Artificial intelligence can flag PTSD markers, study shows
Computer program able to identify those with ailment using MRI images
This is your brain on PTSD.
Researchers in London are opening an electronic window into a mysterious mental illness that affects many Canadians, one that could lead to both better diagnosis of post-traumatic stress disorder and improved treatment. Using brain scans, the researchers — backed by some of the city’s key health organizations — are developing a way to use artificial intelligence to tell whether a patient has the debilitating psychological condition associated with exposure to traumatic events and experiences.
Their computer, which uses statistical models to progressively improve its own performance, was able to tell the difference between patients with PTSD and patients with no PTSD diagnosis with 92 per cent accuracy.
“What this study really showed is that training machine-learning algorithms — artificial intelligence — with brain images, we were able to classify PTSD from health controls, but also to further differentiate PTSD from dissociative subtype PTSD,” said lead study author Andrew Nicholson, a Lawson Health Research Institute scientist. Researchers at Western and Lawson, the research arm of the London Health Sciences Centre and St. Joseph’s Health Care London, compiled resting brain scans of 181 participants through magnetic resonance imaging (MRI). Some had no PTSD diagnosis, others had the disorder.
In about 70 per cent of PTSD cases, the illness makes patients feel hyperactive, feel emotions intensely, be prone to outbursts, have a quick pulse and experience fight-or-flight response. In the other 30 per cent of patients, PTSD makes them feel dissociated and detached, as if they ’re shut down or having an out-of-body experience. The computer program was able to effectively tell the difference between brains with each type of PTSD and psychologically healthy brains.
The promising result? With more research, the London findings could create a new diagnostic tool for the disorder.
“The direction that we want to go in is to take this technology and translate it to more clinically feasible modalities,” said Nicholson, who is also a post-doctoral fellow at Western University’s medical school.
But MRI technology is costly and not always easy to access, he said. Adapting the findings of their MRI brain scan study so it could be completed in clinics using another form of medical imaging, electroencephalogram (EEG) technology, could give doctors a new way to assess patients, Nicholson said. EEG technology uses sensors placed on the scalp to track electrical impulses
in the brain. Diagnosing PTSD involves a clinical assessment, observations and determinations by a trained clinician. There aren’t any purely scientific tests for the illness — yet. The researchers’ findings show there are identifiable bio-markers in the brain that can help doctors and researchers to physically determine if a patient has PTSD. With further study, the findings could help clinicians tailor treatment to patients struggling with both subtypes of the disorder. “What’s critical right now is that our treatments are kind of trial and error and we don’t have a protocol that really matches the right treatment to a specific person,” Nicholson said.
“In the future, we can actually use this to predict the symptom trajectory … but also how likely they are to going to respond to a certain treatment.”
The researchers looked at full brain activity in the study but plan to zero in on specific networks known to be affected by PTSD in future research projects. The brain scan data research could also identify previously unreported variations of the disorder, Nicholson said.
“We really believe there are a lot more sub-types and different diagnoses of PTSD that we have yet to discover,” he said. “Machine learning will allow us to data-mine, in a sense, to objectively differentiate patients.”
The London study is published in the journal Psychological Medicine.