Predicting Behavioral and Mental Traits using Graph Convolutional Networks

Mental illness is widespread, affecting half of the population during their lifetime. One solution to mitigate this is to detect the onset of mental illness early, as well as choose individualized treatment options based on biological markers. These biomarkers can predict behavioral and mental traits, like the risk to develop a mental illness. They can potentially be found using neuroimaging techniques like magnetic resonance imaging. As the brain is highly complex, we need to use state-of-the-art data analysis techniques to find markers that predict behavioral and mental traits. Graph convolutional network is one such prediction technique that makes use of the fact that the brain is organized as a network. In our project, we will explore if graph convolutional networks can reliably predict behavioral and mental traits, which will set the groundwork to develop neuroimaging biomarkers of mental illness.

Faculty Supervisor:

Pierre Bellec

Student:

Partner:

Université de Lausanne

Discipline:

Life Sciences

Sector:

Health and Related Sciences & Technology; Artificial Intelligence

University:

Université de Montréal

Program: