Advocate Aurora Health researchers studying AI that could detect mental health conditions
WISCONSIN (CBS 58) -- We use computers for almost everything these days, and soon we may use a computer to diagnose your mental health. Advocate Aurora Health researchers have joined a clinical trial studying an artificial intelligence telehealth platform that analyzes a patient’s words, voice and facial expressions to detect signs of mental health conditions like depression, anxiety, and suicidal thoughts.
Sangit Rawlley, president and co-founder of Aiberry, says diagnosing mental health issues is a critical need.
“The CDC estimates that there has been a four-fold increase in depression and anxiety cases since last year, and the fact is that less than 5% of adults today get screened for any kind of mental illness,” he said.
Rawlley says the current way to get screened for depression involves filling out a lot of forms.
“There's a PHQ-9 form that you have to fill out, the depression form, anxiety Columbia form, suicide ideation assessment. so on and so forth.”
This study will use traditional forms of diagnosis and also Aiberry’s artificial intelligence. Researchers plan to enroll 1,000 patients ages 13 to 79 in the study. The platform will be used to:
- Take a video of a patient during their interview with a mental health professional
- Analyze multiple data channels, including video, audio and speech, to extract patterns to a particular mental health disorder
- Assign a score showing the likelihood the patient has a particular mental health disorder
Dr. Mindy Waite, research scientist with Advocate Aurora Research Institute’s Ed Howe Center for Health Care Transformation, says the study has already started.
“We're going to have it open for at least a year, our goal is to enroll 200 patients or community members. They do not have to be our own patients, so anybody off the street,” she said.
How good is artificial intelligence at diagnosing mental health issues? A 2009 study of more than 50-thousand patients showed that doctors correctly identified depression just over 47-percent of the time. But Rawlley says early research shows this artificial intelligence has a better success rate.
“We created a proof of concept based on a limited data set, and our accuracy based on that limited data set was around 70% accurate,” he said.
Researchers see a future where this artificial intelligence runs in the background at the doctor's office, any interaction which could catch mental health problems before they get too serious.
“This is maybe offered up at every, every time a person comes into a clinic, whether you're there for your annual screening, whether you're there for surgery, anytime you come in contact with your health care provider I could see this thing offered as a service,” Waite said.
If the studies bear fruit, Aiberry’s president says this AI could be brought to market to spot depression in the first quarter of next year, and anxiety and suicidal ideation could be added after more research comes in
To learn more about Advocate Aurora’s research, visit aurora.org/research.