Spotting suicide risk in the brain

Can we work out who is at risk of suicide by peering into their mind?
07 November 2017

Interview with 

Professor Marcel Just, Carnegie-Mellon University


Cartoon of the brain.


In many western countries, suicide is the leading cause of death in young men. One of the reasons for this is that spotting who’s really at risk - so doctors can prioritise helping them - is extremely difficult. But now a team in the US have discovered a telltale signature of suicide risk written into an individual’s brain activity, which they uncovered using machine learning. The technique could offer a way not only to find out who’s most at risk, but also to tailor therapies to reduce thoughts of suicide in the first place. Marcel Just told Georgia Mills what he’s found...

Marcel - It has to do with acquiring fMRI images of brain activity during a thought process. We ask people to think about some thought. We could ask them to think about a hammer or a chair, or whatever but, in this study, we asked them to think about concepts related to death, positive concepts in life, and some negative aspects of life. This allowed us to acquire the brain activation patterns associated with each of these concepts, and our machine learning techniques allowed us to determine the physical manifestation of an individual thought. We thought this approach might be useful in identifying people who are suicidal ideators.

Georgia - So you put people in a brain scanner - this fMRI. You get them to think about certain topics and then their brains light up in certain ways, depending on what they’re thinking about, and your machine learning is able to go through all the various different brains and find consistent patterns?

Marcel - Correct. Patterns, but not arbitrary patterns. Patterns corresponding to very specific thoughts and we see very specific alterations in the thought patterns, the neural signatures, of the people who are suicidal ideators.

Georgia - So you had people in your study who were people who had contemplated suicide and other people who hadn’t - is that right?

Marcel - That’s correct. We had a group of 17 people who were suicidal ideators and 17 who were neurotypical people who were matched in age and other demographics. At the first level, we just let a computer programme find any difference it could latch onto, and it could accurately distinguish the two groups. But, at a second level, we asked “can we specify what some of the differences are?”, and there we asked whether the emotional component of the neural signature differs between groups, and it did.

Now let me explain how this magic of discerning of the emotional signature came about. We’d previously done a study where people are asked to evoke in themselves various emotions and so we had this repository, this archive of emotion signatures. Then, in the current study of the people with suicidal ideation, we had their neural representations of various concepts such as death, and carefree, and funeral, and so on. We could ask how much of the neural signature of something like sadness that there was in each one. How much there was of shame. The neural representation of death in the people who are suicidal ideators has a larger component of sadness and of shame.

Georgia - What would you say is the significance of a finding like this?

Marcel - A lot of the significance is promissory - potential for the future. Imagine if we could use this to predict who’s going to make a suicide attempt, we could save lives with this if it works. It also has the potential for application to other psychiatric disorders. Many psychiatric disorders consist of an alteration of some kind of thinking. Maybe we can detect it and provide a complementary measure to conventional psychiatric diagnosis.

Georgia - If you are able to see then a difference in the brains of people who are contemplating suicide, could this give us any ideas of how to treat suicidal thoughts?

Marcel - Yes, and let me just say, we’re not just seeing differences in the brains, we’re seeing differences in the thought representations in the brains. I think that’s different, and it gives you an additional traction on potential therapy. If you just know that some area activates too much or too little - well, maybe you can do something to change that. But if you know the thought is altered in a particular way; if you know that death evokes an unusual amount of sadness, you could possibly direct your therapy at changing, eliminating, reducing that specific alteration. I think this gives, potentially, an extremely useful avenue to therapy.


Add a comment