Increasing accessibility to a large brain-computer interface dataset: Curation of physionet EEG motor...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Reliable brain-computer interfaces (BCI) depend on accurate decoding of electroencephalography (EEG) signals.
• Model calibration for motor imagery (MI) BCIs requires extensive, well-processed EEG data.
• The PhysioNet EEG Motor Movement/Imagery Dataset is large but underutilized for MI decoding.

Purpose of the Study:

• To curate and clean the PhysioNet EEG Motor Movement/Imagery Dataset for enhanced accessibility.
• To facilitate quicker exploitation, decoding, and classification of MI trials.
• To support advancements in EEG-based MI-BCI development and real-life applications.

Main Methods:

• Curated and cleaned the PhysioNet EEG Motor Movement/Imagery Dataset.
• Excluded six subjects with anomalous EEG recordings.
• Pre-processed data from 103 subjects across eight tasks (four MI, four motor execution).
• Coded task annotations numerically.
• Stored the dataset in MATLAB structure and CSV formats.

Main Results:

• Created an accessible and organized dataset from the PhysioNet EEG Motor Movement/Imagery data.
• The processed dataset includes 103 subjects and eight distinct motor tasks.
• Data is available in convenient MATLAB and CSV formats for immediate use.

Conclusions:

• Enhanced accessibility of this curated EEG dataset will accelerate research in EEG-based MI-BCI decoding.
• Easier data access is expected to improve cross-subject classification and transfer learning.
• This work aims to enable more reliable and practical real-life BCI applications.