Arnaud Delorme analyzes large scale EEG data to extract reproducible markers of subjective experience.
New book (2026): The Noetic Particle -- a hard science fiction novel encoding a scientifically grounded neural theory of consciousness. Read more
EEGPrep (2025): A Python implementation of the EEGLAB preprocessing pipeline. Learn more
A systematic comparison of EEG preprocessing pipelines across multiple datasets showed that simple baseline removal and 0.5 Hz high-pass filtering recovered 87% of the statistically significant effects that any pipeline could detect. Complex multi-step pipelines offered diminishing returns while introducing variability. Cited over 305 times, the study prompted a field-wide reassessment of preprocessing practices. Scientific Reports, 2023.
Training deep learning models across EEG datasets is difficult because electrode montages differ between laboratories. This benchmark evaluated four channel-adaptation strategies across five pretrained foundation models and found that Riemannian recentering -- a fixed geometric operation with no learned parameters -- performed on par with learned adaptation methods. Kokate, Aristimunha, Truong & Delorme, arXiv 2026.
The Healthy Brain Network EEG dataset comprises 128-channel recordings from over 3,000 participants across six cognitive tasks, distributed through NEMAR in BIDS format. It served as the basis for the NeurIPS 2025 EEG Foundation Challenge, the largest EEG machine learning competition ever organized with 1,100 competing teams, testing cross-task transferability and psychopathology prediction from brain signals.
An open-source MATLAB and Python environment for EEG analysis created in 2001, combining interactive visualization with scriptable pipelines and independent component analysis for artifact rejection. EEGLAB includes over 120 community-contributed plugins, is used in more than 3,000 laboratories worldwide, and has been cited over 28,500 times, making it one of the most cited tools in neuroscience methods.