Quantitative radiomic analysis of MS based on MRI, performed by extracting imaging correlates of MS pathophysiology, has been recognized as critical for more accurate and earlier diagnostics, improved precision in clinical decision-making, and more powerful outcomes in trials for targeted MS therapeutics. Unfortunately, the application of these approaches in MS are still in their infancy and several challenges unique to MS remain to be solved before radiomic analyses can be translated in clinical and research practice. A major challenge for the diagnosis and monitoring of MS is to disentangle the heterogeneity of white matter lesions, both from an etiologic perspective and in the degree of tissue injury. The presence of confluent clusters of lesions that are comprised of multiple lesions, particularly around the ventricular horns, poses a key challenge for dissecting this heterogeneity in lesions: while histopathology shows great phenotypic variability both within and between lesions, most neuroimaging studies average metrics across lesion clusters losing the valuable information about each individual lesion. In this proposal, we propose to use advanced statistical analysis of signal intensity from multi-parametric imaging to distinguish individual lesions and more accurately phenotype them, and thus facilitate much greater understanding of an individual patients burden of disease and easier application to clinical practice and research studies. We will also create tools that will facilitate the adoption of these techniques in the clinic. We will validate these approaches by comparison to expert neuroradiologist assessments and determine added value of these techniques. We further propose to develop a state-of-the-art method for the discovery of covariate effects in diffuse processes in the normal-appearing white matter and gray matter, which will facilitate many potential studies of MS pathology and therapeutics. We will also develop software implementations and educational resources to disseminate the methods developed.
Our overarching goals are to develop statistical methods for the analysis of imaging acquired on multiple scanners for the study of multiple sclerosis (MS). We will develop and validate statistical methods for identifying individual lesions and quantifying the central vein in MS lesions for improved MS diagnostics as well as techniques for automatically identifying lesion phenotypes in a novel coordinate system. Finally, we will develop methods that allow for the study of patterns of neuropathological change across an image to study diffuse changes in the normal-appearing white and gray matter of the brain.