Cerebellar structure is abnormal in schizophrenia and deviates from bipolar disorder.

The cerebellum has been implicated in schizophrenia-related structural and functional deficits, with posterior Crus I and II most consistently affected. Source-based morphometry (SBM) involves applying independent component analysis to gray matter volume to identify spatially distinct structural networks that covary across individuals. We applied SBM and voxel-based morphometry (VBM) to cerebellar structural imaging data to identify patterns of gray matter differences across individuals with schizophrenia (SZ), bipolar disorder with psychotic features (BDwP), and healthy controls (HC). Data were drawn from the Psychosis Human Connectome Project (P-HCP) and included 168 participants: 85 with SZ, 36 with BDwP, and 47 HC. T1-weighted images were processed using the ENIGMA Cerebellum Volumetrics Pipeline, and ICA decompositions were performed using the SBM module of the GIFT Toolbox. One independent component (IC) showed a significant diagnostic group effect (p < 0.05, Bonferroni-corrected), differentiating SZ from HC and BDwP. This cerebellar network included vermis VIIIa, bilateral Crus I, and right lobule IX with positive loadings, and bilateral Crus I and lobule IX with negative loadings. Voxel-based morphometry showed reduced GM volume in negatively loaded regions in SZ. Composite cognitive performance correlated with GM volume (r = 0.27, p < 0.001) and network loadings (r = -0.29, p < 0.001). Mediation analyses showed a strong direct diagnostic effect on IC loadings (-0.42), with small, nonsignificant indirect effects via cognition. These findings identify a cerebellar structural network that differentiates schizophrenia from bipolar disorder and controls, underscoring the cerebellum's unique contribution to the neurobiology of schizophrenia.