A magnetoencephalography dataset for motor and cognitive imagery-based brain-computer interface.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Magnetoencephalography (MEG)-based brain-computer interfaces (BCIs) show promise but are hindered by a lack of accessible datasets.
• MEG systems are costly, limiting data availability for developing advanced BCI signal processing algorithms.

Purpose of the Study:

• To address the scarcity of open-source MEG-BCI data.
• To provide a valuable resource for researchers to develop novel machine learning methods for BCI applications.

Main Methods:

• A 306-channel MEG-BCI dataset was recorded at a 1KHz sampling frequency.
• Data collection involved 17 healthy participants performing four distinct mental imagery tasks (hand, feet, subtraction, word generation) across two sessions on separate days.

Main Results:

• The dataset represents the first publicly available MEG imagery BCI dataset.
• It includes high-density MEG recordings during well-defined cognitive and motor imagery tasks.

Conclusions:

• The release of this dataset is expected to accelerate research and development in MEG-BCI systems.
• It will facilitate the creation of more effective pattern recognition and machine learning algorithms for decoding brain activity from MEG signals.