Investigating structural-functional brain covariation in bipolar disorder using a multimodal fusion approach

Due to the lack of consistent findings across different modalities, the neurobiological underpinning of bipolar disorder (BD) remains elusive. This study aims to employ a multimodal fusion algorithm, integrating multimodal imaging data, to unravel the neurobiological underpinning of BD. A data-driven multimodal fusion algorithm was utilized to analyze covariant patterns across modalities in a cohort of 125 BD patients and 113 healthy controls (HCs). The study focused on fusing regional homogeneity (ReHo), gray matter volume (GMV), and fractional anisotropy (FA) derived from MRI scans to generate group-discriminative joint independent components (jIC). That differentiated BD patients from HCs across three modalities. An inverse functional pattern was observed in the default mode network (DMN) and sensorimotor network (SMN) in BD patients, characterized by increased ReHo in the DMN and decreased ReHo in the SMN compared to healthy individuals. This inverse pattern was also mirrored in GMV, showing increase in the DMN and decreases in the SMN. Meanwhile, significant functional hyperactivation coupled with decreased structural volume in the precuneus underscores its role in cognitive function in BD. Multimodal neuroimaging fusion provides a comprehensive understanding in pathophysiology of BD, offering valuable insights that could be pivotal in advancing the diagnosis and treatment of BD.