Knowledge Vault 3/81 - G.TEC BCI & Neurotechnology Spring School 2024 - Day 9
Hemodynamic brain-computer interfaces for functional
brain assessment, communication and neurotherapy
Bettina Sorger, Maastricht University (NL)
<Resume Image >

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

graph LR classDef hemodynamic fill:#f9d4d4, font-weight:bold, font-size:14px; classDef bci fill:#d4f9d4, font-weight:bold, font-size:14px; classDef neurofeedback fill:#d4d4f9, font-weight:bold, font-size:14px; classDef fnirs fill:#f9f9d4, font-weight:bold, font-size:14px; classDef fmri fill:#f9d4f9, font-weight:bold, font-size:14px; classDef future fill:#d4f9f9, font-weight:bold, font-size:14px; A[Bettina Sorger] --> B[Hemodynamic brain computer
interfacing work presented. 1] A --> C[Brain signals for control,
feedback, inference, communication. 2] C --> D[Neurofeedback monitors, modulates
brain activation. 3] A --> E[Hemodynamic BCIs vs
other imaging methods. 4] E --> F[Hemodynamic signals: delayed
but reliable. 5] E --> G[fMRI: spatially distinct
activation patterns. 5] A --> H[Indirect information encoding
in hemodynamic BCIs. 6] H --> I[fNIRS features encode
multiple choice answers. 6] A --> J[fNIRS pioneers showed
consciousness assessment, communication. 7] A --> K[fMRI BCI: spatial,
temporal encoding. 8] A --> L[fMRI letter speller
BCI with tasks. 9] A --> M[fMRI BCI enabled
locked-in patient communication. 10] A --> N[fNIRS: real-world
hemodynamic BCI. 11] A --> O[Ideal clinical BCI
criteria listed. 12] A --> P[fNIRS: optical absorption
of oxy/deoxyhemoglobin. 13] P --> Q[fNIRS: non-invasive, safe,
compact, affordable, robust. 14] P --> R[fNIRS challenges: variability,
noise, resolution. 15-17] A --> S[fNIRS mental task
spatial differentiation accurate. 18] A --> T[Intuitive tasks, shorter
encoding times explored. 19] A --> U[Multi-sensory encoding improved
multiple choice accuracy. 20] A --> V[fNIRS reliable, spatially
corresponds to fMRI. 21] A --> W[Real-time fNIRS BCI
setup steps described. 22-23] A --> X[Online demo: imagined
drawing decoded letter. 24] A --> Y[Potential fNIRS improvements:
resolution, thresholds, modeling. 25-26] A --> Z[fNIRS considerations: artifacts,
no EM interference. 27] A --> AA[Hemodynamic BCIs promising
for real-world applications. 28] AA --> AB[More fNIRS research
needed to realize potential. 29] A --> AC[Live demo illustrated
practicality, room for development. 30] class B,E,F,G,H,I,N,P,Q,R,V,AA,AB hemodynamic; class C,J,K,L,M,S,T,U,W,X,AC bci; class D neurofeedback; class P,Q,R,V,Y,Z,AA,AB fnirs; class G,K,L,M fmri; class AB,AC future;


1.- Bettina Sorger from Maastricht University presented work on hemodynamic brain computer interfacing, focusing on functional brain assessment, neurofeedback therapy, and BCI.

2.- Brain signals can be used for external device control, neurofeedback, and to infer brain functions and communicate answers in disorders of consciousness.

3.- Neurofeedback enables monitoring and modulating brain activation to influence human functioning, as altered brain activation accompanies disease.

4.- Most BCIs use neuroelectric signals, but hemodynamic signals provide an alternative with pros and cons compared to other functional brain imaging methods.

5.- Hemodynamic signals have a delay but high single trial reliability. fMRI shows spatially distinct activation patterns for different mental tasks.

6.- Indirect information encoding is used in hemodynamic BCIs. Spatial and temporal fNIRS signal features can encode multiple choice answers.

7.- fNIRS pioneers showed basic consciousness assessment and yes/no question answering is possible in disorders of consciousness patients unable to behaviorally communicate.

8.- A multiple choice fMRI BCI combined spatial (assigning letters to tasks) and temporal (task timing) features to increase disentanglement of answers.

9.- A letter speller fMRI BCI encoded 27 letters using mental tasks (imagery vs calculation), task duration, and onset delay for each letter.

10.- An fMRI BCI enabled a locked-in patient to communicate multiple choice answers with 100% accuracy using visual or auditory task encoding.

11.- fMRI is not ideal for daily life BCIs. fNIRS provides an alternative hemodynamic approach that is more amenable to real-world use.

12.- Ideal clinical BCIs should be functional, safe, durable, individualized, comfortable, user-friendly, affordable, aesthetically appealing, and usable in daily life.

13.- fNIRS exploits optical absorption properties of oxy/deoxyhemoglobin. Near-infrared light penetrates tissue and probes cortical hemodynamic responses between source-detector pairs.

14.- fNIRS has two dependent variables (oxy/deoxy-Hb), is non-invasive, safe, compact, usable in any position, relatively affordable, and robust to motion.

15.- Challenges of fNIRS include spatial variability in anatomy/function, lack of individual anatomical data, variable signal quality, and non-neural physiological noise.

16.- Short distance fNIRS channels measure extra-cerebral noise to be regressed out, improving signal quality to better isolate true brain activation.

17.- fNIRS has limited depth penetration, spatial coverage, and resolution, especially in sulcal areas, presenting challenges to be addressed.

18.- A study examined the ability to spatially differentiate mental tasks with fNIRS and found high accuracy but inter-subject variability.

19.- Another study explored shorter encoding times and more intuitive tasks (e.g. imagined drawing checkmarks for yes) and found 68% single-trial accuracy.

20.- Multi-sensory encoding of multiple choice answers using motor imagery timed to visual, auditory or tactile letter cues enabled 85% accuracy.

21.- Oxygenated and deoxygenated fNIRS signals show high single-trial reliability and spatial correspondence to fMRI in an experienced user's motor imagery data.

22.- Real-time fNIRS BCI setup involves cap placement, calibration, and checking signal quality before a localizer run to select optimal channels.

23.- Deoxyhemoglobin from a selected channel over motor cortex clearly shows the expected response during cued periods of hand motor imagery.

24.- In an online demo, imagined drawing was detected on single trials and used to unambiguously decode the letter "B" in a multiple-choice question.

25.- Swapping fNIRS source/detector positions could improve spatial resolution. Signal quality thresholds depend on the speed vs averaging needs of the application.

26.- Clarity of fNIRS signals in motor disorders is unknown but plausible. Individualized modeling of the expected hemodynamic response may help.

27.- External light artifacts affect fNIRS, requiring shielding. Electromagnetic interference is not an issue unlike with electrical brain signals.

28.- In summary, hemodynamic BCIs, especially fNIRS, show promise for real-world functional brain assessment, neurofeedback therapy, and communication/control applications.

29.- More research is needed on fNIRS signal quality, spatial resolution, artifact removal, and clinical populations to realize the full potential.

30.- The engaging live demo showed online fNIRS BCI at work, illustrating the technology's practicality while leaving room for further development.

Knowledge Vault built byDavid Vivancos 2024