A streamable large-scale clinical EEG dataset for Deep Learning.

Area of Science:

• Neuroscience
• Biomedical Research
• Machine Learning

Background:

• Deep Learning (DL) is transforming various scientific fields, including biomedical research.
• In neuroscience, DL offers potential for predictive modeling of electrophysiological data, reducing reliance on manual feature engineering.
• Large-scale datasets are essential for developing and validating DL models in this domain.

Purpose of the Study:

• To introduce the first large-scale clinical electroencephalography (EEG) dataset specifically designed for deep learning applications.
• To facilitate easier data access and management for researchers in the neuroscience community.
• To showcase a practical use case and highlight the importance of such neuroinformatics infrastructure.

Main Methods:

• Compiled a large-scale dataset of eyes-closed EEG recordings from 1,574 juvenile participants.
• Data sourced from the Healthy Brain Network.
• Developed a framework to simplify data access and management for DL model experimentation.

Main Results:

• Successfully curated and published a novel, large-scale clinical EEG dataset.
• Demonstrated a functional use case integrating the dataset within a DL framework.
• Established a valuable neuroinformatics resource for the scientific community.

Conclusions:

• The release of this large-scale EEG dataset is critical for advancing deep learning applications in neuroscience.
• Simplified data access and management accelerate research and discovery in electrophysiological neuroimaging.
• This infrastructure is vital for future scientific breakthroughs in understanding brain function and disorders.