Carnegie Mellon/Pitt team receive $3.8 million NIMH grant to study suicide
A $3.8 million grant from the National Institute for Mental Health (NIMH) has been awarded to Carnegie Mellon's Marcel Just and the University of Pittsburgh's David A. Brent, who will be using the grant to expand on their innovative research on how images of the brain can identify individuals with suicidal thoughts.
Their research, known as the Predicting Risk Imaging Suicidal Minds (PRISM) project, made headlines in 2017 when their initial findings were published. In the study, they examined brain activity in suicidal and non-suicidal participants by comparing images of the brain's response when participants heard positive words, negative words, and words related to suicide. They identified five areas of the brain that were identifiably different between suicidal participants and non-suicidal participants.
Brent and Just then trained a machine learning algorithm on the brain data, and tested it to see how well it could correctly categorize brain images as either suicidal or non-suicidal; it returned results with 91 percent accuracy. When the algorithm was trained only on the data from suicidal brains, it identified images with 94 percent accuracy.
The NIMH grant will fund the research for a further five years.
"The cornerstone of this project is our recent ability to identify what concept a person is thinking about based on its accompanying brain activation pattern or neural signature," said Just, the D.O. Hebb University Professor of Psychology in Dietrich College, in a statement. "We were previously able to obtain consistent neural signatures to determine whether someone was thinking about objects like a banana or a hammer by examining their fMR brain activation patterns. But now we are able to tell whether someone is thinking about 'trouble' or 'death' in an unusual way.”
One limitation of Just's and Brent's past research — as with many psychology studies — is their small sample size (34 participants). While this number was enough to provide valid results, in future work the greater number of participants allowed by the NIMH funding will grant additional confirmation and further insight, and allow them to compare imaging from patients with a greater range of different mental states to broaden the way they are able to interpret the images.
While it's generally agreed that a way to track biological predictors of suicide is needed, some think the PRISM project isn't the best solution.
Some experts cautioned against placing too much confidence, too soon, in the machine learning approach. “They used a method called ‘cross validation’ to both train and test their machine learning algorithm on the same small data set," explained Derek Hill, a professor of medical imaging science at University College London. "While this is a widely used approach, it is not a true replication study, so it isn’t yet clear whether their algorithm would work on another separate group of patients.”
The largest point of criticism, however, was the method's applicability in a clinical setting.
Brain imaging is expensive, and suicide is an extremely frequent problem. It's unrealistic to expect a switch to this advanced method from simple screening surveys in the doctor's office. “In terms of predicting suicide risk, it is unlikely to move the field forward," said psychiatrist Seena Fazel of the University of Oxford, in response to the 2017 study.
However, if the PRISM method is implemented in a clinical setting, it could potentially help reduce the high number of false positives produced by standard screening surveys — as well as increase the accuracy of screenings.
"Suicide is the second leading cause of death among young adults in the U.S., and current assessment methods rely entirely on patients self reporting and doctors' observations," said Brent. "Any new inroads to better diagnosis and treatment have the potential to save lives."