Machine learning-based prediction of antipsychotic efficacy from brain gray matter structure in drug-naive first-episode schizophrenia

Predicting patient response to antipsychotic medication is a major challenge in schizophrenia treatment. This study investigates the predictive role of gray matter (GM) in short- and long-term treatment outcomes in drug-naive patients with first-episode schizophrenia (FES). A cohort of 104 drug-naive FES was recruited. Before initiating treatment, T1-weighted anatomical images were captured. The Positive and Negative Syndrome Scale and the Personal and Social Performance Scale were adopted to assess clinical symptoms and social function. At the 3-month follow-up, patients were categorized into remission and non-remission groups. At 1-year follow-up, patients were categorized into the rehabilitation and non-rehabilitation groups. Machine learning algorithms were applied to predict treatment outcomes based on GM volume, cortical thickness, and gyrification index, and the model performance was evaluated. Widespread regions, such as the superior temporal gyrus, middle frontal gyrus, supramarginal gyrus, the posterior central gyrus, anterior cingulate gyrus, and parahippocampal gyrus showed substantial predictive value for 3-month treatment efficacy (74.32% accuracy). The inferior frontal gyrus, anterior cingulate gyrus, and inferior occipital gyrus demonstrated significant predictive power for treatment outcome at 1-year follow-up (70.31% accuracy). We developed a machine learning model to predict individual responses to antipsychotic treatments, which could positively impact clinical treatment protocols for schizophrenia.