By all accounts, Web Developers are among those most concerned about being replaced by AI. Those concerns, however, may be misplaced.
Even though mental illness begins in the brain, diagnosing it through brain imaging has typically proved elusive for clinicians.
There are a lot of reasons why.
Different psychiatric conditions like bipolar disorder or anxiety share similar symptoms, like irritability. Even conditions like depression can have a myriad of symptoms that can change depending on the person, notes Seth Gillihan, a Clinical Assistant Professor of Psychology at the University of Pennsylvania, in Psychology Today.
And while Researchers are getting closer to pinpointing certain areas of the brain impacted by different illnesses — such as the amygdala, an area that processes fear, in patients with post-traumatic stress disorder — there’s also a lot of overlap and a lot of questions.
But in the years ahead, experts say some of those questions could be answered through artificial intelligence.
“Unfortunately, no one knows how different parts of the brain interact,” says Russ Greiner, Professor in the Department of Computing Science at the University of Alberta. “What we do have are data sets.”
Data that could one day contain information on hundreds or thousands of previous patients, and could be capable of building classification systems to apply to new patients in need of a diagnosis, he says.
While Greiner says the technology isn’t still in the works, his research is already pointing at the widespread clinical possibilities of AI in the diagnosis of mental illness.
Interpreting Brain Scans Through AI
Along with a team of UAlberta Researchers, Greiner conducted a 2019 study using machine learning to identify schizophrenia in brain scans.
The team developed a form of AI which trained on — and learned from — numerous brain scans of previous patients diagnosed with the complex mental disorder, which is often marked by abnormal behavior, hearing voices, and an overall inability to understand what’s real and what’s not.
The technology — dubbed EMPaSchiz, or the Ensemble algorithm with Multiple Parcellations for Schizophrenia prediction — looks at thousands of features in a patients’ brain scan to make a prediction.
The team’s research showed EMPaSchiz could identify schizophrenia accurately 87 percent of the time for new scans.
“One issue with psychiatric diagnosis is that it is subjective,” says Greiner, who’s also the Principal Investigator of the Alberta Machine Intelligence Unit. “One thing I like to do is have measures not based on what a psychiatrist does, but on the actual patient characteristics.”
While widespread practical application is a long way off, he says the goal is developing targeted treatments for each patient, instead of cycling people through different options that may or may not work — a tiring and potentially-dangerous process that can leave patients in limbo for months, or even years.
Researchers hope “to make patients suffer less, and for less time,” he says.
Apps and Wearables Entering Mental Health Sphere
Beyond diagnosing brain scans, Greiner also predicts AI will enter the mental health sphere in other ways in the years ahead.
From Alexa to Siri, various apps and their parent companies are already starting to dabble in the healthcare space, for instance. “Given the Fitbits and smartphones, I think there’s more and more of these other options for psychiatric diseases,” Greiner adds.
It raises big questions: What if our always-on technology could catch issues with our mental state, by sensing our words and actions? What if they checked in the way a helpful colleague or family member would do?
“You could have tools which monitor you and give advice,” Greiner says. “Think of what your mother would do: ‘Fred, I think you seem out of sorts.” People’s mothers may not be doing it, but their cell phones might.”
BioBeats, for instance, is one UK-based app founded in 2013, which identify stress patterns using sensors in smartphones and wearables, Forbes reports.
By coupling the sensors with behind-the-scenes machine learning, the technology allows users to understand how their body is responding to stress and potentially take preventative action for the sake of their mental health.
Greiner says AI could potentially even help prevent addiction relapses or suicide, by flagging unusual behavior or dangerous locations to a person or their loved ones — be it an alert reminding someone they’re in a neighbourhood where they typically buy drugs, or one offering a crisis helpline if an app determined their search history pointed to suicidal thoughts.
“Even if these people withdraw from human contact, they still have their cell phones,” Greiner says.
“A friend could be aware of these things. Why couldn’t the artificial intelligence tool that’s with you?”