1Advanced
MRI Development, Florey Institute of Neuroscience and Mental Health,
Melbourne, VIC, Australia; 2Department of Medicine, University of Melbourne,
Melbourne, VIC, Australia; 3Wellcome Trust Centre for
Neuroimaging, UCL Institute of Neurology, London, England, United Kingdom; 4Department
of Nuclear Medicine, Austin Health, Melbourne, VIC, Australia; 5The
Australian E-Health Research Centre, CSIRO, Brisbane, QLD, Australia
Recent work has focused on the development of more interpretable quantitative diffusion MRI measures that can be associated with a specific fibre bundle within a voxel containing crossing fibres. Unfortunately, traditional methods for voxel-based analysis of 3D images cannot be applied to analyse these measures, since the definition of the local neighbourhood for smoothing and cluster construction is ambiguous when adjacent voxels may contain different fibre populations. We propose a novel statistical framework for whole-brain voxel-based analysis of bundle-specific diffusion MRI measures, which uses group-average fibre tractography to define the local neighbourhood for clustering and smoothing purposes.