Computers to spot patients at risk of suicide

Suicidal patients can be identified by artificial-intelligence analysis of brain scans, according to a study.

A computer program proved 90 per cent accurate at spotting suicidal people from the way their brains responded when prompted to think about death.  Although the study was small and required costly MRI scans, scientists have said that it held out the long-term possibility of methods to spot those in imminent danger of taking their own lives.



Researchers at Carnegie Mellon University tested and scanned 17 suicidal patients and 17 control subjects and trained a machine learning algorithm to use scan data to distinguish the groups.

It correctly identified 15 suicidal patients and 16 non-suicidal ones and correctly distinguished which suicidal patients had actually attempted to kill themselves 16 times out of 17.

Barry Horwitz, of the US National Institutes of Health, said the findings could become ''a major medical tool'' if confirmed by larger studies.