EEG-DaSh: an open data, tool, and compute resource for machine learning on neuroelectromagnetic data

Résumé fourni par la source

EEG-DaSh (EEGDash) is an open-access Python library and catalog that makes 700+ MEEG (EEG, MEG, iEEG, fNIRS, EMG) datasets immediately usable for machine learning and deep learning. It aggregates recordings from OpenNeuro, NEMAR, Zenodo, Figshare, SciDB, OSF, DataRN, and EEGManyLabs behind a single searchable catalog and a uniform EEGDashDataset interface compatible with MNE-Python, braindecode, and PyTorch. EEG-DaSh addresses three bottlenecks: automated BIDS repair, conversion to a PyTorch-native data format for direct use in training loops, and a feature-extraction framework with extractors across six categories (signal, spectral, connectivity, complexity, dimensionality, spatial filtering). All datasets are annotated with Hierarchical Event Descriptors (HED) for cross-study aggregation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.