Researchers Develop Machine-Learning Tool to Predict Psychosis Onset
A groundbreaking machine-learning tool has been developed by researchers, allowing accurate identification of individuals at high risk of psychosis through MRI brain scans. The innovative tool achieved an impressive 85% accuracy rate during training and maintained a 73% accuracy rate when using new data. This development paves the way for early interventions in psychosis and has the potential to greatly improve treatment outcomes.
An Advancement in Psychiatric Care
This significant advancement in psychiatric care involves the detection of structural brain differences before the onset of psychosis. By using MRI scans, the machine-learning tool can accurately distinguish between individuals at high risk of psychosis and those not at risk. This breakthrough offers potential benefits for diverse clinical settings and emphasizes the importance of better prediction and prevention strategies in mental health.
Key Facts
- The machine-learning tool can predict psychosis risk using MRI brain scans with a high level of accuracy.
- Early identification of psychosis risk through MRI scans can lead to more effective and timely interventions, minimizing the negative impact on individuals.
- Further development is needed to ensure the tool’s applicability in different data sets and clinical environments.
The machine-learning tool used in the research can accurately classify MRI brain scans, differentiating individuals at high risk of psychosis from those who are healthy. The tool was trained using data from over 2,000 participants from 21 locations worldwide. Half of the participants in the study had previously been identified as being at high risk of developing psychosis.
During the training phase, the classifier achieved an 85% accuracy rate in accurately distinguishing between individuals at risk and those not at risk. This accuracy rate remained at an impressive 73% when using new data. The tool’s use in future clinical settings holds significant promise since early intervention in psychosis leads to improved outcomes and reduced negative impacts on individuals’ lives.
Psychosis is a condition that may involve delusions, hallucinations, or disorganized thinking and can affect anyone. Potential triggers for psychosis include illness or injury, trauma, substance use, medication, or a genetic predisposition. Effective treatment is available, and most individuals make a full recovery. However, identifying young people who require support can be challenging, as the common age for a first episode of psychosis is during adolescence or early adulthood.
Associate Professor Shinsuke Koike from the Graduate School of Arts and Sciences at the University of Tokyo explained, “Clinicians require assistance in identifying individuals who will go on to have psychotic symptoms, using both subclinical signs and biological markers. Our machine-learning tool, using MRI brain scans, aims to address this need by predicting psychosis onset at the high-risk stage.”
While previous studies have indicated structural brain differences after the onset of psychosis, this research is the first to identify differences in the brains of individuals at high risk of psychosis who have not yet experienced symptoms. The study involved collaboration between 21 institutions from 15 countries, ensuring a large and diverse pool of adolescent and young adult participants.
Accurately researching psychotic disorders using MRI scans can be challenging due to variations in brain development and MRI machine specifications, making it difficult to obtain precise and comparable results. However, the research team successfully developed a machine-learning classifier that can overcome these challenges by accounting for differences in MRI models and developing a robust classifier for predicting psychosis onset.
By dividing the participants into various high-risk groups and using machine-learning algorithms, the researchers were able to classify individuals into healthy controls and those at high risk of developing psychosis. The results were highly accurate, with the classifier achieving an 85% accuracy rate during training and maintaining a 73% accuracy rate using new data.
Further research and testing are required to determine the classifier’s effectiveness across different data sets and clinical environments. The international research consortium involved in this study is committed to refining the classifier to ensure its widespread applicability in routine clinical settings and real-life scenarios.
Funding: The research received support from various sources, including AMED, JST Moonshot R&D, JSPS KAKENHI, and the Takeda Science Foundation. The study was also supported by the International Research Center for Neurointelligence at the University of Tokyo.
About this Psychosis Research News
Joseph Krisher
Source: University of Tokyo
Original Research
Molecular Psychiatry
