Artificial intelligence (AI) has made significant strides in the field of mental health, with researchers developing a machine learning classifier that can predict the risk of psychosis before it occurs. This groundbreaking tool utilizes MRI scans to distinguish between individuals who are at high risk for psychotic episodes and those who are not.
The classifier was trained using MRI scans from over 2,000 people across 21 international locations, approximately half of whom were identified as clinically high risk for developing psychosis. Remarkably, the classifier demonstrated an accuracy rate of 85 percent in distinguishing individuals who were not at risk from those who later experienced psychotic symptoms. Even when tested with new data, the accuracy remained at an impressive 73 percent.
The implications of this advancement are profound, particularly in clinical settings where early intervention in psychosis is vital for better recovery outcomes. Psychosis, characterized by delusions, hallucinations, and disorganized thinking, can have various triggers, including illness, trauma, substance use, or genetic predisposition. However, it is treatable, with most individuals eventually making a full recovery.
Identifying individuals who require help can be challenging, particularly since the onset of psychosis often occurs during the teenage years or young adulthood, a period marked by significant changes in the brain and body. Professor Shinsuke Koike from the University of Tokyo emphasizes the need to not only rely on subclinical signs but also biological markers to identify those who will develop psychotic symptoms.
This study, conducted by an international team from 21 institutions across 15 countries, is among the first to identify brain differences in individuals at high risk of psychosis who have not yet experienced an episode. The research aims to overcome challenges related to brain development variations and MRI machine calibrations by correcting for these differences and creating a fine-tuned classifier for predicting the onset of psychosis.
The study involved 1,165 participants, categorized into three groups: those at clinically high risk who later developed psychosis, those who did not develop psychosis, and those with uncertain follow-up status. Additionally, there were 1,029 healthy controls for comparison. The machine-learning algorithm was trained to recognize patterns in brain anatomy among these groups, resulting in an impressive accuracy rate of 85 percent during training and 73 percent when tested with new data.
However, further testing is required to ensure the classifier works effectively with new sets of data. The researchers are focusing on building a classifier that can robustly classify MRIs from new sites and machines. A national brain science project in Japan called Brain/MINDS Beyond is taking on this challenge. If successful, the team can create more reliable classifiers for new data sets, which can then be implemented in real-life and routine clinical settings.
The findings of this study, published in the journal Molecular Psychiatry, mark a significant step forward in the prediction and early intervention of psychosis. With the aid of AI and MRI scans, clinicians may soon have a powerful tool at their disposal for identifying individuals at high risk of psychosis before it manifests, allowing for timely intervention and improved outcomes. This breakthrough has the potential to revolutionize mental healthcare and provide hope for those at risk of developing psychosis.
