Machine learning can predict individuals at risk of suicide

Analyzing brain patterns with machine learning could predict people at risk of suicide, according to a study published in Nature Human Behavior.

In people contemplating suicide, words like “death” and “trouble” produce a specific signal in the brain that could be used to prevent an attempt. In the study, researchers had volunteers enter a brain scanner with a display that would flash words like “apathy,” “death” and “praise.”

"We could tell what emotion a person was feeling, we could tell what social interaction they're thinking about," Marcel Just, an author of the paper and the D.O. Hebb professor of cognitive neuroscience at Carnegie Mellon University, told NPR. "And we thought well, maybe the brain activation patterns of certain thoughts are altered in people who are thinking about suicide."

Brain activity was collected in response to each word and sent to a computer where a machine learning program would differentiate between suicidal brain signals and others. The program, which was able to differentiate between suicidal individuals 90 percent of the time, also distinguished between people who had attempted suicide and those who just thought about it.

"It correctly identified 15 of the 17 suicidal participants and 16 of the 17 controls," wrote Just.