EEG-BCI Dataset for Motor Imagery and Overt Spatial Attention EEG-BCI Control

This dataset consists of EEG recordings and Brain-Computer Interface (BCI) data from 25 different human subjects performing BCI experiments. More information can be found in the corresponding manuscript: 

Dylan Forenzo, Yixuan Liu, Jeehyun Kim, Yidan Ding, Taehyung Yoon, Bin He: “Integrating Simultaneous Motor Imagery and Spatial Attention for EEG-BCI Control”, IEEE Transactions on Biomedical Engineering (10.1109/TBME.2023.3298957).

Please cite this paper if you use any data included in this dataset. 

The dataset was collected under the support of NIH grants AT009263, EB021027, EB029354, NS096761,

NS124564 to Dr. Bin He at Carnegie Mellon University.

Each file is a MATLAB object (.mat file) which contains data from a single run of BCI control. The MATLAB files are grouped into folders based on the Subject, one for each of the 25 subjects studied. Each subject completed 5 sessions of BCI experiments and each session consisted of either 18 (sessions 1 and 2) or 15 (sessions 3-5) runs, for a total of 81 runs per subject or 2025 total BCI runs.

Each of the MATLAB files contains a single structure with the following fields:

data: An array containing the EEG recordings with the size (channels x time points)

times: A vector containing the timestamps in milisceonds with the size (1 x time points)

fs: sampling frequency (1000 Hz)

labels: A cell array containing the label for each channel

targets: A list of target codes. For LR: 1 is right, 2 is left. For UD: 1 is up, 2 is down. For 2D: 1 is right, 2 is left, 3 is up, and 4 is down

event: A structure of events from BCI2000. Each index corresponds to the start of a trial and includes the time (latency) of when the trial starts, and how long each trial lasted (duration).

results: A vector of which target was hit for each trial (0 if the trial was aborted before a target was hit)

outcome: A vector indicating the outcome of each trail (1: hit, 0: abort, -1: miss)

subject: The coded subject number

session: The session number. Please note that the session numbers are for specific tasks, so even though 2D sessions began on the third day of experiments, the 2D runs are listed as session 1, 2, and 3 as they are the first, second, and third 2D sessions.

axis: The axis of control. Either LR (horizontal only, Left-Right), UD (vertical only, Up-Down), or 2D (both horizontal and vertical control).

task: The control paradigm used. Options are MI (motor imagery), OSA (overt spatial attention), MIOSA (MI and OSA together), MIOSA1 (MI controls horizontal axis, OSA controls vertical. Referred to as MI/OSA in the paper), or MIOSA2 (MI controls vertical axis, OSA controls horizontal. Referred to as OSA/MI in the paper).

run: The run number