imagined speech Brain-Computer ... [https://www.nature.com/articles/s42003-025-07464-7]
• 6. Recommendations for Combining Brain-Computer Interface ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12527325/]
• 7. Toward the integration of mixed reality and brain-computer ... [https://www.sciencedirect.com/science/article/pii/S0208521625000518]
• 8. Brain Computer Interface (BCI) Research Report 2025-2035: [https://www.globenewswire.com/news-release/2025/10/01/3159276/0/en/Brain-Computer-Interface-BCI-Research-Report-2025-2035-EEG-MEG-and-fMRI-Integration-Enhances-Neural-Signal-Detection-and-Drives-Global-Growth.html]
• 9. Locked-in Syndrome - StatPearls - NCBI Bookshelf - NIH [https://www.ncbi.nlm.nih.gov/books/NBK559026/]
• 10. Application and future directions of brain-computer ... [https://www.sciencedirect.com/science/article/pii/S2589238X2500083X]
• 11. BCI in patients with disorders of consciousness: Clinical ... [https://www.sciencedirect.com/science/article/pii/S1877065714018326]
• 12. The Pons - Function - Location - Vasculature [https://teachmeanatomy.info/neuroanatomy/brainstem/pons/]
• 13. Neuroanatomy, Pons - StatPearls - NCBI Bookshelf - NIH [https://www.ncbi.nlm.nih.gov/books/NBK560589/]
• 14. Locked-in Syndrome (LiS): What It Is, Causes & Symptoms [https://my.clevelandclinic.org/health/diseases/22462-locked-in-syndrome-lis]
• 15. Locked-in syndrome [https://en.wikipedia.org/wiki/Locked-in_syndrome]
• 16. A human brain network linked to restoration of ... [https://www.nature.com/articles/s41467-025-61988-4]
• 17. BRAIN 2025: A Scientific Vision - BRAIN Initiative - NIH [http://braininitiative.nih.gov/vision/nih-brain-initiative-reports/brain-2025-scientific-vision]
• 18. World Stroke Organization: Global Stroke Fact Sheet 2025 [https://pmc.ncbi.nlm.nih.gov/articles/PMC11786524/]
• 19. Epidemiology of amyotrophic lateral sclerosis: a population ... [https://pubmed.ncbi.nlm.nih.gov/40644419/]
• 20. ALS Disease Statistics: Insights from Latest Research [https://www.targetals.org/2025/01/30/als-statistics/]
• 21. Global, regional, and national burdens of traumatic brain ... [https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2025.1556147/full]
• 22. Global, regional, and national burdens of traumatic brain ... [https://pubmed.ncbi.nlm.nih.gov/40297033/]
• 23. Eye-Tracking and BCI Integration for Assistive ... [https://arxiv.org/abs/2509.23518]
• 24. Brain-computer interface restores natural speech after ... [https://www.nih.gov/news-events/nih-research-matters/brain-computer-interface-restores-natural-speech-after-paralysis]
• 25. Assessing consciousness in patients with locked ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2025.1604173/full]
• 26. Noninvasive Brain Computer Interfaces for Augmentative and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6525622/]
• 27. Brain–computer interface in critical care and rehabilitation [https://www.accjournal.org/journal/view.php?doi=10.4266/acc.2023.01382]
• 28. Stable long-term BCI-enabled communication in ALS and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6093965/]
• 29. High performance communication by people with paralysis ... [https://elifesciences.org/articles/18554]
• 30. bigP3BCI: An Open, Diverse and Machine Learning Ready ... [https://physionet.org/content/bigp3bci/]
• 31. Exploring the Impact of Brain-Computer Interfaces on ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11805155/]
• 32. Update on How to Approach a Patient with Locked-In ... [https://www.mdpi.com/2076-3425/14/1/92]
• 33. Update on How to Approach a Patient with Locked-In ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10813368/]
• 34. The challenge of controlling an auditory BCI in the case of ... [https://jneuroengrehab.biomedcentral.com/articles/10.1186/s12984-023-01289-3]
• 35. Disorders of Consciousness, Language and ... [https://psychologicabelgica.com/articles/10.5334/pb.1381]
• 36. Assessing consciousness in patients with locked ... [https://pubmed.ncbi.nlm.nih.gov/41017975/]
• 37. mindBEAGLE Brain Assessment and Communication [https://www.gtec.at/product/mindbeagle-brain-assessment-and-communication/?srsltid=AfmBOooznV8s76hRlM2kJWTV-5gvgwfAyGmQZZdXTb755vJ6OlTk8snj]
• 38. Assessing consciousness in patients with locked-in syndrome ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12460305/]
• 39. Non-Invasive Brain-Computer Interfaces: State of the Art and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11861396/]
• 40. Communication in locked-in state after brainstem stroke [https://pmc.ncbi.nlm.nih.gov/articles/PMC4437925/]
• 41. A hybrid P300-SSVEP brain-computer interface speller ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2023.1133933/full]
• 42. Dataset of binocularly coded steady-state visual evoked ... [https://www.nature.com/articles/s41597-025-05696-0]
• 43. Enhancing motor rehabilitation using a hybrid BCI ... [https://www.sciencedirect.com/science/article/pii/S2324242625000609]
• 44. Exploring neural activity changes during motor imagery-based ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12592122/]
• 45. EEG-Based Auditory BCI for Communication in a ... [https://arxiv.org/html/2507.00305v1]
• 46. Electrocorticography (ECoG) surveys the landscape | Intracortical electrodes mine for gold [https://www.paradromics.com/blog/electrocorticography]
• 47. Origami-inspired soft fluidic actuation for minimally invasive ... [https://www.nature.com/articles/s41467-024-50597-2]
• 48. Cortical Electrocorticogram (ECoG) Is a Local Signal - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC6538865/]
• 49. Low-threshold, high-resolution, chronically stable intracortical ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10592461/]
• 50. The “Podcast” ECoG dataset for modeling neural activity ... [https://www.nature.com/articles/s41597-025-05462-2]
• 51. Minimally invasive electrocorticography (ECoG) recording ... [https://www.sciencedirect.com/science/article/pii/S0165027025000500]
• 52. Implementing intraoperative high‐density electrocorticography ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12097464/]
• 53. Decoding visual object categories from temporal ... [https://www.sciencedirect.com/science/article/pii/S105381191301224X]
• 54. Eye-Tracking and BCI Integration for Assistive ... [https://arxiv.org/html/2509.23518v1]
• 55. A new era of brain-computer interfaces: Restored touch ... [https://www.eurekalert.org/news-releases/1106526]
• 56. Completely locked-in man uses brain-computer interface to ... [https://wysscenter.ch/update/completely-locked-in-man-uses-brain-computer-interface-to-communicate/]
• 57. Advancing Brain-Computer Interface Closed-Loop Systems for ... [https://biomedeng.jmir.org/2025/1/e72218]
• 58. Brain-to-voice neuroprosthesis restores naturalistic speech [https://engineering.berkeley.edu/news/2025/03/brain-to-voice-neuroprosthesis-restores-naturalistic-speech/]
• 59. First-of-its-kind technology helps man with ALS 'speak' in ... [https://health.ucdavis.edu/news/headlines/first-of-its-kind-technology-helps-man-with-als-speak-in-real-time/2025/06]
• 60. Spelling interface using intracortical signals in a ... - Nature [https://www.nature.com/articles/s41467-022-28859-8]
• 61. The challenge of controlling an auditory BCI in the case ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10795353/]
• 62. Brain Computer Interfaces 2025-2045: Technologies, ... [https://www.idtechex.com/en/research-report/brain-computer-interfaces/1024]
• 63. Brain-Computer Interfaces: User-Experience Design ... [https://www.coderio.com/biz-tech/brain-computer-ux-design-principles/]
• 64. Brain Computer Interfaces Market Size & Outlook, 2025-2033 [https://straitsresearch.com/report/brain-computer-interfaces-market]
• 65. Brain–computer interfaces: the innovative key to unlocking ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11392146/]
• 66. How To Reduce Noise In EEG Recordings [11 Solutions] [https://mentalab.com/reduce-noise-in-eeg-recordings/]
• 67. The history, current state and future possibilities of the non- ... [https://www.sciencedirect.com/science/article/pii/S2590093525000049]
• 68. A novel approach for noise removal and distinction of EEG ... [https://www.sciencedirect.com/science/article/abs/pii/S174680941730143X]
• 69. Auditory Noise Leads to Increased Visual Brain-Computer ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7783197/]
• 70. Noise benefits in the array of brain-computer interface ... [https://www.sciencedirect.com/science/article/pii/S2352914818300923]
• 71. Think Fast | Setting new standards and new records for ... [https://www.paradromics.com/blog/bci-benchmarking]
• 72. Evaluating the Feasibility of Visual Imagery for an EEG- ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11249027/]
• 73. Complete Locked-in and Locked-in Patients [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2017.00251/full]
• 74. Brain-Computer Interfaces and communication in paralysis [https://pmc.ncbi.nlm.nih.gov/articles/PMC2644824/]
• 75. Brain–Computer Interface–Based Communication in the ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5283652/]
• 76. Semantic and BCI-performance in completely paralyzed ... [https://www.sciencedirect.com/science/article/abs/pii/S0093934X18301792]
• 77. Brain-Computer Interfaces in Rehabilitation [https://pmc.ncbi.nlm.nih.gov/articles/PMC12381593/]
• 78. Brain computer interfaces are poised to help people with ... [https://www.npr.org/sections/shots-health-news/2025/06/30/nx-s1-5339708/brain-computer-interface-implants-disabilities-neuralink]
• 79. The Bold Frontier of Brain–Computer Integration - IEEE Pulse [https://www.embs.org/pulse/articles/silicon-synapses-the-bold-frontier-of-brain-computer-integration/]
• 80. Brain-computer interface study wins 2025 Top Ten Clinical ... [https://health.ucdavis.edu/news/headlines/brain-computer-interface-study-wins-2025-top-ten-clinical-research-achievement-award-/2025/02]
• 81. Decoding inner speech from brain signals [https://www.nih.gov/news-events/nih-research-matters/decoding-inner-speech-brain-signals]
• 82. Brain-computer-interface-based intervention increases ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12445871/]
• 83. Privacy & Neurodata Privacy: Brain , Consent, and... Data Ownership [https://garanord.md/ownership-consent-and-control-the-ethics-of-brain-data-privacy/]
• 84. Informed Consent in Implantable BCI Research [https://pubmed.ncbi.nlm.nih.gov/26497727/]
• 85. Ethical considerations for the use of brain–computer ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11542783/]
• 86. Brain-Computer Interfaces: Privacy and Ethical ... [https://fpf.org/blog/brain-computer-interfaces-privacy-and-ethical-considerations-for-the-connected-mind/]
• 87. What Rights Do You Have Over Your Brain ? The Future of... Data [https://www.linkedin.com/pulse/what-rights-do-you-have-over-your-brain-data-future-privacy-bekaren-czzzf]
• 88. The ethical and legal landscape of brain governance - PMC data [https://pmc.ncbi.nlm.nih.gov/articles/PMC9799320/]
• 89. Informed consent in implantable BCI research [https://pubmed.ncbi.nlm.nih.gov/27247140/]
• 90. Locked‐in syndrome: A qualitative study of a life story - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC11302814/]
• 91. The Long Search for a Computer That Speaks Your Mind [https://www.wired.com/story/the-long-search-for-a-computer-that-speaks-your-mind/]
• 92. Silent Speech, Loud Questions: The Dawn of Brain ... [https://www.rand.org/pubs/commentary/2025/05/silent-speech-loud-questions-the-dawn-of-brain-computer.html]
• 93. A review on the performance of brain-computer interface ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11297882/]
• 94. Locked-In Syndrome [https://www.ohsu.edu/reknew/locked-syndrome]
• 95. Invasive Brain–Computer Interface for Communication [https://www.mdpi.com/2076-3425/15/4/336]
• 96. How does invasive vs non-invasive BCI compare? [https://synapse.patsnap.com/article/how-does-invasive-vs-non-invasive-bci-compare]
• 97. A mental state aware brain computer interface for adaptive ... [https://www.nature.com/articles/s41598-024-82252-7]
• 98. Unlocking Autonomy with Brain-Computer and Brain-Machine ... [https://currents.neurocriticalcare.org/Leading-Insights/Article/unlocking-autonomy-with-brain-computer-and-brain-machine-interfaces-current-insights-and-future-frontiers]
• 99. The top BCI stories of 2025 so far [https://www.massdevice.com/top-bci-stories-2025-so-far/]
• 100. NeuronLab: BCI framework for the study of biosignals [https://www.sciencedirect.com/science/article/pii/S0925231224007987]
• 101. An open-source human-in-the-loop BCI research framework [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2023.1129362/full]
• 102. Development and Validation of a Tool Based on the DART ... [https://www.sciencedirect.com/science/article/abs/pii/S0003999325005957]