Feature similarity: a sensitive method to capture the functional interaction of brain regions and networks to support flexible behavior

Abstract

The brain is a dynamic system where complex behaviours emerge from interactions across distributed regions. Accurately linking brain function to cognition requires tools that are sensitive to these dynamics. We introduce a novel technique, Feature Similarity (FS), which represents a conceptual advancement in measuring brain region interactions by integrating multiple features across various measures, moving beyond traditional methods that focus on a single measure. Our results show that FS can capture functional brain organisation: regions within the same functional network have greater FS compared to those in different networks, and FS also identifies the principal gradient that spans from unimodal to transmodal cortices. FS demonstrated greater sensitivity to task modulation than traditional functional connectivity (FC) and 46 out of 49 statistical measures of pairwise interactions (SPIs). Specifically, FS reveals interaction patterns missed by FC and most SPIs, such as a double dissociation in Dorsal Attention Network: greater interaction with Visual network during working memory tasks and greater interaction with default mode network during long-term memory tasks. These findings position FS as a powerful tool for capturing task-specific brain network dynamics and advancing our understanding of cognitive flexibility. Using FS to reanalyse existing fMRI data could resolve inconsistencies, lead to novel insights, and prompt revisions of existing theories or the development of new theoretical frameworks.