AI Case Study
Researchers at Stanford University identify depression in patients with 80% accuracy using machine learning
Researchers at Stanford University have developed a machine learning system to identify signs of depression. Still at a very early stage, the system is able to successfully identify depression in patients with a 80% accuracy. The model was fed with video footage and trained on facial expressions, voice tone, and spoken words to distinguish between depressed and non-depressed people.
Healthcare Providers And Services
"In a study carried out by a team at Stanford University, scientists found that face and speech software can identify signals of depression with reasonable accuracy.
The researchers fed video footage of depressed and non-depressed people into a machine-learning model that was trained to learn from a combination of signals: facial expressions, voice tone, and spoken words. The data was collected from interviews in which a patient spoke to an avatar controlled by a physician.
While the new work is at an early stage, the researchers suggest that it could someday provide an easier way for people to get diagnosed and helped.
“Compared to physical illnesses, mental disorders are more difficult to detect,” the researchers write in a paper that is being presented at the NeurIPS AI conference in Montreal this week. “The burden of mental health is exacerbated by barriers to care such as social stigma, financial cost, and a lack of accessible treatment options [...] This technology could be deployed to cell phones worldwide and facilitate low-cost universal access to mental health care.”
The researchers caution that the technology would not be a replacement for a clinician. They add that the data used did not include any protected health information, such as names, dates, or locations. They also note that further work would be needed to ensure that the technology is not biased toward a particular race or gender."
"Depression is a huge problem for millions of people, and it is often compounded by poor mental-health support and stigma. Early diagnosis can help, but many mental disorders are difficult to detect. The machine-learning algorithms that let smartphones identify faces or respond to our voices could help provide a universal and low-cost way of spotting the early signs and getting treatment where it’s needed."
"In testing, it was able to detect whether someone was depressed more than 80% of the time."