No Description

Rafael Grigoryan 45bd1caf39 Update 'datacite.yml' 4 years ago
sourcedata 7c38beade4 gin commit from PC-2 4 years ago
sub-10 7c38beade4 gin commit from PC-2 4 years ago
sub-11 7c38beade4 gin commit from PC-2 4 years ago
sub-12 7c38beade4 gin commit from PC-2 4 years ago
sub-13 7c38beade4 gin commit from PC-2 4 years ago
sub-14 7c38beade4 gin commit from PC-2 4 years ago
sub-15 7c38beade4 gin commit from PC-2 4 years ago
sub-16 7c38beade4 gin commit from PC-2 4 years ago
sub-17 7c38beade4 gin commit from PC-2 4 years ago
sub-18 7c38beade4 gin commit from PC-2 4 years ago
sub-19 7c38beade4 gin commit from PC-2 4 years ago
sub-20 7c38beade4 gin commit from PC-2 4 years ago
sub-21 7c38beade4 gin commit from PC-2 4 years ago
sub-22 7c38beade4 gin commit from PC-2 4 years ago
sub-23 7c38beade4 gin commit from PC-2 4 years ago
sub-24 7c38beade4 gin commit from PC-2 4 years ago
sub-25 7c38beade4 gin commit from PC-2 4 years ago
sub-26 7c38beade4 gin commit from PC-2 4 years ago
sub-6 7c38beade4 gin commit from PC-2 4 years ago
sub-8 7c38beade4 gin commit from PC-2 4 years ago
sub-9 7c38beade4 gin commit from PC-2 4 years ago
LICENSE 40f3dc99f8 Add 'LICENSE' 4 years ago
README.md eedb9d8ff1 Update 'README.md' 4 years ago
datacite.yml 45bd1caf39 Update 'datacite.yml' 4 years ago
dataset_description.json 7c38beade4 gin commit from PC-2 4 years ago
participants.json 7c38beade4 gin commit from PC-2 4 years ago
participants.tsv 7c38beade4 gin commit from PC-2 4 years ago
task-bci_events.json 7c38beade4 gin commit from PC-2 4 years ago

README.md

Braille BCI EEG experiment

Brief experiment description

10 blind and 10 sighted subjects were using tactile P300 BCI, implemented with Alva 640 comfort Braille display.

4 fingers on each hand were placed on separate Braille cells, serving as stimuli. Each stimulus was activated 10 times in random order. Participant's task was to count activations of target cell.

For blind subjects, Braille proficiency was assessed with the reading test before the experiment. The experiment included BCI sessions with large (8 dots) and small (1 dot) stimuli.

The present analysis is focused on the comparison of BCI performance for large and small stimuli in blind and sighted individuals.

For more details, see our preprint at https://doi.org/10.1101/2020.06.16.155796

Complete analysis is available at https://github.com/eegdude/erp_analysis/tree/brlbci The scripts are designed to work with contents of sourcedata folder

Sourcedata description

Per-record markup is provided in sourcedata. Per-epoch markup is calculated at preprocessing step. Various stimuli parameters for each epoch are descriped in events.tsv files

Per-record markup

column name description
user user id
folder folder with a record inside the user folder
reg BCI mode. 'large' for large stimuli, 'small' for small
targets list of target stimuli for all learning and stimuli selection cycles
fingers which finger corresponds to specific stimuli code.
right hand 0 - pinky, 1 - ring, 2 - middle, 3 - index
left hand 4 - index, 5 - middle, 6 - ring, 7 - pinky
ecg_r_peak_direction 1 if R-peak is up, -1 if R-peak is down, 0 or 2 if no ECG
leading_hand 'r' if right, 'l' if left
ignore_events_id list of events to not include into analysis
reading_finger list of fingers used for Braille reading
blind 1 if blind, 0 if sighted
order index of the record in the experiment. Failed sessions that have not made it into the dataset are counted.
blindness_age age of complete vision loss
reading_time time of reading Braille text of standard informed consent
braille_display_user 1 if the subject uses Braille display in daily life
daily_braille_time self-reported daily Braille usage, hours
age user's age
music 1 if the subject has musical experience (except vocal), 0 if not
remaining_vision 1 if the user has residual vision, 0 if not
sex male or female
blind_years years passed from total vision loss
congenitaly_blind 1 if congenital blindness, 0 if acquired

Per-epoch markup

The preprocessed dataset has some added columns, which are calculated during the preprocessing stage.

column name description
id unique epoch id
event epoch event (activated stimuli id)
target id of current target stimuli
is_target 1 if epoch is target, 0 if not
session_id stimuli selection cycle index inside experiment
epoch_id epoch index inside stimuli selection cycle (from 0 to 79)
ms_after_r milliseconds after R-peak
ms_before_r milliseconds before next R-peak

Sourcedata folder structure

\
└── markup.csv                               # metadata for all records
└── user 
    └── record
        └── _data_$MODE__play_$TIME.npy      # Raw EEG data
        └── _events_$MODE__play_$TIME.npy    # BCI events
        └── _photocell_$MODE__play_$TIME.npy # sync events
        └── Rpeaks.npy                       # preprocessed R-peaks, if availible

users are numbered from 6 to 26

$MODE can be 'large' for large stimuli and 'small' for small stimuli

Preprocessed data folder structure

\
└── markup.csv                               # metadata for all epochs, created at preprocessing stage
└── 0.pickle                                 # pickle files, corresponding to single epochs

...

└── $N.pickle
datacite.yml
Title Difference in somatosensory event-related potentials in the blind subjects leads to better performance in tactile P300 BCI
Authors Grigoryan,Rafael;Faculty of Biology, Lomonosov Moscow State University, Moscow, Russia;ORCID:0000-0002-9843-3589
Goranskaya,Dariya;ANO Laboratory 'Sensor-Tech', Moscow, Russia
Demchinsky,Andrey;ANO Laboratory 'Sensor-Tech', Moscow, Russia
Ryabova,Ksenia;ANO Laboratory 'Sensor-Tech', Moscow, Russia
Kuleshov,Denis;ANO Laboratory 'Sensor-Tech', Moscow, Russia
Kaplan,Alexander;Faculty of Biology, Lomonosov Moscow State University, Moscow, Russia
Description In this study, we have created an 8-command P300 tactile BCI with two stimuli types, running on a minimally modified consumer Braille display and tested it on 10 blind subjects and 10 sighted controls. Blind subjects have demonstrated 27% higher median accuracy than sighted controls (p<0.05), proving that the blind subjects are not only able to use tactile BCI but also can achieve superior results in comparison with sighted subjects. Median accuracy in the blind group with the best stimuli type has reached 95%. The difference in event-related potentials between groups is located in frontocentral sites before 300 ms post-stimulus and corresponds with early cognitive ERP components. The blind subjects have higher amplitude and lower latency of ERPs. This result is consistent through experimental conditions with different tactile stimuli. The classification performance for the blind subjects is correlated with Braille reading speed. This enables a discussion about mechanisms of plastic changes during sensory compensation after vision loss and its dependence on personal perceptual experience.
License Creative Commons CC0 1.0 Public Domain Dedication (https://creativecommons.org/publicdomain/zero/1.0/)
References
Funding
Keywords Neuroscience
Brain-computer interface
P300
Event-related potentials
Tactile
Blindness
BCI
Resource Type Dataset