Subvoxel quantitative susceptibility mapping of chronic white matter lesions: Enhancing lesion heterogeneity characterization and providing complementary biomarkers for myelin integrity in multiple sclerosis.

OBJECTIVES: Quantitative susceptibility mapping (QSM) has been used to classify chronic white matter lesions (WMLs) in multiple sclerosis (MS), which exhibit varying myelin and iron content. Compared to QSM, subvoxel-QSM can further differentiate susceptibility alterations arising from demyelination and iron accumulation. This study aims to evaluate the performance of subvoxel-QSM in classifying chronic WMLs in relapsing-remitting MS (RRMS) and explore their heterogeneity.

METHODS: In 57 RRMS patients, WMLs segmented manually were classified into 5 QSM lesion types (isointense, hypointense, hyperintense, lesions with hypointense rim, and paramagnetic rim lesions [PRLs]). Then, visual and quantitative characteristics in subvoxel-QSM (χpara and χdia) were compared within each subtype. Next, PRLs were classified into two subgroups by hyperintensity presence in χdia and compared as above. Meanwhile, the neurite density index (NDI), from the multi-compartment diffusion model, quantified lesional microstructural damage.

RESULTS: A total of 1002 WMLs from 57 RRMS patients underwent QSM classification, demonstrating the following distribution: 119 (11.88%) QSM-isointense, 665 (66.37%) QSM-hyperintense, 17 (1.70%) QSM-hypointense, 3 (0.30%) QSM-hypointense rim, and 198 (19.76%) PRLs. Subvoxel-QSM analysis identified 199 PRLs in χpara (195 from QSM-PRL and 4 from QSM-hyperintense), grouped as 130 (65.33%) χpara-PRL without myelin and 69 (34.67%) with myelin. The χpara-PRL without myelin exhibited significantly lower NDI than those with myelin (P =0.006). In QSM-isointense lesions, 84.03% (100/119) showed χpara-isointensity and χdia-isointensity, while 15.97% (19/119) exhibited χpara-hypointensity and χdia-hypointensity. Moreover, the χpara-isointense/χdia-isointense subgroup exhibited significantly higher χdia values (P =0.019). Among QSM-hyperintense lesions, 92.78% (617/665) displayed χpara-hyperintensity with χdia-hypointensity and 6.62% (44/665) showed χpara-isointensity with χdia-hypointensity, while 0.60% (4/665) appeared as PRLs in χpara. Moreover, the χpara-hyperintense/χdia-hypointense exhibited significantly lower NDI and χdia values (all P <0.01).

CONCLUSION: Subvoxel-QSM enhances lesion heterogeneity characterization in RRMS, providing complementary biomarkers for myelin integrity beyond conventional QSM, and its detection of repair gradations may help with personalized therapeutic strategies and clinical trial endpoint optimization.