U of A research shows faster, more accurate schizophrenia diagnosis possible with artificial intelligence
Machine learning can predict instances of schizophrenia with 74 per cent accuracy
A partnership between University of Alberta researchers at the Alberta Machine Intelligence Institute (Amii) and IBM scientists has broken new ground in the use of artificial intelligence to predict cases of schizophrenia.
In a recently published study, the researchers demonstrate that machine learning algorithms can predict instances of schizophrenia with 74 per cent accuracy, a number comparable to what human doctors are currently doing.
The researchers were also able to predict the severity of specific symptoms in schizophrenia patients — something that wasn't possible before.
By using artificial intelligence and machine learning, "computational psychiatry" can be used to help clinicians more quickly assess and treat patients with schizophrenia.
"This was a very fruitful collaboration between the U of A and IBM, " said postdoctoral fellow Mina Gheiratmand, who co-authored the research.
"What we did basically was to use machine learning techniques for predicting the instances of schizophrenia — so who basically has schizophrenia — and this was a little challenging."
Currently, doctors are limited to diagnosing patients while relying on behavioural symptoms, and there is no medical testing that can provide an absolute diagnosis. That can mean significant delays before a symptomatic person is successfully diagnosed.
Computational psychiatry provides doctors with tools that enable them to objectively assess patients where most approaches had been subjective up until that point, according to the research.
"Currently, it's not known what is wrong, for example, what is abnormal in the brains of patients suffering from schizophrenia," said Gheiratmand. "We're using brain images to basically see whether we can discriminate between patients and controls based on brain images.
'Will help Alberta, world in general'
The images are captured using functional magnetic resonance imaging (fMRI), which represents the neural activity of the brain in different regions. It's sort of like capturing moving snapshots that are used to compare with a healthy control group that isn't suffering from the illness.
None of this would be possible without provincial funding for the research, according to Russell Greiner, who helps run the Alberta Machine Intelligence Institute.
Since 2002, the Alberta government has provided funding of up to $2 million a year for machine learning, he said.
"The province started funding us in 2002, before machine learning was so popular," Greiner said. "They had the insight to realize this is an important technology that will help Alberta, and help the world in general."