A 2020 study has shown that neurological disorders can be detected using a combination of machine learning and magnetic resonance imaging (MRI) technology. The research was published in Translational Psychiatry in August 2020.
A team of University of Tokyo researchers, composed of both machine learning experts and medical professionals, engineered a computer algorithm and trained it using the brain scans of 206 Japanese volunteers. The majority of the participants had some form of mental illness (either schizophrenia, autism spectrum disorder, or psychosis), but also included in the study were a number of volunteers without any psychiatric conditions.
The researchers used six different algorithms to differentiate between MRI scans. After feeding their main algorithm the volunteers’ brain images, it learned to distinguish between psychiatric disorders by measuring variations in thickness, surface area, and volume of brain areas taken from the scans.
Once the algorithm had been properly configured, it was put to the test. The scientists fed it 43 additional MRI scans to see if it could detect the presence of a mental disorder. The patients who contributed their MRIs had been previously assessed by psychiatrists.
After running the algorithm, the researchers found that it was able to detect mental illness with 85 percent accuracy–nearly as reliable as the human psychiatrists. Not only was the algorithm able to determine if a patient had autism or schizophrenia, it could also diagnose which one, and if a patient had any risk factors for the latter.
A first of its kind
While similar studies have been conducted in the past, according to the research team, their study is the first to use artificial intelligence to make a distinction between various psychiatric disorders. Shinsuke Koike, M.D., associate professor at the University of Tokyo and senior author of the study, noted the importance of analyzing both autism and schizophrenia in such a trial, due to a potential link between the two diseases.
“Autism spectrum disorder patients have a 10 times higher risk of schizophrenia than the general population. Social support is needed for autism, but generally the psychosis of schizophrenia requires medication, so distinguishing between the two conditions or knowing when they co-occur is very important,” said Koike.
Potential for future use in psychiatry?
Since the team’s algorithm identified mental illness based on certain brain measurements (particularly the thickness of the cerebral cortex), their findings could potentially be used to better understand causes of psychiatric illnesses or possibly to help diagnose schizophrenia by monitoring physical changes in the brain’s cortex area. The study’s results could even pave the way for alternative mental health treatments. For now, though, the researchers plan to further optimize their algorithm and test it on larger datasets in future trials.
Written by Natan Rosenfeld