MACHINE LEARNING FOR MENTAL HEALTH: FOCUS ON AFFECTIVE AND NONAFFECTIVE PSYCHOSIS

Machine learning is developing at a fast pace, providing the ability to analyze large volumes of data quickly and delivering insight without a-priori assumptions. Thus, this technique represents a promising tool in the field of mental healthcare, given the challenges faced by this field including the lack of biological markers of disease and prognosis, the large heterogeneity in clinical presentation, and the lack of reliable predictors of treatment response. Machine learning has been widely applied, especially in neuroimaging studies, but still largely confined in an early phase of deployment and with a varying degree of efficacy. This chapter discusses the most recent applications of this technology in severe psychotic illnesses (schizophrenia spectrum and bipolar disorders), focusing on application of machine learning in early detection, differential diagnosis, treatment management, and response prediction. Pilot attempts to use machine learning for mental health service design and implementation are also discussed. For each key point, a review of updated literature is provided.