This AI uses brain scans to figure out if someone is depressed

Diagnosing depression is tricky and usually boils down to patients providing their doctor with a list of problems. Diagnosing the condition in the future may be far faster, though, as one AI has learned to identify depressed people using nothing more than their brain scans. The AI was trained to identify the brain patterns of individuals who are depressed versus those who aren't, something itself achieved using fMRI machines while quizzing people on how they think and feel about certain words.

Depression is a difficult beast because, unlike many illnesses, it is possible for someone to hide the symptoms until it too late to help them. Suicide prevention measures have largely focused on encouraging individuals experiencing depression symptoms to seek help for an official diagnoses or to call a crisis hotline is suicide seems imminent, but technology like artificial intelligence may usher in a whole new world of intervention.

Researchers with the University of Pittsburgh and Carnegie Mellon trained an AI to notice the pattern differences in a person's brain when they're talking about certain subjects. A depressed person's brain may show a different light pattern during talks about death than someone who isn't depressed, for example. While a person can lie, their brain scan can't.

During testing, the AI was able to achieve a greater than 90% accuracy in some cases, though more testing will have to be done to demonstrate its effectiveness across larger group sizes. Should such technology ever become a typical fixture in medicine, there would still be hurdles to overcome, such as the high cost associated with getting the brain scans and even patient willingness to undergo them.

As it stands, artificial intelligence won't be replacing psychiatrists any day soon, but the technology is promising. In the future, such scans could be a routine part of a mental health patient's assessments, helping a doctor get an idea about which thoughts a patient may otherwise not be relaying to their doctor.

SOURCE: WIRED