Multimodal Fusion of Functional and Structural Data to Recognize Longitudinal Change Patterns in the Adolescent Brain

Functional and structural magnetic resonance imaging (fMRI/sMRI) are extensively used modalities for studying brain development. While individual modalities may overlook crucial aspects of brain analysis, combining multiple modalities allows us to leverage the benefits of revealing hidden brain connections. To analyze multivariate change patterns in brain function and structure with increasing age across the entire brain, we employ a symmetric multimodal fusion approach that combines multiset canonical correlation analysis and joint independent component analysis. In this study, we present a novel approach to analyze linked longitudinal change patterns in functional network connectivity (FNC) and gray matter (GM) data derived from the large-scale Adolescent Brain and Cognitive Development dataset. Our approach uncovers significant pattern changes in both modalities. Specifically, we identify highly structured functional change patterns and structural change patterns that include increased brain functional connectivity between the visual and sensorimotor domains in the fMRI data, as well as changes in the bilateral sensorimotor cortex in the sMRI data. Overall, our study demonstrates the strength of our approach in uncovering longitudinal changes in FNC and GM, provides valuable insights into the dynamic nature of brain connectivity and structure during adolescence, and sheds light on potential gender-related differences in these processes.