Decoding Complex Cognitive States Online by Manifold Regularization in Real-Time fMRI

Toke Jansen Hansen, Lars Kai Hansen, Kristoffer Hougaard Madsen

Abstract: Human decision making is complex and influenced by many factors on multiple time scales, reflected in the numerous brain networks and connectivity patterns involved as revealed by fMRI. We address mislabeling issues in paradigms involving complex cognition, by considering a manifold regularizing prior for modeling a sequence of neural events leading to a decision. The method is directly applicable for online learning in the context of real-time fMRI, and our experimental results show that the method can efficiently avoid model degeneracy caused by mislabeling.