Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Fast Whole-Brain, Geometry-Aware Functional Alignment for Cross-Subject Decoding

Pierre-Louis Barbarant¹, Florent Meyniel², Bertrand Thirion³; ¹INRIA, ²CEA, ³Karavela

Presenter: Pierre-Louis Barbarant

Decoding brain activity is useful for characterizing brain processes and understanding the functional architecture underlying cognition. However, variability in brain response patterns among individuals poses a significant challenge to developing powerful, generalizable decoders. To build these decoders, we create population-level models of brain responses based on functional alignment. The core issue is to strike the right balance between sensitivity to functional features and anatomical consistency while maintaining computational efficiency. We introduce a new functional alignment method that embeds cortical geometry directly into the functional domain using Laplace-Beltrami eigenmodes. Our approach achieves higher decoding accuracy than anatomical alignment and converges several orders of magnitude faster than state-of-the-art methods at the whole-brain scale.

Topic Area: Methods, Tools, Theory & Neural Coding