The End Of Knowledge - Vault 3 - G.TEC BCI & Neurotechnology Spring School 2024 - Day 5

graph LR classDef auditory fill:#f9d4d4, font-weight:bold, font-size:14px; classDef visual fill:#d4f9d4, font-weight:bold, font-size:14px; classDef tactile fill:#d4d4f9, font-weight:bold, font-size:14px; classDef motor fill:#f9f9d4, font-weight:bold, font-size:14px; classDef clinical fill:#f9d4f9, font-weight:bold, font-size:14px; A[Jing Jin] --> B[Auditory BCI aids video-impaired
device control. 1] B --> C[Robot uses auditory BCI
for tracking. 2] B --> D[Three pure tones: high,
middle, low. 3] B --> E[Auditory BCI-robot tested indoors,
face tracking. 4] B --> F[Multiple trials: accuracy nearly
80%, some 100%. 5] B --> G[HLM outperformed DS in
accuracy comparison. 6] B --> H[HLM preferred: feedback optimized
auditory stimuli. 7] A --> I[Nature sounds explored as
pleasant stimuli. 8] I --> J[Drip drops elicited comparable
ERPs to beeps. 9] I --> K[Percussion with/without music
tested: music preferred. 10] A --> L[Minor visual changes engage,
reduce boredom. 11] L --> M[Visual mismatch negativity improved
target/non-target distinction. 12] L --> N[Honeycomb figures increased concentration,
classification accuracy. 13] L --> O[Multi-phase paradigm stabilized P300,
reduced errors. 14] A --> P[ErrP detected, prevented incorrect
P300 outputs. 15] A --> Q[Vibrotactile stimuli for hemiplegic
BCI: bilateral excelled. 16] Q --> R[Cheek-based tactile BCI outperformed
wrist-based. 17] A --> S[Calibration optimized feature extraction,
improved accuracy. 18] S --> T[Covariance matrix augmented motor
imagery performance. 19] S --> U[Genetic algorithm shortened calibration
by 70%. 20] S --> V[Chinese characters elicited stronger
ERD than arrows. 21] A --> W[Stroke minimally affected BCI:
helped limb excelled. 22] W --> X[Motor BCI with FES
improved stroke rehabilitation. 23] A --> Y[P300 BCI speller accurate
in ALS patients. 24] A --> Z[Over 100 stroke patients
used BCI rehabilitation. 25] A --> AA[BCI developed for consciousness
assessment, cognitive training. 26] A --> AB[BCI products obtained medical
device certifications. 27] A --> AC[Videos show BCI-trained stroke
patients' functional gains. 28] A --> AD[Active Chinese BCI community:
conferences since 2006. 29] A --> AE[Future work: algorithms, populations,
user experience. 30] class A,B,C,D,E,F,G,H,I,J,K auditory; class L,M,N,O,P visual; class Q,R tactile; class S,T,U,V,W,X,Y motor; class Z,AA,AB,AC,AD,AE clinical;

Resume:

1.- Auditory BCI helps video-impaired people control devices. Optimizing with intelligent methods improves efficiency and flexibility.

2.- Robot control system with auditory BCI uses automatic search and face recognition to track faces, compensating for video defects.

3.- Three pure tones used as stimuli in auditory BCI: high (1000 Hz), middle (800 Hz), low (200 Hz) frequencies.

4.- Auditory BCI with robot control tested indoors. Subjects trained, then controlled robot to recognize and track faces.

5.- Averaging multiple trials improved classification accuracy close to 80%. Some subjects reached 100% accuracy controlling the robot.

6.- Two auditory paradigms compared - HLM (high/low/middle tones) and DS (do-re-mi tones). HLM performed better in accuracy.

7.- HLM tones were preferred by subjects over DS tones. Feedback used to optimize the auditory stimuli.

8.- Nature sounds like drip drops explored as more pleasant stimuli than beeps in auditory BCI.

9.- Drip drop stimuli intercepted from music clips at different spatial locations. Elicited comparable ERPs to beeps.

10.- Percussion instruments with/without background music tested as auditory BCI stimuli. Music preferred by subjects.

11.- Minor visual changes causing large contrasts used to design engaging visual BCI paradigms and reduce boredom.

12.- Visual mismatch negativity paradigm elicited larger N200/N400 ERPs than single image paradigm, improving target/non-target distinction.

13.- Honeycomb-shaped figures with random red dots used as visual stimuli to increase concentration and classification accuracy.

14.- Multi-phase paradigm with stable flash patterns explored to elicit more stable P300 and reduce errors in visual BCI.

15.- Error-related potential (ErrP) used to automatically detect and prevent incorrect P300 BCI outputs for ALS patients.

16.- Vibrotactile stimuli to left/right forearm and calf used for BCI in hemiplegic patients. Bilateral performed better than unilateral.

17.- Cheek-based tactile BCI developed as cheeks have lower tactile thresholds. Outperformed traditional wrist-based paradigms.

18.- Calibration methods proposed to optimize time windows for feature extraction in motor imagery BCI, improving accuracy.

19.- Covariance matrix calculation methods used to augment motor imagery features and classifier performance from limited data.

20.- Genetic algorithm used to shorten calibration time by 70% in P300 BCI while maintaining accuracy compared to standard calibration.

21.- Chinese character writing paradigm elicited stronger ERD patterns than arrows in motor imagery BCI for Chinese subjects.

22.- Motor imagery BCI performance not significantly affected by stroke. Helped limb may lead to better performance.

23.- Motor imagery BCI with FES applied for stroke rehabilitation, showing functional improvements over control group after training.

24.- P300 BCI speller successfully applied in late-stage ALS patients with high accuracy. Long-term stability needs further study.

25.- Over 100 stroke patients used motor BCI rehabilitation systems across hospitals in China with promising results.

26.- BCI systems developed for consciousness assessment and cognitive training in disorders of consciousness patients.

27.- Several medical device certifications obtained in China and Europe for the BCI products and systems developed.

28.- Video examples show stroke patients regaining limb function and performing daily tasks after motor BCI rehabilitation training.

29.- Active BCI research community in China with hundreds of researchers and regular domestic conferences held since 2006.

30.- Future work to focus on new motor imagery algorithms, expanding to more patient populations, improving user experience.

Knowledge Vault built byDavid Vivancos 2024