Scientists are harnessing the power of artificial intelligence (AI) for early detection of mental health disorders such as depression and schizophrenia.
Unlike, for example, kidney disease, which is relatively easy to diagnose, conditions such as anxiety or depression do not have specific biomarkers that can be detected with a simple test.
Patients with the same mental disorder can present with many different symptoms, which can make it difficult for clinicians to diagnose them early and accurately.
Scientists from Nanyang Technological University in Singapore are working to develop an AI-based diagnostic toolkit that could help solve the problem.
AI’s ability to efficiently process large data sets helps detect signs of different mental health conditions and determine if a patient’s mental impairment becomes more severe.
Dr Eva Bojic, a computer scientist and lead researcher on the project, told Euronews Next that mental health disorders “can be visible in very physical ways”, and that wearables help them pick up on those signals.
“We collected different variants, different signals from variants, something we call digital biomarkers,” Bojic said.
Track heart rate and sleep patterns
Some of the vital signs include heart rate, sleep patterns, energy expenditure, calories expended, and number of steps; “And then we linked those biomarkers to the symptoms we saw, and after a while, the model could learn,” she said.
For example, one of the signals that scientists picked up in their study is that the heart rate of people with depression usually speeds up during the night hours, specifically between the hours of 2 and 4 in the morning.
The biomarkers are then supplemented with a questionnaire for a final assessment and identification of users in the sample who actively suffer from depression.
Finally, users are categorized into two groups, “zero for people who are not depressed and one for, say, people who are depressed. And then we develop machine learning models,” Bojek explained.
The machine learning model is able to make predictions for new users based on their biomarkers, relate knowledge and labels learned from previously analyzed biomarkers, and define a new result: zero or one – healthy or unhealthy.
But of course, “there’s nothing bulletproof,” Bojic said.
She explains how the model predicts is not really binary. The program provides a percentage from zero to 100, “Then it’s basically about where we put the threshold. Would you say depression starts above 50, or are we going to say it starts around 80 percent?”
What is the margin of error in an AI model?
The margin of error depends on the level of detail in the data structure, how clean and accurate the data is, and the number of hours the user used the fitness tracker per day.
Bojic estimated the tool’s accuracy at around 80 percent. “Then sometimes, if you’re really, really strict about data quality, we can really get to 100 percent.”
“But there can be many limitations” – interestingly enough, the error often comes from the human side, she added.
Bojek explained that a questionnaire often poses a problem because depending on how the evaluator asks the question, they can get a different answer.
AI scan tool ‘not perfect’
It also happens that users are disloyal, due to mental health stigma, and sometimes, quite simply, are not aware of their condition.
But despite the challenges, “there is some significant correlation between physical symptoms and the knowledge we elicit. So there is definitely potential to explore,” she said.
“What we’re doing is more than just a screening tool for people…and then, hopefully, they can get into the care process, where their condition can be properly managed.”
“This tool is not perfect,” she added. “So it’s not about eliminating the need for professional psychological help, it’s about how useful it is to help people, and to help professionals identify people they wouldn’t otherwise have access to.”
Not to mention assisting professionals in helping patients as soon as possible. Research shows that early detection of conditions such as depression and schizophrenia is critical to being able to prevent disorders from escalating.
Bojic said one of the strengths of her team’s study, which was published in JMIR mHealth and uHealth, is that it looked at the general public, rather than focusing on people who have already been diagnosed clinically.
What will become of this data?
How to proceed after someone is identified with a possible risk of developing the disease is still an important detail that researchers are working on.
“From a clinical point of view, how can they actually approach someone? Is it a good way of directly saying, you know, ‘You are depressed’? How is that effective?” Bojic said.
“On the other hand, if the system has an error, and a non-depressed person is told that it is, how will that affect that person?”
Then there are the thorny questions about ethics and privacy.
Should governments, agencies or hospitals have access to this information? And how can they harness it to promote good mental health?
“There are many angles in terms of dealing with it, a lot of considerations and certainly ethical, privacy issues. All of that has to be taken into account before this information is actually used in different places,” Bojic said.
Once these concerns are addressed, she believes AI can show great promise in the field of mental health.
“I really believe we can help people with the knowledge and algorithms that we have developed,” she said.
“It’s not just chasing some numbers.”
For now, the team is focused on detecting depression, but they hope to expand it to other conditions such as dementia, loneliness and schizophrenia.