Diagnostic performance of central vein sign,paramagnetic rim lesion and cortical lesion in Asian patients with multiple sclerosis

Abstract

Background:

The 2024 revised McDonald criteria for multiple sclerosis (MS) incorporate central vein sign (CVS) and paramagnetic rim lesion (PRL), but their utility in Asian patients remains unclear.

Objective:

To evaluate the performance of CVS, PRL and cortical lesion (CL) in distinguishing MS from neuromyelitis optica spectrum disorders (NMOSD) and myelin-oligodendrocyte glycoprotein antibody-associated disease (MOGAD) in Japanese patients.

Methods:

We conducted a cross-sectional study of 139 MS, 41 NMOSD and 12 MOGAD patients who underwent 3-Tesla magnetic resonance imaging (MRI), including susceptibility-weighted imaging.

Results:

At least one CVS was identified in 95% of MS, 59% of NMOSD and 42% of MOGAD, while PRL occurred in 63% of MS only. The CVS number (cutoff ⩾ 2) yielded 85.6% sensitivity and 75.5% specificity (area under the curve (AUC) 0.856), which improved in patients <50 (AUC 0.920) or <70 years (AUC 0.907), and when combined with PRL and CL (AUC 0.926). The Select 6 algorithm achieved 69.1% sensitivity and 98.1% specificity (AUC 0.836), which were comparable in patients <50 (AUC 0.879) or <70 years (AUC 0.839), but improved when combined with PRL and CL (AUC 0.925).

Conclusions:

Combining Select 6 or CVS number with PRL and CL, particularly in patients <50 or <70 years, provides robust diagnostic performance for MS in the Asian population.

Keywords

Multiple sclerosis central vein sign paramagnetic rim lesion cortical lesion Asian diagnosis

Introduction

Multiple sclerosis (MS) is a chronic autoimmune disorder of the central nervous system and one of the major causes of physical and mental disabilities in young adults worldwide. The diagnosis of MS can be challenging since MS can have a heterogeneous clinical presentation, which can be mimicked by other diseases.¹ Over the past decade, advances in magnetic resonance imaging (MRI) techniques have enabled the detection of disease-specific biomarkers, such as central vein sign (CVS), paramagnetic rim lesion (PRL) and cortical lesion (CL).^2
–4 Although each reflecting distinct aspects of pathophysiology, CVS and PRL are specific to MS, and thus are incorporated into the 2024 revised McDonald diagnostic criteria for MS.^5,6

The CVS, indicative of perivenular lesion formation, has been reported as a promising imaging biomarker for distinguishing MS from non-MS conditions. However, various cutoff criteria based on the number or frequency of CVS-positive lesions have been proposed in the past.^7,8 Recent studies revealed that a simplified Select 6 algorithm yields high sensitivity and specificity.^9,10 The PRL denotes iron accumulation at lesion edges, suggesting smouldering, chronic active inflammation,^11,12 and is rarely seen in other inflammatory or infectious diseases, with the exception of Susac syndrome.¹³ The CL is also well-known to be MS-specific and associated with progressive disease of MS.¹⁴ Furthermore, several recent papers showed the improved diagnostic performance of CVS by combining with other biomarkers such as PRL, CL or oligoclonal IgG bands (OCB).^9,15

We have previously reported that the frequency of CL is relatively low in Japanese patients with MS, partially due to genetic predisposition, suggesting a less efficient role of CL in the diagnosis of MS than that in other races.¹⁶ In contrast, the diagnostic values of CVS and PRL remain unclear, as there is a limited number of papers that have investigated CVS and PRL in the Asian population. Therefore, we sought to investigate CVS, PRL and CL in our cohort, and to evaluate the diagnostic performance of the Select 6 algorithm and these lesion numbers as single or combined biomarkers in clinical practice. In addition, we aimed to assess the age-dependent performance of CVS, including a simple CVS number in age-restricted subgroups of the younger population.

Method

Participants

We conducted a retrospective cross-sectional study and enrolled 192 patients with MS, neuromyelitis optica spectrum disorders (NMOSD), and myelin-oligodendrocyte glycoprotein antibody-associated disease (MOGAD) who underwent brain MRI between June 2023 and July 2025 and who had been relapse-free at least 3 months prior to the scans (Table 1). At the time of MRI acquisition, all patients were diagnosed according to the established criteria,^17
–19 and their clinical information was retrospectively obtained. This study was approved by the Ethical Committee of Kyushu University and conducted according to the World Medical Association Declaration of Helsinki. Written informed consent was obtained from all participants.

Table 1.

Patient demographics.

	MS(n = 139)	NMOSD(n = 41)	MOGAD(n = 12)	p-values
Age	44.0 (36.5–50.0)	57.0 (49.0–65.0)	41.5 (30.2–52.2)	<0.001
Gender, female	103 (74%)	25 (92%)	4 (44%)	0.002
Disease duration	11.2 (5.2–18.6)	9.4 (4.8–17.2)	7.4 (2.3–12.9)	NS
EDSS score	2.0 (1.0–3.5)	3.5 (2.0–6.0)	0.0 (0.0–1.4)	<0.001
Annualised relapse rate	0.34 (0.13–0.57)	0.42 (0.26–0.77)	0.34 (0.19–0.86)	NS
Progressive disease	24 (17%)	0 (0%)	0 (0%)	0.002
Positive serum AQP4-IgG	0 (0%)	38 (93%)	0 (0%)	<0.001
Positive serum MOG-IgG	0 (0%)	0 (0%)	12 (100%)	<0.001
Positive oligoclonal IgG bands	88/118 (75%)	7/31 (23%)	4/8 (50%)	<0.001

Data are presented as number (%) or median (interquartile range). Annualised relapse rates were available in 131, 33 and 12 cases in patients with MS, NMOSD and NMOSD, respectively. Statistical analysis across the three groups was performed by Kruskal–Wallis for continuous variables and Fisher’s exact test for categorical variables. Abbreviations: AQP4, aquaporin-4; EDSS, Expanded Disability Status Scale; IgG, immunoglobulin G; MOG, myelin-oligodendrocyte glycoprotein; MOGAD, myelin-oligodendrocyte glycoprotein antibody-associated disease; NMOSD, neuromyelitis optica spectrum disorders; NS, not significant.

MR imaging

MR images were acquired on 3-Tesla scanners (Ingenia, Ingenia Elition X, or MR 7700, Philips Medical Systems, Best, the Netherlands) using standardised protocols including 3D-fluid-attenuated inversion recovery (FLAIR), 3D-double inversion recovery (DIR), and susceptibility-weighted imaging (SWI) (Supplemental Table 1). MR data were acquired as Digital Imaging and Communications in Medicine files, and SWI-FLAIR fusion images were generated on an imaging workstation, SYNAPSE VINCENT software (Fujifilm Medical Systems, Tokyo, Japan). The 3D-FLAIR dataset was rigidly co-registered with the SWI images and resampled into the SWI space. The co-registered FLAIR and SWI images were then fused using a linear intensity-blending algorithm, so that FLAIR hyperintensity was overlaid on venous structures depicted on SWI images.

Lesion identification

The numbers of CVS, PRL and CL, and the fulfilment of Select 6 algorithm were assessed by consensus of two experienced examiners, a neurologist (K.S.) and a neuroradiologist (O.T.), both of whom were board-certified with more than 15 years of experience, blinded to clinical information. CVS was identified on SWI-FLAIR fusion images according to the consensus statement of the North American Imaging in Multiple Sclerosis (NAIMS) Cooperative.² A central vein was defined as a hypointense line or dots with a thin diameter (<2 mm), which run through the centre of lesions. Small lesions (<3 mm in diameter), lesions containing multiple veins, and poorly visible lesions were excluded. The fulfilment of Select 6 algorithm was defined as CVS-positive if there are ⩾6 morphologically characteristic lesions with central veins, or if there are <6 morphologically characteristic lesions and CVS-positive lesions outnumber CVS-negative lesions.^8,20 PRL was assessed on both SWI and filtered phase images and defined as lesions with a discrete hypointense rim encompassing at least two-thirds of the outer edge of lesions, excluding cortical and ependymal edges, in accordance with the international consensus statement.³ CL was identified as a hyperintense lesion relative to surrounding normal-appearing grey matter on 3D-DIR images, and was verified on 3D-FLAIR images to exclude artefacts, according to the consensus recommendation⁴ and as previously reported.^16,21 The representative images of CVS, PRL and CL are shown in Figure 1(a) to (c).

Statistical analysis

Group comparisons among MS, NMOSD and MOGAD were conducted using non-parametric and contingency-table approaches. Categorical variables were analysed using the Chi-square test or Fisher’s exact test. Continuous variables across three groups were analysed using the Kruskal–Wallis test. When multiple comparisons were significant, a pairwise Mann–Whitney U test was performed with p-values adjusted using Benjamini-Hochberg false discovery rate (FDR) procedure as needed. The Mann–Whitney U test was used to analyse continuous variables in two groups. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance of single or combined biomarker models. Optimal cutoff values were determined as thresholds maximising Youden’s J statistic (sensitivity + specificity − 1). Areas under the curve (AUCs) were calculated, and ROC curves were compared using DeLong tests. For sets of multiple DeLong comparisons, p-values were corrected by the Benjamini-Hochberg FDR procedure. Statistical analysis was performed using R version 4.5.0. A p-value < 0.05 was considered statistically significant.

Results

Patient demographics

The participants consisted of patients with MS (n = 139), NMOSD (n = 41) and MOGAD (n = 12). As shown in Table 1, patients with NMOSD were relatively older, predominantly female, and had higher EDSS scores compared with other groups. No patients with MS were positive for serum anti-aquaporin-4 antibodies and anti-MOG antibodies.

Frequency and number of CVS, PRL and CL

At least one CVS was identified in 132 (95%), 24 (59%) and 5 (42%) patients with MS, NMOSD and MOGAD, respectively (Table 2). Accordingly, the CVS numbers were significantly higher in MS [median (interquartile range, IQR), 5.0 (3.0–7.0)] than those in NMOSD [median (IQR), 1.0 (0–1.0)] and MOGAD [median (IQR), 0.0 (0–1.5)] (Figure 1(d)). CVS-positive NMOSD or MOGAD patients tended to be older individuals with relatively high qualitative FLAIR white matter lesion loads, though lesion volume was not quantified (Supplemental Figure 1A, B). The PRL was confirmed in 88 (63%) MS patients, but not in any cases of NMOSD and MOGAD (Figure 1(e)). The CL was found in 64 (46%) MS patients, while 2 patients (5%) with NMOSD and no MOGAD patients had a CL (Figure 1(f), Supplemental Figure 1C).

Table 2.

Frequencies of CVS, PRL and CLs in patients with MS, NMOSD and MOGAD.

	MS(n = 139)	NMOSD(n = 41)	MOGAD(n = 12)	p-values
Patients with ⩾1 CVS	132 (95.0%)	24 (58.5%)	5 (41.7%)	<0.0001
Patients with ⩾1 PRL	88 (63.3%)	0 (0%)	0 (0%)	<0.0001
Patients with ⩾1 CL	64 (46.0%)	2 (4.9%)	0 (0%)	<0.0001

Data are presented as number (%). Statistical analysis was performed by Fisher’s exact test.

Abbreviations: CL, cortical lesion; CVS, central vein sign; NMOSD, neuromyelitis optica spectrum disorders; NS, not significant; PRL, paramagnetic rim lesion.

Clinical characteristics of MS patients with CVS, PRL and CL

We next examined the association of CVS, PRL and CL with the pathophysiology of Asian MS by comparing lesion numbers between patients with and without progressive disease course and OCB, both of which have been reported to be less frequently observed in Asian patients with MS than in Caucasian patients.²² The participants with MS consisted of 22 progressive MS (PMS) and 117 relapsing MS (RMS). The CVS numbers were comparable between PMS and RMS [median (IQR), 5.5 (3.0–9.0) vs 5.0 (3.0–7.0), p = 0.545], meanwhile PRL and CL numbers were significantly higher in PMS compared with RMS [PRL number: median (IQR), 4.5 (2.0–6.0) vs 1.0 (0–3.0), p < 0.001; CL number: median (IQR), 2.5 (1.0–5.5) vs 0.0 (0.0–1.0), p < 0.001) (Supplemental Figure 2A–C). The positivity of OCB was 74.6% in total MS patients, which was higher than 53.2% in recently published data of the Japanese nationwide survey.²² OCB-positive patients had increased numbers of CVS, PRL and CL (Supplemental Figure 2D–F). We also examined the correlations of clinical demographics with lesion numbers in MS patients. The CVS, PRL and CL numbers showed a positive correlation with each other. The PRL and CL numbers were positively correlated with EDSS score (PRL: ρ = 0.25, p < 0.008; CL: ρ = 0.25, p < 0.001); however, CVS numbers were not (Supplemental Figure 3). We also evaluated the distribution of lesion number by age and disease duration. Although PRL is reported to be linked to younger age and early smouldering inflammation,¹² PRL was observed across the entire age and disease duration range. The linear associations of PRL with age or disease duration were not significant, while a generalised additive model suggested a mild peak in PRL burden in the middle-aged population (Supplemental Figure 4).

Diagnostic performance of CVS, PRL and CL as a single biomarker

We performed ROC curve analysis of CVS, PRL and CL counts, and of the Select 6 algorithm to evaluate the diagnostic performance of a single biomarker (Figure 2(a)). The CVS numbers exhibited the best performance with an AUC of 0.856, and the optimal cutoff estimated by Youden’s J statistic (⩾2 CVS) presented 85.6% sensitivity and 75.5% specificity (Table 3). A more stringent cutoff of ⩾3 CVS yielded 75.5% sensitivity and 79.2% specificity, indicating slightly higher specificity and lower sensitivity than that of ⩾2 CVS (Supplemental Table 2). The fulfilment of the Select 6 algorithm also distinguished MS and non-MS with AUC 0.836, sensitivity 69.1% and specificity 98.1%, which was not significantly different from CVS number (Table 3, Supplemental Table 3). The PRL numbers had an AUC of 0.817, 63.3% sensitivity and 100% specificity with a cutoff of ⩾1 PRL. Due to the low frequency of CL, which is consistent with our previous reports,¹⁶ the CL numbers showed relatively low performance, with AUC 0.716, sensitivity of 46.0% and specificity of 96.2% at a cutoff of ⩾1 CL. We next performed the same analysis in patients aged <50 or <70 years, considering the recommendation of additional tests in patients older than 50 with vascular comorbidities,⁶ and found that the diagnostic performance of CVS numbers improved in these subgroups. The CVS number with a cutoff ⩾2 CVS exhibited AUCs of 0.907, with corresponding sensitivities of 86.9% and 85.8%, and specificities of 88.9% and 84.1% in patients aged <50 and <70 years, respectively (Table 3, Figure 2(b) and (c)). In contrast, the diagnostic performance of Select 6, PRL and CL numbers did not remarkably improve in this subgroup analysis (Table 3, Figure 2).

Table 3.

Diagnostic performance of CVS, PRL and CL as a single biomarker.

MS vs NMOSD/MOGAD	AUC	95% CI	Cutoff	Sensitivity	Specificity
All patients
CVS number	0.856	0.797–0.915	2	85.6%	75.5%
CVS Select 6	0.836	0.793–0.879	NA	69.1%	98.1%
PRL number	0.817	0.776–0.857	1	63.3%	100.0%
CL number	0.716	0.669–0.763	1	46.0%	96.2%
Patients aged < 50 years
CVS number	0.920	0.863–0.977	2	86.9%	88.9%
CVS Select 6	0.879	0.836–0.921	NA	75.8%	100.0%
PRL number	0.803	0.755–0.851	1	60.6%	100.0%
CL number	0.717	0.668–0.766	1	43.4%	100.0%
Patients aged < 70 years
CVS number	0.907	0.864–0.950	2	85.8%	84.1%
CVS Select 6	0.839	0.795–0.884	NA	70.1%	97.7%
PRL number	0.813	0.772–0.855	1	62.7%	100.0%
CL number	0.714	0.665–0.764	1	46.3%	95.5%

The cutoff values were determined as thresholds maximising Youden's J statistic (sensitivity + specificity − 1). Abbreviations: AUC, area under the curve; CI, confidence interval; CL, cortical lesion; CVS, central vein sign; MS, multiple sclerosis; MOGAD, myelin-oligodendrocyte glycoprotein antibody-associated disease; NA, not available; NMOSD, neuromyelitis optica spectrum disorders; PRL, paramagnetic rim lesion.

Table 4.

Diagnostic performance of CVS in combination with PRL and CL.

MS vs NMOSD/MOGAD	AUC	95% CI	Sensitivity	Specificity
All patients
CVS number	0.856	0.797–0.915	85.6%	75.5%
CVS number + PRL	0.926	0.890–0.961	72.7%	98.1%
CVS number + CL	0.884	0.834–0.933	89.2%	71.7%
CVS number + PRL + CL	0.926	0.891–0.961	76.3%	98.1%
CVS Select 6	0.836	0.793–0.879	69.1%	98.1%
CVS Select 6 + PRL	0.914	0.882–0.946	83.5%	98.1%
CVS Select 6 + CL	0.887	0.850–0.925	79.9%	94.3%
CVS Select 6 + PRL + CL	0.925	0.894–0.956	83.5%	98.1%
Patients aged < 50 years
CVS number	0.920	0.863–0.977	86.9%	88.9%
CVS number + PRL	0.949	0.910–0.988	90.9%	88.9%
CVS number + CL	0.935	0.888–0.983	88.9%	88.9%
CVS number + PRL + CL	0.952	0.914–0.989	91.9%	88.9%
CVS Select 6	0.879	0.836–0.921	75.8%	100.0%
CVS Select 6 + PRL	0.924	0.889–0.960	84.8%	100.0%
CVS Select 6 + CL	0.919	0.883–0.956	83.8%	100.0%
CVS Select 6 + PRL + CL	0.934	0.901–0.968	86.9%	100.0%
Patients aged < 70 years
CVS number	0.907	0.864–0.950	85.8%	84.1%
CVS number + PRL	0.945	0.915–0.976	80.6%	97.7%
CVS number + CL	0.922	0.885–0.960	81.3%	88.6%
CVS number + PRL + CL	0.946	0.916–0.977	81.3%	97.7%
CVS Select 6	0.839	0.795–0.884	70.1%	97.7%
CVS Select 6 + PRL	0.910	0.876–0.944	82.8%	97.7%
CVS Select 6 + CL	0.885	0.845–0.925	79.9%	93.2%
CVS Select 6 + PRL + CL	0.921	0.888–0.954	82.8%	97.7%

For combined biomarker models, positivity was defined on the basis of the predicted probability from the logistic regression model, using the optimal cutoff determined by Youden’s J statistic.

Abbreviations: AUC, area under the curve; CI, confidence interval; CL, cortical lesion; CVS, central vein sign; MS, multiple sclerosis; MOGAD, myelin-oligodendrocyte glycoprotein antibody-associated disease; NMOSD, neuromyelitis optica spectrum disorders; PRL, paramagnetic rim lesion.

Diagnostic performance of CVS in combination with PRL and CL

We next investigated whether the diagnostic performance of the CVS number or the Select 6 algorithm, both of which exhibited the best performance as a single biomarker, improved when co-evaluated with PRL and CL numbers. Although our sample size and dichotomised variables do not afford the same level of clinical validation as in Borrelli et al.,¹⁵ we applied the same regression-plus-ROC workflow to confirm model behaviour and to compare the diagnostic performance in our dataset (Figure 3, Table 4, Supplemental Table 3).

In the total patient cohort, when CVS numbers were combined with PRL numbers, the AUC rose to 0.926 (sensitivity 72.7%, specificity 98.1%), and when combined with CL numbers, the AUC was 0.884 (sensitivity 89.2%, specificity 71.7%) (Figure 3(a)). A combination of all three markers yielded AUC 0.926 (sensitivity 76.3%, specificity 93.1%). Likewise, the diagnostic performance of Select 6 algorithm increased up to AUC 0.914 (sensitivity 83.5%, specificity 98.1%) when PRL number was added, to AUC 0.887 (sensitivity 79.9%, specificity 94.3%) with CL, and to AUC 0.925 (sensitivity 83.5%, specificity 98.1%) when all three were included (Figure 3(b)).

The same ROC curve analysis was applied in a subgroup of patients aged <50 and <70 years. In patients aged <50 years, pairwise comparisons of AUCs using the DeLong test were not feasible due to quasi-perfect discrimination with a limited sample size. However, the AUCs of CVS number appeared to be further improved to AUC 0.949 (sensitivity 90.9%, specificity 88.9%) with PRL, AUC 0.935 (sensitivity 88.9%, specificity 88.9%) with CL, AUC 0.952 (sensitivity 91.9%, specificity 88.9%) with PRL and CL (Figure 3(c)). Likewise, the Select 6 algorithm exhibited an AUC of 0.924 (sensitivity 84.8%, specificity 100.0%) with PRL, an AUC of 0.919 with CL (sensitivity 83.8%, specificity 100.0%), and an AUC of 0.934 (sensitivity 86.9%, specificity 100.0%) with PRL and CL (Figure 3(d)). In patients aged <70 years, the performance of CVS number was meaningfully improved when we added PRL (AUC 0.945, sensitivity 80.6%, specificity 97.7%), and PRL together with CL (AUC 0.946, sensitivity 81.3%, specificity 97.7%), but not with CL (AUC 0.922, sensitivity 81.3%, specificity 88.6%) (Figure 3(e)). The performance of Select 6 algorithm improved to AUC 0.910 (sensitivity 82.8%, specificity 97.7%) with PRL, to AUC 0.885 (sensitivity 79.9%, specificity 93.2%) with CL, and AUC 0.929 (sensitivity 82.8%, specificity 97.7%) with PRL and CL (Figure 3(f)).

Discussion

In this cross-sectional study, the diagnostic performance of CVS, PRL and CL was investigated using clinically feasible 3-Tesla MRI. At least one CVS was identified in 95% of MS patients, and could be detected in NMOSD and MOGAD as well, especially in relatively old subjects with high brain lesion load. The PRL and CL were less frequently observed in MS than CVS; however, they were almost exclusive to MS. We found that the fulfilment of Select 6 algorithm and ‘⩾2 CVS in patients aged <50 or <70 years’ yielded robust diagnostic performance, and that the Select 6 algorithm enhanced its performance in combination with PRL and CL.

The CVS has been repeatedly reported to be useful in distinguishing MS from MS mimics; however, various cutoff criteria have been proposed.² To ensure applicability in routine clinical practice, we simply counted the number of lesions and assessed the fulfilment of the Select 6 algorithm, rather than using an efficacious, but complicated and time-consuming method in which examiners evaluated the frequency of CVS-positive lesions.^7,8 The ROC curve analyses revealed that the Select 6 algorithm alone exhibited near-perfect specificity in distinguishing MS from NMOSD and MOGAD, but integrating PRL and CL into Select 6 further improved its sensitivity, as recently reported.¹⁵ We also considered incorporating OCB as a combined biomarker; however, OCB did not improve diagnostic accuracy, likely due to the relatively low frequency of OCB in our patients with MS (data not shown).

Although several papers have shown that the Select 3 or ‘three or more CVS’ can efficiently distinguish MS from non-MS,^8,23,24 we found that CVS numbers exhibited a robust performance, especially in subgroups of younger population, and that the optimal cutoff values were consistently ‘two or more CVS’ in our cohort, considering the Youden’s J statistic and trade-offs of sensitivities and specificities. This simple algorithm can be briefly checked and is useful for distinguishing MS from NMOSD and MOGAD, especially in patients aged <50 or <70 years who have a relatively low risk of confounding cerebral small vessel disease. However, it should be noted that the specificity is not as high as the Select 6 algorithm.

Regarding the ethnic aspects, there was limited number of papers that investigated CVS or PRL in Asian population,^25
–27 and no previous paper has evaluated the diagnostic performance of CVS in combination with PRL and CL in Asian patients with MS. According to our data, CVS and PRL were identified as commonly as those in the previous reports which investigated using 3-Tesla scanners^25,28
–30; meanwhile, CL was less frequently detected than that in the literature³¹ as our group has previously described.^16,21 This finding is compatible with a recently published Chinese 7-Tesla MRI study²⁷ describing that 87% of patients were positive for at least one PRL and that only 58% of patients exhibited CL, despite a systematic review reporting that more than 90% of patients, primarily Caucasian, were positive for CL on 7-Tesla MRI.³² The relatively low frequency of CL might be, at least partly, explained by lower disabilities, given the modest median EDSS scores in this cohort. The milder disease burden in Asians compared with Caucasians could be due to the distinct genetic background, as previously suggested.^16,33 However, the frequency of PRL was comparable with the literature, suggesting that PRL can reflect MS-specific pathology, which is present even in mild disease, more efficiently than CL, which is also highly specific to MS.

This study has several limitations. First, this study is performed retrospectively and with a modest sample size of non-MS cases, which may limit generalisability. It would be necessary to include other non-MS diseases, such as ischemic cerebrovascular disease, systemic lupus erythematosus, Sjögren syndrome and neuro-Behçet disease, for differential diagnosis in clinical practice. Second, there is room for improvement in magnetic field strength and sequences. In general, lesion detection on 3-Tesla MRI (especially for CVS and CL) is inferior to that on ultrahigh-field 7-Tesla MRI,^32,34 and numerous susceptibility-based imaging methods are available for detecting CVS.^2,35 We applied fusion images of SWI and FLAIR images for the CVS identification. Third, the percentages of CVS-positive lesions were not evaluated, which have been reported to be efficacious in distinguishing MS from non-MS.^7,8 Fourth, this study did not include an independent external validation cohort. Our findings and their reproducibility should be re-evaluated using other sequences in the future. Prospective, multicenter validation – ideally incorporating cognitive and longitudinal outcomes – will be essential to confirm clinical utility and refine cutoff values across age groups. Fifth, brain and lesion volumes were not quantified. This limited the statistical evaluation of the association between PRL counts and EDSS scores.

In conclusion, CVS and PRL are commonly observed on 3-Tesla MRI scanners in Asian patients with MS. The fulfilment of the CVS Select 6 algorithm, which is now incorporated into the 2024 revised McDonald diagnostic criteria,^5,6 demonstrates a robust diagnostic performance in distinguishing MS from NMOSD and MOGAD, particularly when combined with PRL and CL. Among patients aged <50 or <70 years, two or more CVS could be strongly suggestive of MS in the Asian population.

Supplemental Material

sj-docx-1-msj-10.1177_13524585261420650 – Supplemental material for Diagnostic performance of central vein sign, paramagnetic rim lesion and cortical lesion in Asian patients with multiple sclerosis

Supplemental material, sj-docx-1-msj-10.1177_13524585261420650 for Diagnostic performance of central vein sign, paramagnetic rim lesion and cortical lesion in Asian patients with multiple sclerosis by Koji Shinoda, Ayano Matsuyoshi, Hajime Takeuchi, Mitsuru Watanabe, Katsuhisa Masaki, Eikichi Igeta, Hironori Kamano, Koji Yamashita, Osamu Togao and Noriko Isobe in Multiple Sclerosis Journal

Footnotes

Acknowledgements

The authors are grateful to Professor Kousei Ishigami (Department of Clinical Radiology, Kyushu University), Dr. Masaya Harada, Dr. Kazunori Iwao, Dr. Yuu-ichi Kira, Dr. Keisuke Mizutani and Dr. Yukiko Inamori (Department of Neurology, Kyushu University) for their generous support of the present study.

Author Contribution

K.S. and N.I. designed and conceptualised the study. K.S., E.I., H.K., K.Y., and T.O. contributed to the development of MRI sequences. K.S. and O.T. evaluated MR images. A.M. and H.T. assisted with the data acquisition. M.W. and. K.M. assisted with data interpretation. K.S. and N.I. wrote the paper. All the authors reviewed, amended, and agreed on the final version of the manuscript.

Declaration of Conflicting Interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: K.S. received honoraria from Alexion, Biogen, Chugai, Mitsubishi Tanabe, Novartis, and UCB. A.M, H.T., E.I., H.K., K.Y. and O.T. have nothing to disclose. M.W. received honoraria from Alexion, Argenx, Biogen, Chugai, Mitsubishi Tanabe, Novartis, Takeda, UCB, and Viatris. K.M. received honoraria from Alexion, Biogen, Chugai, Novartis, and Tanabe Mitsubishi. N.I. received honoraria from Alexion, Biogen, Chugai, Eisai, Mitsubishi Tanabe, Novartis, Takeda, and Teijin and research support from Chugai, Japan Blood Products Organisation, and Sumitomo.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported in part by JSPS KAKENHI (JP23K06964, JP24K02371, JP25K10629), the MHLW Research Program on Rare and Intractable Diseases (JPMH23FC1009), the Nakatomi Foundation, and the Takeda Science Foundation.

ORCID iDs

Koji Shinoda

Hajime Takeuchi

Mitsuru Watanabe

Katsuhisa Masaki

Koji Yamashita

Data Availability Statement

Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.

Supplemental Material

Supplemental material for this article is available online.

References

Solomon

Arrambide

Brownlee

, et al. Differential diagnosis of suspected multiple sclerosis: An updated consensus approach. Lancet Neurol 2023; 22(8): 750–768.

Sati

Constable

, et al. The central vein sign and its clinical evaluation for the diagnosis of multiple sclerosis: A consensus statement from the North American Imaging in Multiple Sclerosis Cooperative. Nat Rev Neurol 2016; 12(12): 714–722.

Bagnato

Sati

Hemond

, et al. Imaging chronic active lesions in multiple sclerosis: A consensus statement. Brain 2024; 147(9): 2913–2933

Geurts

JJG

Roosendaal

Calabrese

, et al. Consensus recommendations for MS cortical lesion scoring using double inversion recovery MRI. Neurology 2011; 76(5): 418–424.

Barkhof

Reich

, et al. 2024 MAGNIMS-CMSC-NAIMS consensus recommendations on the use of MRI for the diagnosis of multiple sclerosis. Lancet Neurol 2025; 24(10): 866–879.

Montalban

Lebrun-Frénay

, et al. Diagnosis of multiple sclerosis: 2024 revisions of the McDonald criteria. Lancet Neurol 2025; 24(10): 850–865.

Maggi

Absinta

Grammatico

, et al. Central vein sign differentiates multiple sclerosis from central nervous system inflammatory vasculopathies. Ann Neurol 2018; 83(2): 283–294.

Mistry

Abdel-Fahim

Samaraweera

, et al. Imaging central veins in brain lesions with 3-T T2*-weighted magnetic resonance imaging differentiates multiple sclerosis from microangiopathic brain lesions. Mult Scler 2016; 22(10): 1289–1296.

Toljan

Daboul

Raza

, et al. Diagnostic performance of central vein sign versus oligoclonal bands for multiple sclerosis. Mult Scler 2024; 30(10): 1268–1277.

10.

Cagol

Cortese

Barakovic

, et al. Diagnostic performance of cortical lesions and the central vein sign in multiple sclerosis. JAMA Neurol 2024; 81(2): 143–153.

11.

Dal-Bianco

Grabner

Kronnerwetter

, et al. Slow expansion of multiple sclerosis iron rim lesions: Pathology and 7 T magnetic resonance imaging. Acta Neuropathol 2017; 133(1): 25–42.

12.

Absinta

Sati

Masuzzo

, et al. Association of chronic active multiple sclerosis lesions with disability in vivo. JAMA Neurol 2019; 76(12): 1474–1483.

13.

Maggi

Sati

Nair

, et al. Paramagnetic rim lesions are specific to multiple sclerosis: An international multicenter 3T MRI study. Ann Neurol 2020; 88(5): 1034–1042.

14.

Calabrese

Magliozzi

Ciccarelli

, et al. Exploring the origins of grey matter damage in multiple sclerosis. Nat Rev Neurosci 2015; 16(3): 147–158.

15.

Borrelli

Martire

Stölting

, et al. Central vein sign, cortical lesions, and paramagnetic rim lesions for the diagnostic and prognostic workup of multiple sclerosis. Neurol Neuroimmunol Neuroinflamm 2024; 11(4): e200253.

16.

Shinoda

Matsushita

Nakamura

, et al. HLA-DRB1*04:05 allele is associated with intracortical lesions on three-dimensional double inversion recovery images in Japanese patients with multiple sclerosis. Mult Scler J 2018; 24(6): 710–720.

17.

Thompson

Banwell

Barkhof

, et al. Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria. Lancet Neurol 2018; 17(2): 162–173.

18.

Wingerchuk

Banwell

Bennett

, et al. International consensus diagnostic criteria for neuromyelitis optica spectrum disorders. Neurology 2015; 85(2): 177–189.

19.

Banwell

Bennett

Marignier

, et al. Diagnosis of myelin oligodendrocyte glycoprotein antibody-associated disease: International MOGAD Panel proposed criteria. Lancet Neurol 2023; 22(3): 268–282.

20.

Ontaneda

Sati

Raza

, et al. Central vein sign: A diagnostic biomarker in multiple sclerosis (CAVS-MS) study protocol for a prospective multicenter trial. Neuroimage Clin 2021; 32: 102834.

21.

Shinoda

Matsushita

Nakamura

, et al. Contribution of cortical lesions to cognitive impairment in Japanese patients with multiple sclerosis. Sci Rep 2020; 10(1): 5228

22.

Watanabe

Isobe

Niino

, et al. Prevalence of, and disability due to, multiple sclerosis and neuromyelitis optica spectrum disorder in Japan by the Fifth nationwide survey. Neurology 2024; 103(10): e209992

23.

Sinnecker

Clarke

Meier

, et al. Evaluation of the central vein sign as a diagnostic imaging biomarker in multiple sclerosis. JAMA Neurol 2019; 76(12): 1446–1456.

24.

Solomon

Watts

Ontaneda

, et al. Diagnostic performance of central vein sign for multiple sclerosis with a simplified three-lesion algorithm. Mult Scler 2018; 24(6): 750–757.

25.

Jang

Nam

Choi

, et al. Paramagnetic rims in multiple sclerosis and neuromyelitis optica spectrum disorder: A quantitative susceptibility mapping study with 3-T MRI. J Clin Neurol 2020; 16(4): 562–572.

26.

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.

27.

Liu

Wang

Wei

, et al. The characteristics and influencing factors of paramagnetic rim lesions in Chinese MS patients: A 7T MRI study. Mult Scler 2025; 31(8): 964–974.

28.

Clarke

Pareto

Pessini-Ferreira

, et al. Value of 3T susceptibility-weighted imaging in the diagnosis of multiple sclerosis. AJNR Am J Neuroradiol 2020; 41(6): 1001–1008.

29.

Maggi

Absinta

Sati

, et al. The ‘central vein sign’ in patients with diagnostic ‘red flags’ for multiple sclerosis: A prospective multicenter 3T study. Mult Scler J 2020; 26(4): 421–432.

30.

Meaton

Altokhis

Allen

, et al. Paramagnetic rims are a promising diagnostic imaging biomarker in multiple sclerosis. Mult Scler 2022; 28(14): 2212–2220

31.

Calabrese

Battaglini

Giorgio

, et al. Imaging distribution and frequency of cortical lesions in patients with multiple sclerosis. Neurology 2010; 75(14): 1234–1240.

32.

Madsen

MAJ

Wiggermann

Bramow

, et al. Imaging cortical multiple sclerosis lesions with ultra-high field MRI. Neuroimage Clin 2021; 32: 102847.

33.

Nakamura

Gaetano

Matsushita

, et al. A comparison of brain magnetic resonance imaging lesions in multiple sclerosis by race with reference to disability progression. J Neuroinflammation 2018; 15(1): 255

34.

Okromelidze

Patel

Singh

, et al. Central vein sign in multiple sclerosis: A comparison study of the diagnostic performance of 3T versus 7T MRI. AJNR Am J Neuroradiol 2023; 45(1): 76–81.

35.

Kerpel

Tal

Menascu

, et al. Acquisition of susceptibility-sensitive sequences for central vein sign detection: A systematic review of MRI protocols. Mult Scler Relat Disord 2025; 102: 106608.