Photo: Carnegie Mellon University

More than 44,000 Americans commit suicide in the U.S. each year. For young adults, it is the second-leading cause of death.

Doctors and mental health professionals commonly have to rely solely on behavioral analysis to evaluate an individual’s risk of self-harm or suicide. This method makes it extremely difficult to pinpoint who may follow through and act on suicidal thoughts.

Marcel Just, of the Department of Psychology at Carnegie Mellon University, and David Brent, Endowed Chair in Suicide Studies at the University of Pittsburgh, focused on finding a way to highlight biological indicators that could better predict suicide risk.

They used functional magnetic resonance imaging (fMRI) to collect brain scan images of 34 young adult volunteers – 17 of whom were considered “suicidal ideators” and the other 17 acted as the control group.

The researchers created two lists of words they believed may reveal brain activity patterns associated with suicide – the first list of 10 words related to negative concepts while the second list of 10 words was associated with positive ideas.

While the participants were in the fMRI scanner, they were asked to think about each of the words as they appeared on a screen for a few seconds at a time. As the participants thought about each of the words, the team captured their brain activity.

The researchers also developed a machine-learning algorithm, which was applied to six of the word-concepts that best discriminated between the two participant groups, according to the team. The six words were death, cruelty, trouble, carefree, good and praise.

The machine-learning program was trained to detect patterns and distinguish people with suicidal thoughts from neurotypical, or healthy-minded, individuals. The program did so with a 91 percent success rate.

The program was also able to identify nine suicidal ideators who had previously made a life-ending attempt, and distinguished them from eight individuals who thought about suicide, but didn’t actually go through with acting on it.

“Our latest work is unique insofar as it identifies concept alterations that are associated with suicidal ideation and behavior, using machine-learning algorithms to assess the neural representation of specific concepts related to suicide. This gives us a window into the brain and mind, shedding light on how suicidal individuals think about suicide and emotion related concepts,” said Just.

The findings were published in the journal Nature Human Behavior. Just and Brent do acknowledge that the findings would need to be replicated with a larger participant pool for validation, but the study does demonstrate the potential for using brain scanning technology and machine-learning to evaluate an individual’s mental health more effectively by identifying distinct neural signatures that suggest an elevated risk in attempting suicide.