Predicting subclinical psychotic-like experiences on a continuum using machine learning

Abstract

= 174 words Full Text = 3962 words Abstract Previous studies of psychosis using machine learning methods have primarily been concerned with binary classification of patients and healthy controls. The aim of this study was to use electroencephalographic (EEG) data and pattern recognition to predict individual psychotic experiences on a continuum between these two extremes in otherwise healthy people. From responses evoked by an auditory oddball paradigm, behavioural measures, neural activity, and effective connectivity were extracted as potential contributing features. Optimal performance was achieved using spatiotemporal maps of neural activity in response to frequent sounds, with late-P50 and early-N100 time windows contributing most to higher schizotypy scores. Effective connectivity estimates, in particular top-down frontotemporal connections, were also predictive of psychotic symptoms.As a proof-of-concept, these findings demonstrate that individual psychotic experiences in healthy people can be predicted from EEG data alone, whilst also supporting the idea of altered sensory responses and the dysconnection hypothesis in schizophrenia, as well as the notion that psychosis may exist on a continuum.