Depression and suicidal thoughts are usually identified through conversations, questionnaires and a clinician’s judgment.
These methods are essential, but researchers at the University of Southern California are exploring whether signals from the body could provide doctors with another source of information.
The work is part of a research project called PRECOG, which began in June 2023. The project brings together specialists in engineering, brain science, artificial intelligence, language and psychiatry to search for measurable biological patterns linked to depression and suicidal thinking.
The idea addresses a difficult problem in mental health care. Unlike many physical illnesses, depression does not have a single blood test, brain scan or other laboratory test that can confirm the diagnosis, and people may not always report what they are experiencing.
This problem can be especially important when doctors are trying to understand suicide risk. The USC team is therefore studying signals that people do not easily control on purpose, including electrical activity in the brain, eye movements and tiny changes in sweating.
Participants completed a task built around language. They read 160 statements about themselves that differed in emotional tone, then indicated whether they agreed or disagreed while researchers recorded several body and brain responses at the same time.
One part of the project used electroencephalography, or EEG, which records the brain’s electrical activity through sensors placed on the scalp. The researchers used a 64-channel system to see how quickly the brain responded while participants processed emotionally meaningful words and sentences.
The brain patterns were not identical in healthy participants, people with depression and people with suicidal thoughts. One analysis found that some of the clearest differences appeared about 300 to 600 milliseconds after words were shown, a period when the brain is working out meaning and emotional importance.
Other analyses focused on particular brain responses. The researchers reported changes in signals involved in processing negative information and found another response that appeared to become weaker as suicidal thinking became more severe.
The team also studied eye movements because where a person looks can reveal how attention is being directed. An infrared eye-tracking system recorded gaze while participants completed the same sentence task, and computer models searched the recordings for patterns associated with mental health status.
Horizontal eye movements contained especially useful information in the researchers’ models. When negative statements appeared, healthy participants tended to show more organized viewing patterns, while participants with suicidal thoughts showed gaze patterns that were more spread out or disengaged.
A third approach examined skin conductance, which changes when sweat glands respond to emotional or physical arousal. Because this response is controlled largely automatically, it may offer information that is different from what a person says on a questionnaire.
The researchers found that reactions to negative words provided the strongest information in this part of the work. People with depression and suicidal thoughts showed different patterns of physical arousal, suggesting that the body’s automatic response to distressing information may contain useful clues.
Artificial intelligence was then used to find combinations of these complex signals that could separate groups of participants. The long-term aim is not to let a computer diagnose someone on its own, but to give clinicians additional evidence that could support a broader mental health assessment.
The PRECOG project has produced several research papers. The eye-movement study, led by Kleanthis Avramidis, was published in npj Digital Medicine, while related EEG studies were published in the Journal of Affective Disorders, Communications Biology and Translational Psychiatry. A study of skin conductance was made available on the arXiv preprint server and had not yet gone through the same journal publication process described for the peer-reviewed papers.
The findings are promising because they show that depression and suicidal thinking may be associated with measurable patterns across several different systems in the body. Using several signals together may eventually be more useful than relying on any single biological marker.
There are important limits, however. These studies show associations and classification patterns, not a proven test that can predict whether a particular person will attempt suicide, and AI models can perform less well when used with people who differ from those included in the original research.
The tasks were also performed under controlled research conditions with specialized equipment. Before such methods could become part of everyday care, they would need to be tested in larger and more diverse groups and shown to add meaningful information beyond careful clinical interviews and existing assessments.
The most important message is therefore not that brain signals, eye movements or sweat can replace psychiatrists. Instead, the research suggests that objective body signals may one day give clinicians an extra layer of evidence, particularly when a person’s symptoms or level of risk is difficult to judge from self-report alone.
Overall, the USC work is an interesting step toward more measurable mental health assessment, but it remains an emerging research approach rather than a ready-to-use suicide prediction tool. Its real value will depend on future studies showing that the methods are accurate, fair across different populations and useful enough to improve clinical decisions and patient care.
If you care about mental health, please read studies about 6 foods you can eat to improve mental health, and B vitamins could help prevent depression and anxiety.
For more health information, please see recent studies about how dairy foods may influence depression risk, and results showing Omega-3 fats may help reduce depression.
Source: University of Southern California.
