Accounting for covariance in studies of multimodal microstructure - a model for multivariate quantification in health and disease
Stefanie A Tremblay1,2, Amir Pirhadi3, Zaki Alasmar4, Felix Carbonell5, Yasser Iturria-Medina6,7,8, Claudine J Gauthier1,2, and Christopher J Steele4,9
1Physics, Concordia University, Montreal, QC, Canada, 2Montreal Heart Institute, Montreal, QC, Canada, 3Electrical and Computer Engineering, Concordia University, Montreal, QC, Canada, 4Psychology, Concordia University, Montreal, QC, Canada, 5Biospective Inc., Montreal, QC, Canada, 6Neurology and Neurosurgery, Montreal Neurological Institute and Hospital, Montreal, QC, Canada, 7McConnell Brain Imaging Centre, Montreal Neurological Institute, Montreal, QC, Canada, 8Ludmer Centre for NeuroInformatics and Mental Health, Montreal, QC, Canada, 9Neurology, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany
Multivariate approaches have recently gained in popularity to address the physiological unspecificity of neuroimaging metrics and to better characterize the complexity of biological processes underlying behavior. However, approaches commonly used are biased by the covariance between variables. Here, we propose computing the voxel-wise Mahalanobis distance (MhD), as a measure of deviation from normality that accounts for covariance between metrics. We show that this measure can be linked to behavior and to potential physiological underpinnings by extracting metrics contributing most to the MhD. Integrative multivariate models are crucial to expand our understanding of the multiple factors underlying disease development and progression.
This abstract and the presentation materials are available to members only;
a login is required.