David Benrimoh wanted to improve mental health outcomes for those suffering from, or at risk from suffering from, psychosis. But he needed support to bring his research project to life.
That’s where CDSI came in: For three straight years, CDSI’s Convergent Research Themes (CRT) program has been supporting researchers whose interests don’t fit neatly into a box. Since its inception in 2022, the program has offered more than $136,000 to a dozen projects that cross disciplinary boundaries. In the program’s first year, the areas of focus included extreme weather events in smart cities, improving clinical care, and predicting the decisions of consumers.
As a ş«ąúÂăÎč-trained neuropsychiatrist (MD 2016), Benrimoh and his team wanted to understand how machine learning could be used to make predictions with different forms of data that aren’t easily directly compared when diagnosing and treating patients with a range of mental health conditions, including psychosis and psychosis risk. These could include everything from a patient’s responses to questionnaires, to data from their cell phones. These data could potentially be used to inform the development and targeting of new treatments aimed at preventing or delaying the onset of these conditions.
In previous studies, researchers sometimes found that adding these additional variables, such as phone data, to their data didn’t improve predictions much, or sometimes made them worse. But Benrimoh and his team found that “if you use what's called intermediate fusion, which is where you basically create latent representations of the data using models like auto encoders, and then you use that during your prediction, you can actually perform better when you have all the data together, without overfitting.”
“And so we actually were able to show that we could improve upon the results of some previous work, and now, having all the data together — questionnaires and phone-based data — actually improved outcomes, improved predictions, over having less data, which was which was very nice, because it gives us confidence that we can use this approach in future projects, like the ones that we're working on right now.”
Benrimoh says the CRT funding was especially useful for someone like him, still early in his career and without the same access to funding as more senior researchers. “The CDSI grant was super helpful in giving me a quick boost of initial funding, which let me bring on a really great student who was interested in working with me, but who otherwise I wasn't able to fund” he says. This work, as well as other work which was completed using follow-on funding from the Fonds de recherche du Québec, is currently being readied for publication, and Dr. Benrimoh's lab will be releasing an open-source toolbox for other researchers in this space to use.
If you’re a researcher with an idea for a project that crosses disciplinary boundaries, start planning now so you can be ready for our fall applications period.
Learn more about past projects and how the CRT can help you in your research.
