Study: Machine learning could lead to objective measures for schizophrenia diagnosis

Images of regions of the brain that showed a statistically significant difference between patients with schizophrenia and patients without it. (IBM)

Diagnosing schizophrenia relies on subjective methods, but IBM and the University of Alberta are working to change that. The duo published research suggesting machine learning could help scientists uncover objective measures in brain imaging to predict when someone has schizophrenia.

There is no single test for schizophrenia, so diagnosis typically involves observing symptoms and ruling out other potential causes for them, such as substance abuse or another mental health disorder. And while scientists have observed differences in brain scans of healthy people and those with schizophrenia, but these are not currently used to diagnose the neurological disorder, according to the National Institute of Mental Health (NIMH).

Algorithms predicted schizophrenia occurrence by analyzing brain networks

In the study, the researchers used machine learning algorithms to analyze functional magnetic resonance imaging (fMRI), which measures changes in blood flow to gauge activity in different parts of the brain. Using an open data set, Function Biomedical Informatics Research Network, they looked at fMRI data taken from 95 people while they did an auditory test, IBM said in a statement. These included healthy people, as well as patients with schizophrenia and those with schizoaffective disorder, a condition where patients experience symptoms of schizophrenia and a mood disorder.

The team used machine learning to create a model that identifies schizophrenia based on connections in the brain, IBM said. The fMRI data was taken from different sites, using different machines, but the algorithm could differentiate between the patients with schizophrenia and without 74% of the time.

And the algorithm doesn't just determine if a patient has schizophrenia or not, said Guillermo Cecchi, who directs the Computational Psychiatry and Neuroimaging groups at IBM. More importantly, it can also predict the severity of disease to provide a better representation of the individual's condition. This can guide physicians as they develop treatment plans for schizophrenia patients.

“We’ve discovered a number of significant abnormal connections in the brain that can be explored in future studies, and AI-created models bring us one step closer to finding objective neuroimaging-based patterns that are diagnostic and prognostic markers of schizophrenia, ”said Serdar Dursun, a professor of psychiatry and neuroscience at the University of Alberta, in the statement.

The researchers hope the findings, published in the journal Schizophrenia, will build our understanding of the disorder and improve treatments.

“The ultimate goal of this research effort is to identify and develop objective, data-driven measures for characterizing mental states, and apply them to psychiatric and neurological disorders” said Ajay Royyuru, Vice President of Healthcare & Life Sciences, IBM Research. “We also hope to offer new insights into how AI and machine learning can be used to analyze psychiatric and neurological disorders to aid psychiatrists in their assessment and treatment of patients.”

The researchers will continue to look at areas and connections in the brain implicated in schizophrenia. They plan to conduct studies in bigger cohorts, Cecchi said, and to investigate the universality of the measures. The goal is to find out if the schizophrenia affects the brain "no matter what you do," or if its effects are limited only to certain tasks, such as the auditory test.

Objective measures for mental health

The emerging field of computational psychiatry aims to develop new approaches to improve evidence-based decision-making while treating mental health disorders.

Mindstrong Health, cofounded by former NIMH chief Tom Insel, is working on technology that analyzes smartphone data to determine a person’s mental state. The company’s tech collects information on which words are used, or a person’s location when using certain apps, for example, and turns them into objective measures of brain function. The company recently raised $14 million in its Series A.

Meanwhile, Boston-based Akili Interactive and Pfizer reported data last year showing that a video game-based diagnostic test could distinguish between people with and without brain amyloidosis, a hallmark of Alzheimer’s disease. And PureTech's Sonde Health is working on the analysis of "vocal biomarkers," or changes in nonlinguistic characteristics of a person's voice, to indicate changes in health.