Suicidal thoughts are difficult to diagnose. People who have these thoughts often don't disclose them to other people, let alone healthcare professionals. Finding out whether people have these thoughts could lead to better interventions and targeted suicide prevention.
Scientists at Carnegie Mellon University and the University of Pittsburg have therefore attempted to identify suicidal thoughts as they occur in the brain. In a study published in Nature Human Behaviour, researchers scanned the brains of people who have suicidal thoughts. By studying how their brains light up upon hearing certain words (such as "death"), the researchers were able to train a machine-learning algorithm to spot certain activation patterns that indicate suicidal tendencies.
The researchers looked at 17 people between the ages of 18 and 30 who had reported suicidal ideation (thoughts of suicide or unusual preoccupation with suicide) to therapists. As a control, they also recruited 17 people who hadn't had these thoughts.
They then scanned the paricipants' brains using an fMRI machine whilst being shown a series of 30 words, both positive (e.g. "carefree" and "bliss") and suicide-related concepts ("death", "desperate", "apathy"). They were also shown words such as "boredom", "trouble", and "cruelty", which were seen as negative but not related to suicidal thoughts.
When the participants were shown the words that were related to suicide, the brains of the patients who had reported suicidal ideation lit up in a way that wasn't seen in the control group. Of the words shown, "death" was the concept that showed the most striking difference between the two groups.
The resulting algorithm was able to identify people with suicidal thoughts with an astonishing 91 percent accuracy.