Subvoxel quantitative susceptibility mapping reveals distinct susceptibility patterns and clinical associations of juxtacortical paramagnetic rims in multiple sclerosis

Abstract

Background:

Juxtacortical paramagnetic rims (JPRs) have emerged as potential biomarkers of cortical pathology in multiple sclerosis (MS), yet their susceptibility characteristics and clinical significance remain unclear.

Objectives:

To characterize the heterogeneity of cortical lesions (CLs), with a particular focus on JPRs and clinical relevance.

Methods:

Sixty-four patients underwent 3T magnetic resonance imaging (MRI), including structural imaging, multi-echo gradient echo, and Magnetic Resonance Image Compilation (MAGiC) sequences. Paramagnetic (χ_para) and diamagnetic (χ_dia) maps were reconstructed using subvoxel quantitative susceptibility mapping. CLs were identified, and those with JPRs were defined as JPR-associated CLs. Susceptibility metrics were compared across lesion subtypes and correlated with clinical and imaging measures.

Results:

Among 499 CLs, 15% were JPR-associated CLs, which exhibited elevated χ_para measurements and reduced χ_dia measurements. Patients with JPRs had a higher number of paramagnetic rim lesions (95% CI [2.445, 4.738], p < 0.001), annualized relapse rates (95% CI [0.130, 0.297], p < 0.001), and lower digit span test scores (95% CI [−2.504, −0.473], p = 0.004). The absolute χ_para and χ_dia values in all CLs were correlated with longer disease duration (r = −0.461, p = 0.001; r = 0.457, p = 0.001).

Conclusion:

JPRs are a characteristic feature of CLs associated with greater clinical and radiological burden in MS.

Keywords

Multiple sclerosis cortical lesion juxtacortical paramagnetic rims quantitative susceptibility mapping

Introduction

Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system characterized by demyelination, neurodegeneration, and widespread tissue heterogeneity involving both white and gray matter.^1,2 Although most imaging studies have focused on white matter lesions (WMLs), increasing evidence indicates that cortical pathology plays a critical role in disease progression and clinical outcomes.^3–5

Cortical lesions (CLs) in MS differ from WMLs in their inflammatory profile, cellular composition, and tissue microstructure.^6,7 Compared with white matter, CLs generally exhibit less overt inflammatory activity.^8–10 Although previous studies have demonstrated the sensitivity of quantitative susceptibility mapping (QSM) to cortical pathology, reported susceptibility patterns in CLs remain variable and sometimes conflicting,^11,12 suggesting lesion heterogeneity in cortical regions. Recently, juxtacortical paramagnetic rims (JPRs) have been described as a distinct susceptibility feature at the white matter-cortex interface in a subset of CLs.¹³ Although JPRs have been proposed as potential markers of cortical pathology, their susceptibility characteristics and clinical relevance remain incompletely understood.

Magnetic susceptibility measured by QSM is influenced by multiple tissue components.^14,15 Several source separation approaches have been developed to decompose QSM signals into paramagnetic and diamagnetic components.^16–18 Among these, subvoxel QSM employs iterative voxel-wise optimization and refined signal modeling to characterize paramagnetic and diamagnetic susceptibility features in the cortex.¹⁷

In this study, we applied subvoxel QSM to investigate paramagnetic and diamagnetic alterations in CLs, with a particular focus on JPRs, to delineate their distinct susceptibility patterns and clinical relevance.

Materials and methods

This study was approved by the Institutional Review Board of the First Affiliated Hospital of Chongqing Medical University, China (approval number: 2022-027). Written informed consent was obtained from all patients.

Patients

Sixty-seven consecutive patients with relapsing-remitting MS (RRMS) were prospectively recruited between September 2023 and December 2024 from the Department of Radiology, the First Affiliated Hospital of Chongqing Medical University. RRMS diagnosis followed the 2017 revised McDonald criteria.¹⁹ Exclusion criteria included magnetic resonance imaging (MRI) contraindications, intravenous corticosteroid treatment within 2 months before imaging, and scans with artifacts or incomplete clinical data. Clinical disability was assessed with the Expanded Disability Status Scale (EDSS) and cognitive performance with the digit span test (DST) and symbol digit modalities test (SDMT).

Data acquisition

MRI was performed on a 3.0 T magnetic resonance (MR) scanner (Signa Premier, GE Healthcare) with a 48-channel phased-array head coil. The standard MS protocol included three-dimensional (3D) magnetization-prepared rapid gradient echo (MPRAGE), 3D fluid-attenuated inversion recovery (FLAIR), and 3D double inversion recovery (DIR). Sequences for susceptibility separation consisted of 3D multi-echo gradient echo (GRE) and Magnetic Resonance Image Compilation (MAGiC). Detailed imaging parameters are provided in the Supplemental Material.

Reconstruction of paramagnetic and diamagnetic susceptibility maps

3D GRE data were processed using the STI-Suite toolbox (version 3.0, Martinos Center, Harvard Medical School) to reconstruct QSM. The processing pipeline included Laplacian-based phase unwrapping, background field removal using the V-SHARP algorithm, and susceptibility inversion with the STAR-QSM method.

MAGiC data were processed using SyMRI software (version 8.0, SyntheticMR, Linköping, Sweden) to generate quantitative T₂ maps, from which R₂ maps were calculated. GRE data and R₂ maps were then integrated using the subvoxel QSM separation algorithm described by Li et al.¹⁷ to separate paramagnetic (χ_para) and diamagnetic (χ_dia) susceptibility sources. This method employs a voxel-specific signal model and an iterative optimization framework to estimate χ_para and χ_dia components. The signal model is expressed as:

\begin{array}{l} S (t) = M_{0} e^{- (R_{2} + a | χ_{p a r a} | + a | χ_{d i a} |) t} . \\ e^{i (ϕ_{r e s} + ϕ_{b g (t)} + 2 π γ B_{0} (D * (χ_{p a r a} + χ_{d i a})) t)} \end{array}

where M₀ is the extrapolated signal magnitude at zero echo time, R₂ is the transverse relaxation rate, φ_res denotes the time-independent residual phase, φ_bg(t) is the echo time-dependent background phase, D is the magnetic dipole kernel, γ is the gyromagnetic ratio, B₀ is the main magnetic field strength, and * indicates spatial convolution.

All subvoxel QSM processing was performed in MATLAB (R2019a, MathWorks, Natick, MA). The resulting χ_para, χ_dia, and QSM maps were co-registered to the MPRAGE images using Advanced Normalization Tools (ANTs).²⁰

Image analysis

CLs were assessed by consensus between two radiologists (Q.Z. and Z.Y.) using combined MPRAGE and DIR sequences across multiple continuous image sections. CLs were required to be visible on at least two consecutive slices to exclude artifacts, and lesions smaller than three voxels were excluded. All lesions included in this study were non-acute, as confirmed by the absence of gadolinium enhancement. Lesions were manually segmented on MPRAGE using ITK-SNAP (v3.8.0).²¹ For leukocortical lesions (LCLs), only the cortical portion was manually outlined. After segmentation, all lesion regions of interest (ROIs) were overlaid on QSM images for quality control, during which JPR assessment was performed. JPRs were identified as linear QSM-hyperintensities extending along the cortex and closely adherent to the cortical ribbon. QSM-hyperintense juxtacortical lesions (JCLs), which are WMLs located adjacent to the cortex, were additionally included for comparison with JPR-associated CLs to determine whether lesions with similar hyperintense appearances on QSM exhibit distinct paramagnetic and diamagnetic susceptibility patterns. Reference ROIs were outlined on adjacent normal-appearing (NA) tissues (cortical gray matter for all CLs and white matter for JCLs) to minimize the influence of anatomical location and intrinsic differences between tissue types for further quantitative analysis (Figure 1). Specifically, the distance between the NA ROI and lesion ROI was constrained to two to five voxels. NA ROIs were carefully positioned to avoid the white matter-cortex boundary and outer cortical surface adjacent to the meninges. In addition, to minimize potential gradient effects related to surface-in susceptibility artifacts, NA ROIs were placed on the same imaging layer as the corresponding lesion ROIs. Since CLs were jointly identified by two raters via consensus on 3D DIR/MPRAGE, with QSM verification, formal inter-rater reliability testing was deemed unnecessary for lesion delineation. To validate NA tissue reliability, the second rater re-drew NA ROIs in 50% of randomly selected CLs, and intraclass correlation coefficients (ICCs) for susceptibility values were calculated.

On QSM, χ_para, and χ_dia maps, lesion signals were visually classified as hypointense, isointense, or hyperintense relative to NA tissues. Mean χ_para and χ_dia values for each ROI were extracted using the FSLSTATS command of the FMRIB Software Library (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/).²² To validate the visual classification of lesion susceptibility signals, an additional healthy control (HC) analysis was performed (see the Supplemental Materials). Following Müller et al.,²³ lesion-to-NA ratios were calculated as susceptibility indices. For clinical correlations, patient-wise weighted mean susceptibility values (absolute χ_para/χ_dia) were derived from all CLs and NA ROIs. Additional MRI-derived parameters, including WML volume (WMLV), the number of paramagnetic rim lesions (PRLs), and normalized brain tissue volumes, are detailed in the Supplemental Material.

Statistical analysis

Statistical analysis was performed using SPSS version 26.0 (IBM Corp., Armonk, NY, United States) and Python version 3.7.7 (Python Software Foundation, Wilmington, DE, USA). Data normality was tested with the Kolmogorov–Smirnov test. Differences in demographics, disease duration, annualized relapse rate (ARR), imaging features, EDSS, and neuropsychological scores between RRMS patients with and without JPRs were assessed using chi-square tests for categorical data and generalized linear models for continuous data, adjusted for age, sex, and disease-modifying therapy (DMT). Spearman correlation analyses were performed to evaluate the associations between absolute χ_para and χ_dia values in all CLs and clinical as well as imaging variables, including the number and normalized volume of all CLs, the number and volume of JPR-associated CLs, disease duration, ARR, imaging features, EDSS, and neuropsychological scores. These analyses were adjusted for age, sex, DMT, years of education, and absolute χ_para and χ_dia values in NA tissue using the Pingouin package (v0.5.4). In analyses involving the number and volume of JPR-associated CLs, the corresponding number and volume of all CLs were additionally included as covariates. For lesion-wise comparisons, mixed-effects models were used to assess χ_para and χ_dia susceptibility indices between LCLs and ICLs, with patient ID as a random effect. Then, JPR-associated CLs were separated from LCLs to assess their individual contribution. Mixed-effects models were further applied among non-JPR LCLs, ICLs, JPR-associated CLs, and QSM-hyperintense JCLs. For the post hoc pairwise comparisons among these four groups, p values were adjusted using the Bonferroni correction to control the family-wise error rate across all six possible pairwise tests. Normally distributed data were expressed as mean ± SD and non-normal data as median (IQR). p < 0.05 was considered statistically significant.

Results

Patient and lesion characteristics

Table 1 summarizes the demographic, clinical, and imaging characteristics of the overall study population, stratified by the presence or absence of JPRs, with between-group comparisons provided. Demographic characteristics of the HC cohort used for supplementary validation are summarized in the Supplemental Table S1. Among 67 RRMS patients, three were excluded for poor image quality and lack of clinical data, and a total of 64 RRMS patients were included for analysis. Our cohort was predominantly female (75.0%), with a median age of 29.0 years (IQR = 24.3–38.5), a median EDSS of 1.0 (IQR = 0.0–2.0), and a median disease duration of 33.5 months (IQR = 12.3–58.0).

Table 1.

Demographic characteristics, clinical characteristics, and imaging features of all patients.

	All RRMS patients (n = 64)	Without JPRs (n = 41)	With JPRs (n = 23)	Adjusted mean difference (95% CI)	p
Age (years), median (IQR)	29.0 (24.3–35.8)	28.0 (23.0–36.5)	30.0 (25.0–35.0)	-	0.889
Gender, female, n (%)	48 (75.0)	32 (78.0)	16 (69.6)	-	0.452
Disease duration, median (IQR)	33.5 (12.3–58.0)	32.0 (18.0–59.5)	35.0 (8.0–56.0)	-	0.839
Education (years), median (IQR)	15.0 (12.0–16.0)	15.0 (12.0–16.0)	15.0 (12.0–16.0)	-	0.698
EDSS score, median (IQR)	1.0 (0.0–2.0)	0.5 (0.0–2.0)	1.5 (0.0–2.5)	0.461 (−0.323 to 1.245)	0.631
ARR, median (IQR)	0.0 (0.0–1.0)	0.0 (0.0–0.5)	1 (0.0–1.0)	0.213 (0.130 to 0.297)	<0.001
DST score, median (IQR)	13.0 (12.0–15.0)	14.0 (12.0–15.0)	12.0 (12.0–14.0)	−1.489 (−2.504 to −0.473)	0.004
SDMT score, median (IQR)	48.5 (35.5–58.8)	51.0 (40.0–60.5)	45.0 (30.0–55.0)	−6.518 (−14.435 to 1.399)	0.107
Normalized WML volume, median (IQR)	0.005 (0.001–0.012)	0.002 (0.001–0.006)	0.009 (0.006–0.018)	0.007 (0.003 to 0.011)	0.001
Normalized all CL volume (% of TIV), median (IQR)	0.007 (0–0.033)	0.003 (0–0.009)	0.039 (0.014–0.054)	0.024 (0.011 to 0.038)	<0.001
No. of all CLs, median (IQR)	2.5 (0.0–12.8)	1.0 (0.0–5.0)	12.0 (3.0–20.0)	8.733 (3.220 to 13.808)	0.001
Normalized brain volume, median (IQR)	0.746 (0.726–0.808)	0.746 (0.729–0.812)	0.745 (0.714–0.800)	−0.010 (0.024 to 3.658)	0.580
Normalized white matter volume, median (IQR)	0.306 (0.282–0.325)	0.307 (0.283–0.331)	0.300 (0.276–0.319)	−0.008 (−0.026 to 0.009)	0.346
Normalized cortical GM volume, median (IQR)	0.314 (0.296–0.342)	0.314 (0.297–0.344)	0.312 (0.292–0.331)	0.000 (−0.017 to 0.017)	0.971
Normalized deep GM volume, median (IQR)	0.108 (0.101–0.116)	0.110 (0.104–0.117)	0.103 (0.097–0.114)	−0.004 (−0.009 to 0.002)	0.166
JPR-associated CL volume, median (IQR)	0.0 (0.0–75.0)	-	127.0 (64.0–340.0)	-	-
No. of JPR-associated CLs, median (IQR)	0.0 (0.0–1.0)	-	2.0 (1.0–5.0)	-	-
No. of PRLs, median (IQR)	1.0 (0.0–3.0)	0.0 (0.0–1.0)	0.0 (0.0–1.0)	3.592 (2.445 to 4.738)	<0.001
DMT, n (%)
No DMT	14 (21.9)	12 (29.3)	2 (8.7)		0.073
Orals	38 (59.4)	23 (56.1)	15 (65.2)		0.466
High effective DMTs	12 (18.7)	6 (14.6)	6 (26.1)		0.346

Data are presented as means ± SD or median (IQR). Categorical variables are presented as frequencies with percentages in parentheses. ARR: annual relapsing rate; CL: cortical lesion; DST: digit span test; DMT: disease-modifying therapy; EDSS: Expanded Disability Status Scale; GM: gray matter; JPRs: juxtacortical paramagnetic rims; PRLs: paramagnetic rim lesions; SDMT: Symbol Digit Modalities Test; WML: white matter lesion.

Of the 64 patients, 50 (78.1%) were receiving DMT. Forty-six patients (71.9%) had at least one CL, 23 (35.9%) had at least one JPR-associated CL, and 36 (56.3%) had at least one PRL.

In all generalized linear models, patients without JPRs served as the reference group, and the regression coefficients (β) represent the estimated differences calculated as patients with JPRs minus patients without JPRs. Compared with patients without JPRs, those with JPRs showed a significantly higher number of PRLs (β = 3.592, 95% CI [2.445, 4.738], p < 0.001), higher ARR (β = 0.213, 95% CI [0.130, 0.297], p < 0.001), and lower DST scores (β = −1.489, 95% CI [−2.504, −0.473], p = 0.004). However, no significant differences were observed in demographics, disease duration, EDSS, or SDMT scores between the two groups.

Lesion-wise analysis

After excluding lesions with severe artifacts (n = 12), a total of 499 CLs and 112 JCLs were included in the analysis. Among all CLs, 404 (81.0%) were classified as LCLs, whereas 95 (19.0%) were ICLs. JPR-associated CLs accounted for 15.2% (76/499) of all CLs, all of which were leukocortical. When analyzed separately, all JPR-associated CLs appeared hyperintense on χ_para maps. On χ_dia maps, 36 (47.4%) were hypointense, 36 (47.4%) were isointense, and 4 (5.2%) were hyperintense (Figure 2). After excluding JPR-associated CLs, 423 non-JPR CLs remained, of which 360 (85.1%) were hypointense on χ_para maps and 277 (65.5%) were hyperintense on χ_dia maps (Supplemental Figure S1). The validity of this visual classification of all CLs was further supported by the additional HC analysis (the Supplemental Materials).

NA-ROI reliability showed excellent consistency for both χ_para (ICC = 0.87) and χ_dia (ICC = 0.84) values, confirming reference stability. Compared with ICLs, LCLs exhibited higher χ_para (95% CI [0.135, 0.461], p < 0.001) and lower χ_dia (95% CI [−0.503, −0.099], p = 0.004) indices. When JPR-associated CLs and JCLs were analyzed together with non-JPR LCLs and ICLs, the four lesion types showed significant overall differences in both χ_para (F = 830.221, p < 0.001) and χ_dia (F = 933.955, p < 0.001) indices (Figure 3).

Post hoc pairwise tests revealed no difference between non-JPR LCLs and ICLs in either χ_para (95% CI [−0.174, 0.231], p = 1.000) or χ_dia (95% CI [−0.401, 0.107], p = 0.757) indices. JPR-associated CLs showed higher χ_para indices than JCLs (95% CI [0.174, 0.690], p < 0.001), whereas χ_dia indices did not differ (95% CI [−0.185, 0.463], p = 1.000). All other pairwise comparisons among lesion types remained significant (false discovery rate (FDR)-adjusted p < 0.001; Table 2).

Table 2.

Group comparisons in χ_para and χ_dia indices between four lesion subtypes.

	χ_para index		χ_dia index
	Mean difference (95% CI)	p	Mean difference (95% CI)	p
Non-JPR LCLs vs. ICLs	0.029 (−0.174 to 0.231)	1.000	−0.147 (−0.401 to 0.107)	0.757
Non-JPR LCLs vs. JPR-associated CLs	−1.543 (−1.772 to −1.315)	<0.001	0.866 (0.588 to 1.144)	<0.001
Non-JPR LCLs vs. JCLs	−1.112 (−1.302 to −0.921)	<0.001	1.005 (0.767 to 1.244)	<0.001
ICLs vs. JPR-associated CLs	−1.572 (−1.844 to −1.300)	<0.001	1.013 (0.677 to 1.349)	<0.001
ICLs vs. JCLs	−1.140 (−1.382 to −0.898)	<0.001	1.152 (0.848 to 1.456)	<0.001
JPR-associated CLs vs. JCLs	0.432 (0.174 to 0.690)	<0.001	0.139 (−0.185 to 0.463)	1.000

CLs: cortical lesions; ICLs: intracortical lesions; JCLs: juxtacortical lesions; JPRs: juxtacortical paramagnetic rims; LCLs: leukocortical lesions.

Patient-wise correlation analysis

A total of 46 RRMS patients with CLs were included for correlation analysis (Table 3). Longer disease duration correlated with lower absolute χ_para (r = −0.461, p = 0.001) and higher absolute χ_dia (r = 0.457, p = 0.001) values within all CLs. Absolute χ_para values of all CLs were negatively associated with DST scores (r = −0.292, p = 0.049) and positively with the number of PRLs (r = 0.383, p = 0.009). Disease duration was negatively correlated with the number (r = −0.323, p = 0.029) and volume (r = −0.350, p = 0.017) of JPR-associated CLs. Higher ARR was positively correlated with the number of JPR-associated CLs (r = 0.293, p = 0.048). Among volumetric MRI parameters, normalized WM volume was inversely correlated with the number (r = −0.396, p = 0.006) and normalized volume (r = −0.333, p = 0.024) of all CLs. In contrast, the number of PRLs was positively correlated with all CL and JPR measures (r = 0.434–0.541, all p ⩽ 0.004).

Table 3.

Correlations between CL susceptibility metrics, imaging features, and clinical characteristics.

	Absolute χ_para value in all CLs		Absolute χ_dia value in all CLs
	r	p	r	p
Disease duration	−0.461	0.001	0.457	0.001
ARR	0.146	0.331	0.096	0.527
EDSS scores	0.001	0.993	0.006	0.969
DST scores	−0.292	0.049	0.136	0.367
SDMT scores	−0.056	0.712	0.016	0.917
Normalized white matter volume	0.059	0.697	−0.050	0.739
Normalized cortical GM volume	−0.014	0.924	−0.017	0.913
Normalized deep GM volume	−0.119	0.429	0.147	0.330
Normalized WML volume	0.014	0.926	0.206	0.170
No. of PRLs	0.383	0.009	−0.189	0.209

	Normalized all CL volume		No. of all CLs		JPR-associated CL volume		No. of JPR-associated CLs
	r	p	r	p	r	p	r	p
Disease duration	0.181	0.230	0.303	0.041	−0.350	0.017	−0.323	0.029
ARR	0.252	0.091	0.086	0.569	0.215	0.150	0.293	0.048
EDSS scores	−0.029	0.848	−0.005	0.976	0.149	0.322	0.084	0.581
DST scores	−0.322	0.029	−0.350	0.017	−0.159	0.291	−0.163	0.279
SDMT scores	−0.146	0.334	−0.212	0.156	−0.065	0.670	−0.055	0.716
Normalized white matter volume	−0.333	0.024	−0.396	0.006	0.342	0.020	0.252	0.092
Normalized cortical GM volume	−0.236	0.115	−0.158	0.294	0.250	0.094	0.141	0.351
Normalized deep GM volume	−0.286	0.054	−0.290	0.051	0.088	0.560	0.005	0.976
Normalized WML volume	0.617	<0.001	0.610	<0.001	−0.056	0.714	0.081	0.595
No. of PRLs	0.541	<0.001	0.434	0.003	0.417	0.004	0.492	<0.001

ARR: annual relapsing rate; CLs: cortical lesions; DST: digit span test; DMT: disease-modifying therapy; EDSS: Expanded Disability Status Scale; GM: gray matter; JPRs: juxtacortical paramagnetic rims; PRLs: paramagnetic rim lesions; SDMT: symbol digit modalities test; WML: white matter lesion.

Discussion

In this study, we applied subvoxel QSM to characterize paramagnetic and diamagnetic susceptibility patterns in CLs, with a specific focus on the imaging features and clinical relevance of JPRs. Our results highlight JPRs as a distinct susceptibility feature characterizing a subset of CLs with unique susceptibility patterns and clinical associations in MS.

Across all CLs, susceptibility patterns were heterogeneous. When JPR-associated CLs were excluded, most CLs demonstrated reduced paramagnetic susceptibility and increased diamagnetic susceptibility relative to the NA cortex. Reduced paramagnetic susceptibility in non-JPR CLs is, to some extent, consistent with previous reports.¹¹ Given the relatively low inflammatory burden in CLs, it may reflect loss of ferritin-rich oligodendrocytes and myelin-associated iron during demyelination.^24,25 Diamagnetic signals are typically associated with myelin in white matter,^23,26 whereas their sources in the cortex may differ. Given the inherently lower myelin density in cortical tissue and the well-established histological evidence of myelin loss in CLs,^27,28 increased χ_dia is unlikely to reflect remyelination. Instead, χ_dia variations likely reflect model-dependent susceptibility redistribution and microstructural complexity in subvoxel QSM in thin cortical regions. Within this model-dependent framework, membrane- and lipid-rich components associated with tissue remodeling may contribute to χ_dia variations;^8,29 however, this remains a hypothesis requiring further validation. Non-JPR CLs, therefore, exhibit mixed susceptibility patterns, with most lesions showing reduced paramagnetic susceptibility consistent with iron depletion, accompanied by relatively elevated but model-dependent diamagnetic susceptibility features.

In contrast, JPR-associated CLs exhibited a distinct susceptibility pattern characterized by elevated χ_para and decreased or comparable χ_dia measurements relative to other CL subtypes. Despite the shared hyperintense appearance on QSM, JPR-associated CLs also demonstrated higher χ_para indices than JCLs. However, interpretation of these findings warrants caution. Although ROIs were carefully restricted to the cortical component, residual partial volume effects from adjacent juxtacortical white matter cannot be completely excluded, particularly given the pronounced hyperintensity of JPRs on susceptibility maps. Furthermore, histopathological studies have demonstrated that CLs generally exhibit substantially less inflammatory activity than WMLs, and inflammatory changes in LCLs are often more prominent in the juxtacortical white matter component.^6,8 Therefore, it remains challenging to determine whether the elevated χ_para measurements observed in JPR-associated CLs originate from the cortical tissue itself. Nevertheless, the consistent susceptibility pattern observed in JPR-associated CLs suggests that JPRs represent a characteristic MRI susceptibility signature at the white matter-cortex interface, while the precise biological substrate underlying this imaging feature remains to be clarified.

Beyond their imaging appearance, JPR-associated CLs showed clear clinical relevance. The prevalence of JPRs in our cohort (35.9%) was substantially higher than previously reported (10%),¹³ and patients with JPRs exhibited higher PRL burden, higher ARR, and worse cognitive performance. The observed association between JPRs and PRLs may be facilitated by population-specific factors, as PRLs have been reported to occur more frequently in Chinese MS populations,³⁰ and JPRs were also more prevalent in our cohort, increasing the sensitivity to detect their relationship. Given that PRLs are considered markers of chronic active inflammation,³¹ the association between JPRs and PRLs may reflect a shared chronic inflammatory process involving both cortex and white matter pathology,⁶ which could account for the higher relapse activity and cognitive impairment observed in patients with JPRs.³²

In patient-wise correlation analysis, a positive correlation was observed between the absolute χ_para values in all CLs and PRL burden, which may be partly driven by the contribution of JPR-associated CLs exhibiting higher χ_para measurements and being associated with greater PRL burden. Accordingly, the effect of JPR-associated CLs should be considered when interpreting subsequent χ_para findings involving all CLs. Moreover, disease duration showed a positive correlation with absolute χ_dia values and a negative correlation with absolute χ_para values in CLs, indicating that iron depletion and the model-dependent diamagnetic susceptibility features evolve with disease progression. Furthermore, after adjustment for total CL burden, JPR-associated CL burden was inversely associated with disease duration, suggesting that JPRs may be relatively more frequent in patients with shorter disease duration. However, whether JPRs gradually diminish or disappear over the disease course remains to be determined in longitudinal studies.

Methodologically, our study extends prior susceptibility source separation work by analyzing individual CL subtypes rather than pooling all lesions. This lesion-level approach revealed marked susceptibility differences among CL subtypes and identified JPRs as a major contributor to the observed heterogeneity. Despite using a 3T MRI, the combination of high-resolution structural imaging and susceptibility indices enabled robust characterization of lesion-specific magnetic patterns.

This study has several limitations. First, the biological correlates of χ_dia in CLs remain uncertain and require MRI-histopathological validation. Second, applying subvoxel QSM to the cortex may be more challenging than in white matter because of the thin cortical ribbon and potential partial volume effects, particularly in JPR-associated CLs with marked susceptibility hyperintensity. Nevertheless, the consistent susceptibility patterns and clinical and radiological associations observed in JPRs support their potential relevance as a distinct feature in MS. Third, the cross-sectional design limited assessment of the temporal evolution of JPRs and did not allow determination of whether CL heterogeneity reflects a time-dependent process. Finally, gadolinium-enhancing lesions were not included, preventing evaluation of associations between acute inflammatory activity and cortical susceptibility abnormalities.

In summary, subvoxel QSM revealed distinct susceptibility patterns among CLs in MS and identified JPR-associated CLs as a characteristic lesion subtype with a distinct MRI susceptibility pattern involving the white matter-cortex interface. JPRs were associated with greater clinical and radiological burden, warranting longitudinal studies to clarify their temporal evolution and potential relevance to disease progression in MS.

Supplemental Material

sj-docx-1-msj-10.1177_13524585261471330 – Supplemental material for Subvoxel quantitative susceptibility mapping reveals distinct susceptibility patterns and clinical associations of juxtacortical paramagnetic rims in multiple sclerosis

Supplemental material, sj-docx-1-msj-10.1177_13524585261471330 for Subvoxel quantitative susceptibility mapping reveals distinct susceptibility patterns and clinical associations of juxtacortical paramagnetic rims in multiple sclerosis by Qiyuan Zhu, Zhuowei Shi, Ermelinda De Meo, Frederik Barkhof, Huajiao Wang, Lisha Nie, Jinzhou Feng, Feiyue Yin, Xiaoya Chen, Hongjiang Wei, Zichun Yan and Yongmei Li in Multiple Sclerosis Journal

Footnotes

Acknowledgements

All authors would like to thank all the subjects who participated in this study. The first author, Qiyuan Zhu, would like to thank Professor Declan Chard (NMR Research Unit, Queen Square MS Centre, Department of Neuroinflammation, UCL Queen Square Institute of Neurology, UCL) for the valuable and insightful comments regarding the interpretation of the results.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the China Postdoctoral Science Foundation (Certificate Number: 2024MD754034), the National Natural Science Foundation of China (82302150), Chongqing Postdoctoral Science Foundation (2023CQBSHTB3104), Natural Science Foundation of Chongqing Municipality (CSTB2022NSCQ-MSX0973, CSTB2024NSCQ-MSX0052), Chongqing Medical Scientific Research Project (Joint Project of Chongqing Health Commission and Science and Technology Bureau, 2023ZDXM006), and China Scholarship Council (No. 202409240047).

ORCID iDs

Qiyuan Zhu

Ermelinda De Meo

Frederik Barkhof

Huajiao Wang

Xiaoya Chen

Yongmei Li

Data availability statement

All data included in this study are available upon request by contact with the corresponding author.

Supplemental material

Supplemental material for this article is available online.

References

Kuhlmann

Moccia

Coetzee

, et al. Multiple sclerosis progression: Time for a new mechanism-driven framework. Lancet Neurol 2023; 22(1): 78–88.

Kiljan

Preziosa

Jonkman

, et al. Cortical axonal loss is associated with both gray matter demyelination and white matter tract pathology in progressive multiple sclerosis: Evidence from a combined MRI-histopathology study. Mult Scler 2021; 27(3): 380–390.

Mainero

Miscioscia

Treaba

CA.

Cortical remyelination in multiple sclerosis: A target for disease monitoring and intervention. Brain 2024; 147: 1124–1126.

Harrison

Roy

, et al. Association of cortical lesion burden on 7-T magnetic resonance imaging with cognition and disability in multiple sclerosis. JAMA Neurol 2015; 72(9): 1004–1012.

Zivadinov

Schweser

Jakimovski

, et al. Decoding gray matter involvement in multiple sclerosis via imaging. Neuroimaging Clin N Am 2024; 34(3): 453–468.

Kuhlmann

Ludwin

Prat

, et al. An updated histological classification system for multiple sclerosis lesions. Acta Neuropathol 2017; 133(1): 13–24.

Albert

Antel

Brück

, et al. Extensive cortical remyelination in patients with chronic multiple sclerosis. Brain Pathol 2007; 17(2): 129–138.

Lassmann

Multiple sclerosis pathology. Cold Spring Harb Perspect Med 2018; 8: a028936.

Louapre

Lubetzki

Neurodegeneration in multiple sclerosis is a process separate from inflammation: Yes. Mult Scler 2015; 21(13): 1626–1628.

10.

Trapp

Vignos

Dudman

, et al. Cortical neuronal densities and cerebral white matter demyelination in multiple sclerosis: A retrospective study. Lancet Neurol 2018; 17(10): 870–884.

11.

Bian

Tranvinh

Tourdias

, et al. In vivo 7T MR quantitative susceptibility mapping reveals opposite susceptibility contrast between cortical and white matter lesions in multiple sclerosis. AJNR Am J Neuroradiol 2016; 37(10): 1808–1815.

12.

Castellaro

Magliozzi

Palombit

, et al. Heterogeneity of cortical lesion susceptibility mapping in multiple sclerosis. AJNR Am J Neuroradiol 2017; 38(6): 1087–1095.

13.

Galbusera

Bahn

Weigel

, et al. Characteristics, prevalence, and clinical relevance of juxtacortical paramagnetic rims in patients with multiple sclerosis. Neurology 2024; 102: e207966.

14.

Wang

, et al. A method for estimating and removing streaking artifacts in quantitative susceptibility mapping. Neuroimage 2015; 108: 111–122.

15.

Langkammer

Schweser

Krebs

, et al. Quantitative susceptibility mapping (QSM) as a means to measure brain iron? A post mortem validation study. Neuroimage 2012; 62(3): 1593–1599.

16.

Shin

Lee

Yun

, et al. χ-separation: Magnetic susceptibility source separation toward iron and myelin mapping in the brain. Neuroimage 2021; 240: 118371.

17.

Feng

Liu

, et al. APART-QSM: An improved sub-voxel quantitative susceptibility mapping for susceptibility source separation using an iterative data fitting method. Neuroimage 2023; 274: 120148.

18.

Dimov

Nguyen

Gillen

, et al. Susceptibility source separation from gradient echo data using magnitude decay modeling. J Neuroimaging 2022; 32(5): 852–859.

19.

Thompson

Banwell

Barkhof

, et al. Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria. Lancet Neurol 2018; 17: 162–173.

20.

Avants

Yushkevich

Pluta

, et al. The optimal template effect in hippocampus studies of diseased populations. Neuroimage 2010; 49: 2457–2466.

21.

Yushkevich

Piven

Hazlett

, et al. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. Neuroimage 2006; 31: 1116–1128.

22.

Jenkinson

Beckmann

Behrens

, et al. FSL. Neuroimage 2012; 62: 782–790.

23.

Müller

Cagol

, et al. Quantifying remyelination using χ-separation in white matter and cortical multiple sclerosis lesions. Neurology 2024; 103: e209604.

24.

Fukunaga

Van Gelderen

, et al. Layer-specific variation of iron content in cerebral cortex as a source of MRI contrast. Proc Natl Acad Sci U S A 2010; 107: 3834–3839.

25.

Yao

Hametner

Van Gelderen

, et al. 7 Tesla magnetic resonance imaging to detect cortical pathology in multiple sclerosis. PLoS ONE 2014; 9(10): e108863.

26.

Kim

Shin

Lee

, et al. χ-separation imaging for diagnosis of multiple sclerosis versus neuromyelitis optica spectrum disorder. Radiology 2023; 307(1): e220941.

27.

Glasser

Goyal

Preuss

, et al. Trends and properties of human cerebral cortex: Correlations with cortical myelin content. Neuroimage 2014; 93(Pt. 2): 165–175.

28.

Jonkman

Fleysher

Steenwijk

, et al. Ultra-high field MTR and qR2* differentiates subpial cortical lesions from normal-appearing gray matter in multiple sclerosis. Mult Scler 2016; 22(10): 1306–1314.

29.

Popescu

Pirko

Lucchinetti

CF.

Pathology of multiple sclerosis: Where do we stand. Continuum 2013; 19(4): 901–921.

30.

Zhang

Gao

, et al. Brain lesion characteristics in Chinese multiple sclerosis patients: A 7-T MRI cohort study. Ann Clin Transl Neurol 2025; 12(2): 300–310.

31.

Absinta

Sati

Schindler

, et al. Persistent 7-tesla phase rim predicts poor outcome in new multiple sclerosis patient lesions. J Clin Invest 2016; 126: 2597–2609.

32.

Dal-Bianco

Grabner

Kronnerwetter

, et al. Long-term evolution of multiple sclerosis iron rim lesions in 7 T MRI. Brain 2021; 144: 833–847.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

2.17 MB