Detection of schizophrenia: A machine learning algorithm for potential early detection and prevention based on event-related potentials

Abstract

We show that event-related potentials can be used to detect schizophrenia with a high degree of precision. With our machine learning algorithm we achieve a balanced accuracy of 96.4 , which exceeds all results with comparable approaches. For this we use additional sensors on the left and right hemisphere in addition to the common central sensors. The experimental design when recording the data takes into account the dysfunction of the schizophrenic efference copy. Due to its serious consequences, schizophrenia is a social issue in which early detection and prevention plays a central role. In the future, machine learning could be used to support early interventions. When the first symptoms appear, potential patients could be tested for the dysfunction typical for schizophrenia. In this way, risk groups and potential patients could be adequately treated before the onset of psychosis.