Researchers have developed a machine-learning-based classifier that could assist in the early diagnosis of psychosis, according to a study published in Molecular Psychiatry. The tool utilizes MRI brain scans to classify individuals into those who are healthy and those at risk of experiencing a psychotic episode. The international consortium of researchers, including scientists from the University of Tokyo, used the classifier to analyze scans from over 2,000 people across 21 global locations. Approximately half of the participants had been identified as clinically at high risk of developing psychosis.
By using training data, the classifier was able to differentiate between individuals who were not at risk and those who later experienced overt psychotic symptoms with an accuracy rate of 85%. When new data was introduced, the accuracy rate decreased slightly to 73%. These findings suggest that the machine-learning tool could be valuable in future clinical settings, where early intervention can lead to better outcomes and minimize the negative impact on individuals’ lives.
Psychosis is a condition characterized by delusions, hallucinations, or disorganized thinking, and its onset can be triggered by various factors such as illness, injury, trauma, substance use, medication, or genetic predisposition. Although treatable, identifying individuals in need of help, especially young people during adolescence or early adulthood when the brain and body undergo significant changes, can be challenging.
Associate Professor Shinsuke Koike from the University of Tokyo’s Graduate School of Arts and Sciences explained that only about 30% of clinically high-risk individuals eventually develop overt psychotic symptoms, leaving the remaining 70% without a clear indication of their risk. Therefore, clinicians require assistance in identifying those individuals who will go on to experience psychotic symptoms using not only changes in thinking, behavior, and emotions but also biological markers.
Previous studies using brain MRI scans have indicated structural differences in the brains of individuals after the onset of psychosis. However, this study is reportedly the first to identify differences in the brains of individuals at high risk of psychosis who have not yet experienced symptoms. The research team from 21 different institutions in 15 countries gathered a diverse group of adolescent and young adult participants for their study.
MRI research related to psychotic disorders is challenging due to variations in brain development and MRI machines, making it difficult to obtain accurate and comparable results. Moreover, distinguishing between changes in brain development and those caused by mental illness in young people can be particularly complex.
According to Koike, the researchers were able to overcome these challenges and create a tuned classifier by correcting for differences in MRI models and parameters. The participants were divided into groups based on clinical risk, and the researchers used a machine-learning algorithm to identify patterns in their brain anatomy. The algorithm successfully classified the participants into healthy controls and those at high risk of developing psychosis, with an accuracy rate of 85% during training and 73% during the final test.
The researchers propose that providing brain MRI scans to individuals identified as being at high clinical risk could be helpful in predicting future psychosis onset. However, they acknowledge the need to test the classifier with new sets of data and build a classifier that can robustly classify MRI scans from different sites and machines. A national brain science project in Japan called Brain/MINDS Beyond is currently taking on this challenge.
If successful, the creation of more robust classifiers for new data sets could enable the application of this tool in real-life clinical settings, leading to improved outcomes for individuals at risk of psychosis. Early intervention and diagnosis play a crucial role in minimizing the impact of psychosis and promoting recovery.
