Prediction of outcome in the psychosis prodrome using neuroanatomical pattern classification

Abstract

To date, research into the biomarker-aided early recognition of psychosis has focused on predicting the transition likelihood of clinically defined individuals with different at-risk mental states (ARMS) based on structural (and functional) brain changes. However, it is currently unknown whether neuroimaging patterns could be identified to facilitate the individualized prediction of symptomatic and functional recovery. Therefore, we investigated whether cortical surface alterations analyzed by means of multivariate pattern recognition methods could enable the single-subject identification of functional outcomes in twenty-seven ARMS individuals. Subjects were dichotomized into 'good' vs. 'poor' outcome groups on average 4years after the baseline MRI scan using a Global Assessment of Functioning (GAF) threshold of 70. Cortical surface-based pattern classification predicted good (N=14) vs. poor outcome status (N=13) at follow-up with an accuracy of 82% as determined by nested leave-one-cross-validation. Neuroanatomical prediction involved cortical area reductions in superior temporal, inferior frontal and inferior parietal areas and was not confounded by functional impairment at baseline, or antipsychotic medication and transition status over the follow-up period. The prediction model's decision scores were correlated with positive and general symptom scores in the ARMS group at follow-up, whereas negative symptoms were not linked to predicted poorer functional outcome. These findings suggest that poorer functional outcomes are associated with non-resolving attenuated psychosis and could be predicted at the single-subject level using multivariate neuroanatomical risk stratification methods. However, the generalizability and specificity of the suggested prediction model should be thoroughly investigated in future large-scale and cross-diagnostic MRI studies.

Keywords: At risk mental state; Functional outcome; Multivariate prediction; Neuroimaging biomarkers; Psychosis.