Reproduce “Neural decoding from stereotactic EEG: accounting for electrode variability across subjects” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Georgios Mentzelopoulos, Evangelos Chatzipantazis, Ashwin G. Ramayya, Michelle J. Hedlund, Vivek P. Buch, Kostas Daniilidis, Konrad P. Kording, Flavia Vitale

Researchers at the University of Pennsylvania and Stanford University developed 'seegnificant', a deep learning framework to enable multi-subject neural decoding from sEEG data despite electrode variability. The framework improved decoding performance to R² = 0.54 when trained on pooled data from multiple subjects and demonstrated transfer learning capabilities for new individuals.

Key insights

  • A specially designed deep learning architecture can effectively overcome the challenges of variable electrode placement in sEEG data to enable robust multi-subject neural decoding.
  • Pooling sEEG data across multiple individuals allows for the learning of more generalized neural representations, leading to improved decoding performance compared to single-subject models.
  • Incorporating explicit spatial information about electrode locations, such as through positional encodings and spatial attention, is crucial for successful multi-subject sEEG decoding.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.