BIDSAlign: a library for automatic merging and preprocessing of multiple EEG repositories.

Area of Science:

• Neuroscience
• Data Science
• Machine Learning

Background:

• Electroencephalography (EEG) data analysis faces challenges due to data heterogeneity.
• Lack of large, standardized EEG datasets hinders the development of robust deep learning models.
• Existing tools often struggle with diverse data formats and preprocessing pipelines.

Purpose of the Study:

• To introduce BIDSAlign, a standardized library for processing and merging heterogeneous EEG datasets.
• To create an environment for preprocessing public EEG datasets to train deep learning architectures.
• To facilitate the use of public EEG data for clinical and non-clinical research.

Main Methods:

• BIDSAlign handles both Brain Imaging Data Structure (BIDS) and non-BIDS datasets.
• It unifies EEG recordings via a common pipeline and channel template.
• Includes visualization functions and a graphical user interface for accessibility.

Main Results:

• BIDSAlign effectively processes public EEG datasets from sources like OpenNeuro.
• Enables extraction of significant medical insights through an end-to-end workflow.
• Facilitates group analysis, visual comparison, and statistical testing for non-experts.

Conclusions:

• BIDSAlign addresses the scarcity of large EEG datasets by aligning data to a standard template.
• Unlocks the potential of public EEG data for training deep learning models.
• Promotes advancements in EEG research for neurological disease diagnosis and treatment.