Paramagnetic rim lesion formation is predicted by the initial gadolinium-enhancing lesion diameter

Abstract

Background: Paramagnetic rim lesions (PRLs) are a magnetic resonance imaging (MRI) marker of compartmentalized intraparenchymal inflammation.

Objectives: The primary objective was to investigate clinical, demographic, and MRI factors that may be predictive of the future formation of PRL.

Methods: This is a retrospective analysis of longitudinal data. Patients were included if they had ⩾1 gadolinium-enhancing lesion on any historical MRI and follow-up scan(s) ⩾6 months afterward on a standardized 3T MRI using the filtered phase component of a susceptibility-sensitive sequence ("SWAN"). Regression and machine-learning models were used to identify the predictive ability of demographic, clinical, immunological, treatment-related, and MRI predictors of PRL formation.

Results: A total of 64 patients having 229 contrast-enhancing lesions (CELs) were included. Among all predictors, the diameter of the initial enhancing lesion was the most influential for determining subsequent PRL formation; every millimeter increase in diameter increased the risk of PRL formation by 44%. Other factors did not contribute additional information; the administration of steroids was not associated with any effect.

Conclusions: The long-axis diameter of a CEL is the best translational predictor of subsequent PRL formation at follow-up. This measure holds promise as a method to identify patients at high risk of chronic active lesion formation during the acute inflammatory window.

Keywords: MRI; biomarkers; multiple sclerosis; paramagnetic rim lesion; steroids glucocorticoids.

Figure 1:

Overview of the process to characterize baseline MRI features of contrast-enhancing lesions and their follow-up assessment on susceptibility-sensitive imaging A 34-year-old woman with relapsing-remitting multiple sclerosis. Axial T2-weighted and T1-weighted post-contrast scans (left) at baseline showing multiple ring-enhancing lesions with hypointense T2 rims and moderate perilesional edema. Follow-up scans show these lesions 8 years later, on T2-FLAIR and the filtered-phase SWAN at 3T (right). Two of the 3 highlighted lesions were determined to be PRL at follow-up, as labeled and seen on magnified areas.

Figure 2:

Flow diagram for patient assessment and inclusion

Figure 3:

Association matrix between baseline clinical, demographic and MRI variables with subsequent PRL formation Stars represent p-values: *<0.05; **<0.01; ***<0.001. Association sizes between continuous variables are Pearson correlations; between continuous and dichotomous variables are point-biserial correlations, and between dichotomous variables is the phi coefficient. “DWI: bright rim” and “T2: bright rim” categories were not included here due to insufficient sample size in relation to variable categories with missing data (baseline phase characteristics).

Figure 4:

Results of the extreme gradient boost model The final model for XGBoost converged on a single dominant factor, the size of the contrast-enhancing lesion in the long axis.

Figure 5:

Scatter plot comparison of CEL-to-PRL conversion, as compared by the initial CEL long-axis diameter.