急性発症治療抵抗性統合失調症における構造的および機能的異常のマルチモーダル融合解析

Patients with treatment-resistant schizophrenia (TRS) experience more severe clinical symptoms, which may derive from greater structural and functional brain abnormalities compared to non-TRS. However, the neural underpinnings of TRS remain poorly understood, and conventional unimodal neuroimaging approaches are limited in capturing cross-modal interactions. Therefore, we used multimodal fusion via multiset canonical correlation and joint independent component analysis (mCCA+jICA) to integrate MRI data to examine shared and modality-specific features in TRS. Methods65 TRS patients, 65 non-TRS patients, and 54 healthy controls (HCs) underwent 3 T MRI. Structural MRI and resting-state functional MRI data were preprocessed using DPARSFA, and grey matter volume (GMV), fractional amplitude of low-frequency fluctuations (fALFF), and regional homogeneity (ReHo) maps were extracted. The mCCA+jICA was performed to derive joint independent components (ICs) and subject-specific mixing coefficients. Group differences in mixing coefficients were assessed via MANCOVA, and clinical correlations were evaluated using partial correlations.

Patients showed significant differences in mixing coefficients for one modality-specific group-discriminative IC (GMV-IC3) and two joint group-discriminative ICs (ReHo-IC6, and GMV-IC6) compared to HCs. These components also differentiated TRS from non-TRS, with abnormalities concentrated in the precuneus in both ReHo and GMV. Partial correlation analysis revealed significant positive associations between the mixing coefficients of GMV-IC3, GMV-IC6, and total PANSS scores and CPZ equivalents specifically in the TRS group.

This study demonstrates that co-existing structural and functional brain alterations are associated with the pathophysiology of TRS, highlighting their potential as novel multimodal biomarkers.