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

2023-12-28

arXiv.org

AI Summary

To address the lack of quantitative identification and characterization methods for chronic active lesions (Rim+ lesions) in multiple sclerosis, this study proposes the first end-to-end quantitative susceptibility mapping (QSM)-based analytical framework. The method integrates unsupervised level-set segmentation (RimSeg) with local binary pattern (LBP) texture features to enable pixel-wise detection, morphological modeling, and radiomic characterization of Rim+ lesions. A novel deep regression network, QSMRim-Net, significantly improves quantification accuracy, achieving a correlation coefficient of 0.91 (95% CI: 0.88–0.93) against expert annotations and an MSE of 0.85. RimSeg attains a Dice score of 78.7%, and Rim+ lesion detection yields a precision-recall AUC of 0.737. This work establishes the first fully automated, reproducible, and multidimensional quantitative assessment pipeline for Rim+ lesions, offering a new paradigm for monitoring disease progression and evaluating therapeutic response in multiple sclerosis.

Abstract

Background: Rim+ lesions in multiple sclerosis (MS), detectable via Quantitative Susceptibility Mapping (QSM), correlate with increased disability. Existing literature lacks quantitative analysis of these lesions. We introduce RimSet for quantitative identification and characterization of rim+ lesions on QSM. Methods: RimSet combines RimSeg, an unsupervised segmentation method using level-set methodology, and radiomic measurements with Local Binary Pattern texture descriptors. We validated RimSet using simulated QSM images and an in vivo dataset of 172 MS subjects with 177 rim+ and 3986 rim-lesions. Results: RimSeg achieved a 78.7% Dice score against the ground truth, with challenges in partial rim lesions. RimSet detected rim+ lesions with a partial ROC AUC of 0.808 and PR AUC of 0.737, surpassing existing methods. QSMRim-Net showed the lowest mean square error (0.85) and high correlation (0.91; 95% CI: 0.88, 0.93) with expert annotations at the subject level.

Problem

• Quantitatively identifying chronic active MS lesions
• Characterizing rim+ lesions using QSM imaging
• Addressing lack of quantitative analysis in existing literature

Innovation

• Unsupervised segmentation using level-set methodology
• Radiomic measurements with Local Binary Pattern descriptors
• Quantitative identification of rim+ lesions on QSM