HCP-EP Functional Dysconnectivity
Under the supervison of Dr. Katie Lavigne, I completed a one-year research project employing Multivariate Distance Matrix Regression (MDMR) to investigate functional connectivity differences between early-psychosis patients and healthy controls. We used resting state functional MRI and cognitive data from the Human Connectome Project for Early Psychosis, an open (but access-controlled) dataset.
Background
Connectivity changes in patients with early psychosis, broadly defined as within five (5) years of psychosis symptoms, has been studied using the Regions of Interest (ROI) approach which narrows the scope of investigation and limits our understanding of how the brain functions as a whole that may have impacted cognition. Additionally, analysis models typically impose a normality assumption of the underlying distribution of data, which, in a multivariate case, is impractical. Thus, we investigated how functional connectivity elucidated by resting-state functional MRI differs between early-psychosis patients and healthy controls in relation to cognition.
Multivariate Distance Matrix Regression (MDMR) offers a non-parametric alternative to multivariate linear regression and has been employed to Connectome Wide Association Studies (see Shezad et al., 2014 and Misaki et al., 2018). The mathematics behind MDMR is described in Zapala, & Schork (2012) and McArtor, Lubke, & Bergeman (2017).
Code
If you’re intersted in using MDMR, I have shared the code I used for my project here.
Citation
Fang, C., Danyluik, M., Lavigne, KM. (2025). Understanding Cognitive Impairment in Early Psychosis through Functional Brain (Dys)connectivity: A Whole-Brain Voxel-wise Analysis Approach. McGill Science Undergraduate Research Journal, 20(2).