Heterogeneity characterization of psychiatric diseases based on common orthogonal basis extraction

Psychiatric disorders exhibit significant heterogeneity. The systematic identification of individual differences in neurofunctional deficits can inform both diagnosis and individual intervention strategies. In this study, we investigated whether individual brain functional connectivity (FC) of electroencephalogram (EEG) could define more predictable features of clinical indicators of psychiatric disorders. Specifically, the common orthogonal basis extraction (COBE) algorithm was employed to isolate individual-specific components by eliminating common components, and then a prediction model was constructed using connectome-based predictive modeling (CPM) and Lasso regression. Our findings indicate that the definition of individual-specific FC significantly improves the prediction performance compared to original FC-based models. Furthermore, the relationship between functional subnetworks and the Positive and Negative Syndrome Scale (PANSS) score was investigated, which revealed different interaction patterns among schizophrenia, bipolar disorder, and schizoaffective disorder. Our results suggest that individual-specific FC enhances the predictive performance compared to the original FC by removing common components. These findings may shed more light on disease heterogeneity and provide enhanced methods for leveraging EEG FC for individual treatment selection, thereby advancing precision medicine in psychiatric disorders.