Applying artificial intelligence to identify schizophrenia

Posted Jul 21, 2017 by Tim Sandle
Identifying the neuroimaging-based patterns which may offer clues for schizophrenia is complex. This task could be made easier thanks to developments with artificial intelligence.
A neuropsychologist points to a brain scan showing the brain activity of a paedophile at the Hudding...
A neuropsychologist points to a brain scan showing the brain activity of a paedophile at the Huddinge hospital near Stockholm
Jonathan Nackstrand, AFP
The mental health disorder schizophrenia has been associated with disrupted brain connectivity. One way of identifying this is through study of specific neuroimaging-based patterns, which are medically described as “pathognomonic” when they indicate schizophrenia. Moreover, such patterns can offer cues about the degree of severity of the symptoms. The task, however, is heavily data reliant and requires complex analysis. The complexity is down to the analysis not being simply due to looking for patterns of significance. Reviews also need to process data drawn from multiple datasets, contexts and cohorts.
Schizophrenia is a mental disorder characterized by abnormal social behavior and failure to understand what is real. A combination of genetic and environmental factors are considered to play a role in the development of schizophrenia. Symptoms of schizophrenia are divided into ‘positive’ and ‘negative’. Positive symptoms include experiencing things that are not real (hallucinations) and having unusual beliefs (delusions); whereas negative symptoms include may be a lack of motivation and being withdrawn. They often last longer than positive symptoms.
To help improve the accuracy of predicting schizophrenia research from the Department of Computing Science at the University of Alberta in Edmonton, Canada has developed a predictive method based on the analysis of biological markers. To construct the predictive model functional network feature patterns were collected from functional magnetic resonance images of the brain.
For the research the scientists used IBM artificial intelligence to analyze images from 95 subjects. The participants were divided into two groups. The first group was the control group and the second group, the test group, was composed of people diagnosed with schizophrenia. By measuring the ‘connectivity’ of each participant’s’ brain, the artificial intelligence accurately diagnosed patients 74 percent of the time.
From this the researchers concluded that this multi-step methodology can help to identify more reliable multivariate patterns going forwards, which will allow for accurate prediction of schizophrenia and the symptoms severity.
The research has been published in the journal Nature. The paper is titled “Learning stable and predictive network-based patterns of schizophrenia and its clinical symptoms.”
For those interested in research into schizophrenia, a companion article looks at recent research to test whether there is a biological explanation for the condition. See: “New tests underline biological connection for schizophrenia.”