Knowledge Vault 3/79 - G.TEC BCI & Neurotechnology Spring School 2024 - Day 9
BCI for the assessment, prediction, communication &
rehabilitation of patients with disorders of consciousness
Christoph Guger, g.tec medical engineering GmbH (AT)
<Resume Image >

Concept Graph & Resume using Claude 3 Opus | Chat GPT4 | Llama 3:

graph LR classDef mindbeagle fill:#f9d4d4, font-weight:bold, font-size:14px; classDef doc fill:#d4f9d4, font-weight:bold, font-size:14px; classDef paradigms fill:#d4d4f9, font-weight:bold, font-size:14px; classDef studies fill:#f9f9d4, font-weight:bold, font-size:14px; classDef guidelines fill:#f9d4f9, font-weight:bold, font-size:14px; A[Christoph Guger] --> B[MindBeagle: assesses DOC,
enables communication. 1] A --> C[DOC: coma, UWS, MCS,
LIS, CLIS. 2] B --> D[MindBeagle: auditory, vibrotactile,
motor imagery paradigms. 3] D --> E[Auditory oddball: P300 indicates
command following. 4] D --> F[Vibrotactile: yes/no questions,
counting vibrations. 5] D --> G[Motor imagery: left/right hand
movement, EEG desynchronization. 6] D --> H[Healthy controls: high accuracy,
brief recordings. 7] B --> I[Recorded data: UWS, MCS,
LIS, CLIS, ALS patients. 8] I --> J[UWS: fluctuations, some
achieve communication. 9] I --> K[MCS, LIS: clear P300,
high accuracies. 10] I --> L[No clear ERPs, but
BCIs detect responses. 11] I --> M[LIS study: high vibrotactile
accuracy, communication. 12] I --> N[UWS study: fluctuations, accuracy
predicts recovery. 13] N --> O[Vibrotactile accuracy predicts
6-month outcome. 14] I --> P[Vibrotactile paradigm: rehabilitation,
CRS-R improvement. 15] B --> Q[Guidelines: best time, meds,
language, family, repetition. 16] B --> R[MindBeagle: command following, prediction,
rehabilitation, brief tests. 17] B --> S[Resting EEG biomarkers, stimulation,
FES in Recoveriix. 18] B --> T[ERP significance: targets vs
non-targets, time points. 19] I --> U[CLIS: yes/no communication,
but not full. 20] D --> V[Vibrotactile: short tests,
patients fall asleep. 21] D --> W[Motor imagery: LDA classification,
topography validates activations. 22] B --> X[Small DOC market, ethically
and scientifically important. 23] B --> Y[Principles extend: stroke, Alzheimer's,
Parkinson's, concussion. 24] B --> Z[Biomarkers: cortex/brain health
in recovery. 25] D --> AA[Auditory P300: optimize target/non-target
ratio, patient's name. 26] I --> AB[Videos: vibrotactile protocol, therapist
instructions, patient impairments. 27] I --> AC[Healthy control data: benchmarks
for patient comparison. 28] I --> AD[MindBeagle locked-in study:
most viewed paper. 29] B --> AE[MindBeagle: quick, objective detection,
communication, prediction, rehab. 30] class A,B,Q,R,S,T,X,Y,Z,AE mindbeagle; class C doc; class D,E,F,G,H,V,W,AA paradigms; class I,J,K,L,M,N,O,P,U,AB,AC,AD studies; class Q guidelines;


1.- MindBeagle system assesses disorders of consciousness patients to identify command following and enable yes/no communication using brain-computer interfaces.

2.- Disorders of consciousness include coma, unresponsive wakefulness syndrome (UWS), minimally conscious state (MCS), locked-in syndrome (LIS), and completely locked-in state (CLIS).

3.- MindBeagle uses auditory, vibrotactile, and motor imagery paradigms with active EEG electrodes to get clean data in brief recording times.

4.- Auditory oddball paradigm with high/low tones elicits P300 response if patient can discriminate the tones, indicating command following.

5.- Vibrotactile stimulators on each hand allow yes/no questions - counting vibrations on one hand for yes, other for no.

6.- Motor imagery of left/right hand movement assessed by event-related desynchronization in EEG, enabling BCI communication in some patients.

7.- Healthy controls achieve near 100% accuracy in vibrotactile and 80-90% in motor imagery paradigms in 2-8 minute recordings.

8.- Recorded data from collaborators in Palermo, Liege, Nimes, Poland on UWS, MCS, LIS, CLIS and ALS patients.

9.- Some UWS patients exhibit fluctuations - no response in one session but significant accuracy and yes/no communication in another.

10.- MCS and LIS patients often have clear P300 responses and high accuracies like healthy controls if able to do task.

11.- Lack of clear ERPs in clinical recordings, but BCIs can still detect responses and enable communication in many patients.

12.- Study with 12 LIS patients - most achieved 100% vibrotactile accuracy and 8/10 correct yes/no responses. MI worked for 3 patients.

13.- Study with 12 UWS patients showed fluctuations across days, some achieved communication; accuracy predicts 6-month outcome/recovery.

14.- Vibrotactile accuracy >40% predicted UWS patients would improve on CRS-R scale 6 months later; low accuracy predicted decline/death.

15.- Vibrotactile paradigm used as rehabilitation - 100 sessions over 10 days; 10/20 UWS/MCS patients improved on CRS-R.

16.- Guidelines: assess at patient's best time, consider meds, use native language, involve family. Repeat to catch fluctuations.

17.- MindBeagle provides objective command following detection, outcome prediction, and rehabilitation for these patients in brief tests.

18.- Also acquiring resting state EEG for biomarkers; can combine with prefrontal DC stimulation and FES used in Recoveriix.

19.- Statistical significance of ERPs calculated between targets and non-targets at each time point. Rupert Ortner will explain in more detail.

20.- Communication beyond yes/no is still limited for CLIS; 2/3 CLIS patients answered yes/no but not full communication.

21.- Vibrotactile tests kept short (2.5min) as patients fall asleep; 8min motor imagery test works least well, likely due to duration.

22.- Motor imagery uses standard linear discriminant analysis for classification; topography helps validate expected activations but patients may use other strategies.

23.- Small market for DOC but ethically and scientifically important. Principles can extend to stroke, Alzheimer's, Parkinson's, concussion.

24.- Biomarkers help track if cortex/brain is getting healthier in recovery. Quick objective brain health measurement has broad uses.

25.- Auditory P300 can be optimized by adjusting target/non-target ratio and using patient's own name to speed up recording.

26.- Video examples showed the vibrotactile yes/no communication protocol in action with therapists instructing patients in native language.

27.- Video of patient attempts to lift arm, showing the severe motor impairments in this population that necessitate BCI.

28.- Data examples from healthy controls provide benchmarks to compare patient performance and set expectations for signals in non-responsive patients.

29.- Paper on MindBeagle locked-in study was the most viewed Frontiers paper, demonstrating the high interest and need.

30.- MindBeagle provides a quick, objective way to detect awareness, enable communication, predict recovery, and provide rehab for DOC patients.

Knowledge Vault built byDavid Vivancos 2024