Brain Imaging Technology Uses Machine Learning to Identify Suicidal Thoughts
Researchers were able to identify subjects with suicidal ideation with 90 percent accuracy using a combination of brain imaging and AI algorithms.
A team of US medical researchers has developed a new and decidedly high-tech system for potentially preventing suicide. Using advanced brain-imaging technology and machine learning — a kind of subset of artificial intelligence — the system can flag patients with suicidal thoughts by essentially reading their minds.
It might sound like science fiction, but it's a technically accurate description of the new technique. The system detects and analyzes brain activity when the subject is asked to consider specific keywords and concepts related to suicide, such as “death” or “cruelty.”
When the brain activity data are processed by the system, electrical activity shows up on a map of the brain, with more intense feelings and thoughts generating specific color patterns in particular areas.
That's where the machine learning comes in. Using specifically coded algorithms, the AI system can detect significant pulses and patterns associated with suicidal thoughts. In a series of experiments using the technique, the system was able to accurately identify suicidal individuals with upwards of 90 percent accuracy.
“This type of analysis can assess a number of different component of the neural representation of a concept,” Marcel Just, professor of psychology at Carnegie Mellon University, told Seeker.
“One of the components that is measurable is the presence of various emotions,” Just said. “A classifier was able to distinguish between suicidal ideators and controls based on the how much of each emotion there was in the neural representation.”
The research, funded in part by the National Institutes of Mental Health, was published Oct. 30 in the journal Nature Human Behavior.
RELATED: AI Interviewers Entice Reluctant Soldiers to Report PTSD Symptoms
Researchers led by Just and the University of Pittsburgh's David Brent designed the experiments by preparing lists of keywords —10 related to death, 10 related to positive concepts, and 10 related to negative ideas.
The team presented the keywords to two groups of 17 people each. The first group consisted of patients with known suicidal tendencies. The second group – the control group – was made up of “neurotypical” persons, randomly selected people with no history of mental illness or suicidal behavior.
Both groups were outfitted with advanced brain scanning systems as they considered the list of keywords in various combinations. By analyzing their brain activity during this period, the system was able to distinguish between the suicidal group and the control group with 91 percent accuracy.
In a second set of experiments, the researchers used a similar approach to see if the machine learning system could distinguish between patients who had made a previous suicide attempt and those who had just thought about it. The program was able to identify those who had previously tried to take their own lives with 94 percent accuracy.
The new technique could have practical value for front-line clinicians who might be worried about potentially suicidal patients.
“We hope that information about what has been altered in a neural representation would be useful to a therapist or a therapy designer,” Just said. “Furthermore, one could subsequently assess the effectiveness of the therapy in terms of whether the alteration has been eliminated or reduced.”
RELATED: Brain-Computer Interface Allows Users to Compose Music With Only Their Thoughts
Just and other Carnegie Mellon researchers first developed the imaging system from brain activation signatures. The research has since been extended to identify emotions and multi-concept thoughts from their neural signatures.
For now, the technique requires specially calibrated brain scanning technology in the laboratory, Just said. But he hopes to change that.
“We are working on another project to determine whether suicidal ideation can be identified using [an electroencephalogram], which is a much cheaper and more widely available technology.”
WATCH: Which Countries Have the Highest Suicide Rates?