An Epicenter-guided computational framework for personalized target of repetitive transcranial magnetic stimulation in schizophrenia.

BACKGROUND: The clinical efficacy of repetitive transcranial magnetic stimulation (rTMS) in schizophrenia is limited by significant inter-individual heterogeneity, largely due to the lack of personalized targeting. HYPOTHESIS: To develop and validate an epicenter-guided computational framework that identifies personalized rTMS targets by integrating individual neuroanatomical and connectome profiles. STUDY DESIGN: We established an epicenter-guided framework in a discovery cohort including 532 schizophrenia and 526 healthy controls. The pipeline involved: (1) identifying subject-specific pathological epicenters by correlating regional gray-matter volume reductions with normative functional connectivity profiles; (2) mapping the voxels within the left dorsolateral prefrontal cortex that exhibited maximal functional connectivity to these epicenters; and (3) computing the harmonic centroid of these voxels to define the optimal personalized target. The framework was retrospectively validated in a cohort of schizophrenia receiving a 4-week rTMS treatment. STUDY RESULTS: The epicenter-guided framework defined a personalized candidate target for each participant with an identifiable individual epicenter. Individual epicenters converged within higher-order association cortices and subcortical structures, consistent with previous findings in schizophrenia. In the retrospective treatment cohort, nodal integration capability within individual epicenter regions increased significantly after rTMS. Smaller target distance was associated with greater concordance between activity-flow-predicted and observed FC changes and with greater improvement in PANSS negative symptoms. CONCLUSIONS: We developed an epicenter-guided framework to derive personalized DLPFC targets and, in retrospective validation, found that target proximity relates to both FC normalization and symptom improvement. These findings support further prospective testing of connectivity-informed personalized targeting for rTMS in schizophrenia.