PRISM: Profiling the Risk for Schizophrenia with Machine Learning

Under the title "PRISM: Profiling the Risk for Schizophrenia with Machine Learning," Christian Valt, PhD, a leading expert in the field of cognitive neuroscience will present his latest research findings on risk assessment for schizophrenia using machine learning.

Identifying individuals at risk for schizophrenia (SCZ) remains a challenge for personalized psychiatry. The heterogeneity of the prodromal symptomatology of SCZ has hindered group-based approaches. However, neurophysiological biomarkers may offer clues for precisely identifying at-risk individuals. A key finding in SCZ and those at risk for psychosis is the blunted mismatch negativity (MMN), elicited by the pre-attentive detection of deviant auditory stimuli among a series of standard stimuli.

Dr. Valt and his team investigated the brain network underlying this MMN deficit through effective connectivity between auditory regions. Using Granger causality on two independent MMN datasets, they revealed that people with psychosis show an overall reduction of effective connectivity within the auditory network, with the most significant deficit in communication from the left to the right auditory cortex. A follow-up machine learning study demonstrated that MMN effective connectivity and other MMN parameters are good predictors for a diagnostic model of SCZ, with potential applications in risk detection.