Prediction of Psychosis before its onset

An international consortium, including researchers from the University of Tokyo, developed a machine-learning tool capable of predicting the onset of psychosis before it happens. This tool classifies MRI brain scans into two categories: those from healthy individuals and those at risk of a psychotic episode.

Psychotic episodes, characterized by delusions, hallucinations, or disorganized thinking, can happen to anyone due to triggers like illness, trauma, drug use, medication, or genetics. Though potentially distressing, they are treatable, and most people recover but identifying those in need of help can be difficult, especially since these episodes most commonly begin in adolescence or early adulthood, a time of significant changes.

“At most only 30% of clinical high-risk individuals later have overt psychotic symptoms, while the remaining 70% do not,” explained Associate Professor Shinsuke Koike from the Graduate School of Arts and Sciences at the University of Tokyo. “Therefore, clinicians need help to identify those who will go on to have psychotic symptoms using not only subclinical signs, such as changes in thinking, behavior and emotions, but also some biological markers.”

The team from 21 different institutions in 15 different countries gathered a large and diverse group of adolescent and young adult participants. According to Koike, MRI research into psychotic disorders can be challenging because variations in brain development and in MRI machines make it difficult to get very accurate, comparable results. Also, with young people, it can be difficult to differentiate between changes that are taking place because of typical development and those due to mental illness.

For the research, participants were divided into three groups of people at clinical high risk: those who later developed psychosis; those who didn’t develop psychosis; and people with uncertain follow-up status, and a fourth group of healthy controls for comparison. Using the scans, the researchers trained a machine-learning algorithm to identify patterns in the brain anatomy of the participants. From these four groups, the researchers used the algorithm to classify participants into two main groups of interest: healthy controls and those at high risk who later developed overt psychotic symptoms.

In training, the tool was 85% accurate at classifying the results, while in the final test using new data it was 73% accurate at predicting which participants were at high risk of psychosis onset.

“We still have to test whether the classifier will work well for new sets of data. Since some of the software we used is best for a fixed data set, we need to build a classifier that can robustly classify MRIs from new sites and machines, a challenge which a national brain science project in Japan, called Brain/MINDS Beyond, is now taking on,” said Koike. “If we can do this successfully, we can create more robust classifiers for new data sets, which can then be applied to real-life and routine clinical settings.”

Reference: Using Brain Structural Neuroimaging Measures to Predict Psychosis Onset for Individuals at Clinical High-Risk. Molecular Psychiatry. DOI: 10.1038/s41380-024-02426-7