Cerebellar-Cortical Circuitry and Treatment Response in Early Psychosis - Abstract Identifying circuitry-based biomarkers of treatment response to antipsychotic medication is critically important for precision psychiatry, which may help physicians make optimal treatment plans at the early stage, reducing the course of ineffective treatment and benefiting long-term life quality of patients. Using functional magnetic resonance imaging (fMRI) and two clinical samples of first-episode psychotic patients (n = 49 and 24, respectively), our recent work has identified that connectivity in the cerebellar-cortical circuitry is a state- independent predictor of antipsychotic response that shows the best predictive value across the whole brain, regardless of fMRI paradigm. This finding strongly points to a promising prognostic biomarker with potential to be clinically translated for assessment of treatment efficacy. The goal of the proposed study is therefore to collect new patient data to validate this identified predictive biomarker in a larger sample (120 patients with first-episode psychosis, Aim 1). Given the state-independent nature of the identified predictive biomarker, we will use naturalistic stimuli instead of traditional high-demand cognitive tasks as fMRI paradigms, which are more patient-friendly and thus may facilitate the application of this biomarker in a real clinical scenario. Using this new dataset, we will further refine functional predictors in the cerebellar-cortical circuitry based on connectivity of each cerebellar system (Aim 2), and develop white matter structural predictors related to the superior cerebellar peduncle (the major efferent fiber of the cerebellum sending information to the cerebral cortex) and middle cerebellar peduncle (the major afferent fiber of the cerebellum receiving information from the cerebral cortex, Aim 3). The results derived from multimodal imaging data will further be used to train an updated prediction model based on both regularized regression and deep learning, with the model performance to be externally validated in our in-house dataset (Aim 4). This will eventually establish an optimized, validated, and generalizable cerebellum-based predictive model for future clinical use.
