MPIDS: Multi-Group Perturbation Based Individualized Differential Structural Brain Network

MRI-based brain structural network plays a key role in understanding the organization architecture of the human brain. However, current structural brain networks are inferred only at the group-level, missing individual heterogeneity at the single subject level. In this study, we propose a multi-group perturbation based individualized differential structural brain network (MPIDS), which reveals subject-specific deviations from a reference group in structural correlations among brain regions, enabling precision capture of individual differences. Each group is sequentially used as the reference network, with perturbations introduced by incorporating subjects from other groups. Results show that the proposed MPIDS consistently achieve the highest accuracy in predicting cognition across two datasets, three brain atlases, three prediction models comparing with voxel-based and region-based methods. Moreover, in diagnosis of SZ and ASD comparing to HC, MPIDS robustly outperformed voxel-based and region-based methods across three datasets, three brain atlases, and three classifiers. Overall, results demonstrate the effectiveness of the proposed MPIDS in both prediction and classification, which can be used to investigate the heterogeneity of brain disorders.