EEG datasets for motor imagery brain-computer interface.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interface (BCI) research predominantly focuses on induced cortical activity, not evoked activity.
• Motor imagery (MI)-based BCI, a standard approach, faces significant performance variations across sessions and subjects.
• Investigating these performance variations is crucial for advancing MI BCI technology.

Purpose of the Study:

• To present comprehensive EEG datasets for MI BCI research from 52 subjects.
• To provide metadata including psychological/physiological questionnaires, EMG data, and 3D EEG electrode locations.
• To facilitate the investigation of factors contributing to MI BCI performance variability.

Main Methods:

• EEG datasets were validated using bad trial rejection, event-related desynchronization/synchronization (ERD/ERS) analysis, and classification.
• Analysis focused on identifying known MI patterns, such as contralateral ERD and ipsilateral ERS in the somatosensory cortex.
• Performance was assessed by the discriminative capability of the datasets.

Main Results:

• Conventional bad trial rejection was applied to the EEG data.
• Contralateral ERD and ipsilateral ERS patterns, characteristic of MI, were observed.
• A significant portion of the datasets, 73.08% (38 subjects), contained reasonably discriminative information.

Conclusions:

• The presented EEG datasets include both well- and less-discriminative data, enabling statistical significance analysis.
• These datasets can support research into human factors influencing MI BCI performance variation.
• Metadata facilitates potential subject-to-subject transfer learning for MI BCI applications.