RimSet: Quantitatively Identifying and Characterizing Chronic Active Multiple Sclerosis Lesion on Quantitative Susceptibility Maps

Background: Rim+ lesions in multiple sclerosis (MS), detectable viaQuantitative Susceptibility Mapping (QSM), correlate with increased disability.Existing literature lacks quantitative analysis of these lesions. We introduceRimSet for quantitative identification and characterization of rim+ lesions onQSM. Methods: RimSet combines RimSeg, an unsupervised segmentation method usinglevel-set methodology, and radiomic measurements with Local Binary Patterntexture descriptors. We validated RimSet using simulated QSM images and an invivo dataset of 172 MS subjects with 177 rim+ and 3986 rim-lesions. Results:RimSeg achieved a 78.7partial rim lesions. RimSet detected rim+ lesions with a partial ROC AUC of0.808 and PR AUC of 0.737, surpassing existing methods. QSMRim-Net showed thelowest mean square error (0.85) and high correlation (0.91; 95with expert annotations at the subject level.