AI Can Help Doctors Diagnose Tricky Mood Disorders Like Bipolar 1

Jack Crosbie
AUGUST 22, 2018
ai mental health,artificial intelligence mental health,augmented intelligence mdd,hca news

For all the advancements of modern medicine, doctors still struggle to decipher the human brain. But a new machine-learning algorithm could drastically simplify the complicated calculus that psychiatrists go through to diagnose conditions such as bipolar disorder and major depressive disorder (MDD), giving them a far easier way to fine-tune medication and treatment.

Researchers at the Lawson Health Research Institute, the Mind Research Network and the Brainnetome Center recently designed an artificial intelligence (AI) system purpose-built to analyze brain scans and pick apart the subtle differences in mood disorders. The algorithm — described in the journal Acta Psychiatrica Scandinavica — could help doctors prescribe and predict a patient’s response to a given medication, lessening the period of tortuous trial-and-error before a patient can get the help they need.

>> READ: mHealth Might Solve Our Prisons’ Mental Health Problem

The researchers used brain scans from 78 young adult patients at the London Health Sciences Center in Ontario, Canada. Of those, 66 patients had already been diagnosed and treated for either major depressive disorder or bipolar type 1. The researchers also used 33 participants with no history of mental illness as a control group for their data set, which was compiled using fMRI scans. Under an fMRI, brains exhibiting MDD and bipolar 1 look different next to each other and compared to the group with no history of mental illness, which the researchers designed the algorithm to identify. They set the AI loose on the data and checked its work based on the patients with an existing diagnosis. All told, the AI was promising — classifying illnesses correctly 92.4 percent of the time.

“The technique can always be improved, and we have lots of ideas there,” Vince Calhoun, Ph.D., president of the Mind Research Network and a professor of neuroscience, computer science and psychiatry at the University of New Mexico, told Healthcare Analytics News™ in an email. “But given the performance was very good for the current algorithm, I think the next steps include scaling up.”

Researchers want to scale the program up to see how it runs on different kinds of MRI machines, and to further automate how the AI processes data and comes to a conclusion (the “analysis pipeline,” as Calhoun put it). They also want to develop a protocol that can help human doctors to perform similar analysis to the algorithm, looking for the same distinctions that the computer does.

Down the line, the technique still needs much more data from additional studies, as well as some way of certifying imaging facilities to do the scans correctly. Calhoun said a widespread implementation would also need cloud-based tools to process the data.

For now, the program is still in its infancy.

“This is not ready for ‘prime time’ or routine clinical care yet,” Elizabeth Osuch, M.D., a clinician-scientist at Lawson Health Research Institute and medical director at the First Episode Mood and Anxiety Program (FEMAP) at the London Health Sciences Center, told HCA News in an email. “But advances in technology may lead to cheaper proxies for an fMRI scan.”

And in any case, Osuch said the AI is designed to be used in cases where traditional DSM diagnoses fails, not on every patient who comes through the door.

But, if implemented properly, it could make the difficult job of distinguishing a patient’s particular mood disorder much easier. The researchers said they already have algorithms that can differentiate bipolar disorders from schizophrenia and classify forms of dementia, and they are working toward applying the approach to everything from autism to ADHD — any disorder that the standard DSM diagnostic approach struggles to nail down.

The study, Osuch said in a news release, “suggests that we may one day have an objective measure of psychiatric illness through brain imaging that would make diagnosis faster, more effective and more consistent across health care providers."

Get the best insights in healthcare analytics directly to your inbox.

Related
Mental Health AI Start-Up Woebot Labs Earns $8M
Digital Therapy Reduces Social Anxiety, Study Finds
The Long-Term Effects of Participating in Reddit’s Mental Health Communities

SHARE THIS SHARE THIS
93
Become a contributor