The Diagnostic Value of Contrast‐Enhanced Vessel Wall MRI for Diagnosing Neuropsychiatric Systemic Lupus Erythematosus

ABSTRACT

Background

Imaging biomarkers for neuropsychiatric systemic lupus erythematosus (NPSLE) are highly needed, and intracranial contrast‐enhanced vessel wall imaging (CE‐VWI) can potentially detect cerebral vessel wall abnormalities in lupus.

Purpose

To evaluate the diagnostic value of CE‐VWI in differentiating NPSLE from non‐NPSLE.

Study Type

Cross‐sectional, retrospective.

Subjects

Forty‐seven patients with NPSLE (mean age, 44.3 years ± 13.2 standard deviation [SD], 40 females, 85%) and 52 patients without NPSLE (mean age, 43.0 years ± 16.5 SD, 49 females, 89%). The non‐NPSLE group consisted of SLE patients who had no NP symptoms or were diagnosed with comorbidities from other diseases.

Field Strength/Sequence

3‐T, three‐dimensional (3D) contrast‐enhanced vessel wall imaging (3D‐T1‐CUBE).

Assessment

Vessel wall lesions (VWLs) were visually assessed across 15 segments, from the internal carotid artery and basilar artery to A1–A2 for ACA, M1–M2 for MCA, and P1–P2 for PCA, for wall thickening and enhancement. Conventional MRI and MR angiography were also used to assess infarction, hemorrhage, atrophy, and arterial stenosis.

Statistical Tests

Paired comparisons using the chi‐square and unpaired t‐tests were followed by multivariate logistic regression analysis incorporating factors with significant group differences to identify associations with NPSLE. Receiver operating characteristic (ROC) analysis with the area under the curve (AUC) assessed the diagnostic performance of CE‐VWI. A p‐value < 0.05 was considered statistically significant.

Results

The NPSLE group showed a significantly higher number of contrast‐enhancing VWLs (CE‐VWLs; median [interquartile range]: 2 [0.5–4] vs. 0 [0–1]). Cerebral infarctions and arterial stenotic lesions were more common in NPSLE, occurring in 12 (26%) vs. 2 (3%) and 19 (40%) vs. 5 (9%) of patients, respectively. A multivariate logistic regression analysis identified CE‐VWLs as the sole significant factor associated with NPSLE (odds ratio, 1.97; 95% confidence interval, 1.23–3.16). The ROC analysis showed an AUC of 0.78 for CE‐VWLs, with a sensitivity of 60% and a specificity of 87%.

Data Conclusion

CE‐VWI may demonstrate high specificity and good diagnostic performance in differentiating NPSLE from non‐NPSLE.

Evidence Level

3.

Technical Efficacy

Stage 2.

Plain Language Summary.

• This study explored the use of contrast‐enhanced vessel wall MRI to detect brain artery involvement in patients with systemic lupus erythematosus (SLE), an autoimmune disease. Some patients develop neuropsychiatric symptoms, known as neuropsychiatric SLE (NPSLE), which are serious and difficult to diagnose with conventional MRI.

• The researchers examined 102 patients and found that vessel wall enhancement was more common in those with NPSLE, even when conventional MRI appeared normal.

• These findings suggest that this advanced MRI technique may help physicians more accurately detect NPSLE and support better diagnosis and treatment planning.

1. Introduction

Systemic lupus erythematosus (SLE) is a chronic autoimmune disease that mainly affects young to middle‐aged women, with symptoms including skin erythema, nephritis, and neurological symptoms [1]. Neuropsychiatric SLE (NPSLE) is a severe complication noted in nearly half of SLE patients, causing psychiatric symptoms, cerebral infarctions, and cognitive dysfunction that can impair daily life [2]. Specifically, NPSLE contributes to up to 19% of deaths among patients with SLE, highlighting the need for proper management [3]. Accurate diagnosis is crucial because therapeutic strategies differ between NPSLE and non‐NPSLE. Patients with NPSLE often require high‐dose corticosteroids and immunosuppressive therapies, while those without neuropsychiatric involvement may avoid such intensive treatment. Although diagnostic criteria are used to assess NPSLE activity [4], they largely rely on clinical judgment and the exclusion of other conditions, limiting diagnostic objectivity and consistency. No established biomarkers for disease activity currently exist, and significant overlap between NPSLE and treatment‐related or comorbid neuropsychiatric disorders further complicates diagnosis.

Imaging studies that assess NPSLE activity are of some interest. Conventional MRI may show brain atrophy, white matter lesions (WMLs), and vascular abnormalities such as cerebral infarctions or hemorrhages [5]. However, these findings are non‐specific and may also be observed in various other neurological or systemic conditions. Previous studies have reported abnormal conventional MRI findings in only approximately 50% of cases [6], and even when present, such findings do not always correlate with clinical symptoms or disease activity [7]. Therefore, conventional MRI is often insufficient to support a definitive diagnosis of NPSLE. Thus, advanced MRI methods are needed to improve diagnostic accuracy, especially in cases of NPSLE that show no abnormalities on conventional MRI [8]. Advanced MRI techniques have been increasingly studied in recent years to better detect subtle abnormalities. These include structural and volumetric MRI [9], diffusion tensor imaging [10], dynamic contrast‐enhanced MRI [11], arterial spin labeling [12], and quantitative susceptibility mapping [13], which allow for the evaluation of white matter changes, blood–brain barrier (BBB) integrity, cerebral perfusion, and iron deposition in SLE. However, while these methods show potential utility in assessing NPSLE, the findings remain inconsistent, and definitive imaging biomarkers for NPSLE have yet to be identified [14].

Vessel wall imaging (VWI) is an advanced MRI method used to evaluate intracranial vasculopathy [15, 16, 17, 18]. A previous study on SLE found that vessel wall lesions (VWLs) were particularly common in the anterior circulation and associated with cerebral infarctions, though the absence of contrast agents limited the scope of VWI [19]. Recently, contrast‐enhanced VWI (CE‐VWI) has gained importance for the detailed assessment of VWLs [20, 21, 22, 23, 24]. While its application in SLE has been described in two case reports [25, 26], studies investigating its effectiveness in diagnosing NPSLE in larger cohorts are currently lacking. These enhancing VWLs are thought to reflect underlying vascular abnormalities, such as non‐inflammatory vasculopathy and inflammatory vasculitis, both of which have been implicated in the pathogenesis of NPSLE [8]. CE‐VWI may help overcome the limitations of conventional MRI by directly visualizing pathological changes in vessel walls, enabling the detection of subtle vascular lesions even in the absence of overt parenchymal abnormalities. Improved diagnostic accuracy could facilitate earlier and more appropriate treatment decisions, potentially preventing irreversible neurological damage and optimizing outcomes in patients with suspected NPSLE.

Therefore, this study aimed to evaluate the diagnostic value of CE‐VWI in differentiating NPSLE from non‐NPSLE, and to elucidate the clinical relevance of contrast‐enhancing VWLs (CE‐VWLs).

2. Materials and Methods

This single‐center study was approved by the institutional review board (approval no. CRB21‐038) for prospective data collection. Written informed consent for prospective data use was obtained from all patients at the time of the MRI examination. The present analysis was conducted retrospectively using the accumulated data.

2.1. Study Participants

This study included consecutive patients with a definite diagnosis of SLE [27], who underwent CE‐MRI at our institution between March 2019 and October 2022. Patient enrollment was conducted in accordance with the inclusion and exclusion criteria outlined in Table S1. Specifically, SLE patients in this study were classified into NPSLE and non‐NPSLE groups based on the diagnostic criteria described in the following paragraph, with diagnoses confirmed by a board‐certified rheumatologist with 35 years of experience (Y.T.).

2.2. NPSLE Diagnosis

The diagnosis of NPSLE was determined based on a comprehensive multidisciplinary evaluation, in accordance with the 1999 American College of Rheumatology case definitions [4] and the 2010 European League Against Rheumatism recommendations [28]. This included clinical assessments by a rheumatologist, psychiatric evaluation by a psychiatrist using the Structured Clinical Interview for DSM‐5 (SCID‐5) in patients with evident psychiatric symptoms, conventional brain MRI, cerebrospinal fluid (CSF) analysis, electroencephalography, and nerve conduction studies. Importantly, CE‐VWI findings were not included in the diagnostic decision‐making process and were assessed independently for the purposes of this research.

2.3. Clinical Data Collection

Demographic and clinical information, including age, gender, disease duration, comorbid antiphospholipid syndrome, history of cardiovascular events, cardiovascular risk factors (hypertension, hyperlipidemia, diabetes mellitus), smoking history, laboratory markers (erythrocyte sedimentation rate, anti‐ deoxyribonucleic acid antibodies, CSF interleukin‐6, CSF protein, CSF immunoglobulin G index), Systemic Lupus Erythematosus Disease Activity Index scores, corticosteroid dosage, and immunosuppressant use, were obtained from electronic medical records. Neuro British Isles Lupus Assessment Group (Neuro‐BILAG) index scores were also obtained to evaluate NPSLE disease activity.

2.4. Patient Selection and Classification

A total of 117 consecutive SLE patients with suspected neuropsychiatric involvement were initially screened. After applying the exclusion criteria, 102 patients were included in the final analysis. Of these, 47 were classified as NPSLE and 55 as non‐NPSLE. The mean age was 44 years (SD = 13.2) in the NPSLE group and 43 years (SD = 16.5) in the non‐NPSLE group (Figure 1). Baseline clinical characteristics of all participants are summarized in Table 1. The 47 NPSLE patients were further subclassified into diffuse (n = 11), focal (n = 32), and peripheral nervous syndrome (n = 4), as detailed in Table 2.

FIGURE 1.

Flowchart of participant inclusion and exclusion. NPSLE, neuropsychiatric systemic lupus erythematosus; CSF, cerebrospinal fluid; EEG, electroencephalogram; CKD, chronic kidney disease; CE MRI, contrast‐enhanced MRI; MCTD, mixed connective tissue disease; SLE, systemic lupus erythematosus. *A high disease activity is indicated by a British Isles Lupus Assessment Group index of A for ≥ 1 item or B for ≥ 2 items. **CSF abnormalities were defined as a protein level > 40 mg/dL or IgG index > 0.73, and abnormal EEGs included those with spike or intermittent slow waves. ***CKD was defined as an estimated glomerular filtration rate of < 30 mL/min.

TABLE 1.

Summary of baseline participant characteristics.

	NPSLE (n = 47)	Non‐NPSLE (n = 55)	p
Age (y) a	44.3 ± 13.2	43.0 ± 16.5	0.66
Gender (female)	40 (85%)	49 (89%)	0.55
Disease duration (months) a	98.5 ± 108.5	95.6 ± 141.6	0.91
Antiphospholipid antibody syndrome	8 (17%)	20 (36%)	0.03
Any cardiovascular risk factor b	8 (17%)	10 (18%)	0.89
Current or past smoking	14 (30%)	12 (22%)	0.36
History of cardiovascular events	0 (0%)	0 (0%)	N/A c
ESR (mm) a	48.9 ± 36.8	36.4 ± 32.2	0.07
anti‐DNA antibody (IU/mL) a	1401 ± 2388	1051 ± 1845	0.46
anti‐dsDNA antibodies (IU/mL) a	54.9 ± 104.6	44.7 ± 112.6	0.65
CSF‐IL‐6 (pg/mL) a	20.0 ± 51.4	3.0 ± 2.1	0.09
CSF‐protein (mg/dL) a	43.1 ± 27.3	48.2 ± 34.7	0.52
CSF‐IgG index a	0.54 ± 0.16	0.54 ± 0.11	0.90
SLEDAI score (median IQR)	8 (3.5–16)	5 (0–13)	0.07
Corticosteroids (mg/day) a	5.2 ± 9.6	4.6 ± 7.4	0.72
Immunomodulatory drugs use	21 (45%)	31 (56%)	0.13

Abbreviations: CSF‐IL‐6, cerebrospinal fluid interleukin‐6; DNA, deoxyribonucleic acid; dsDNA, double‐stranded deoxyribonucleic acid; ESR, erythrocyte sedimentation rate; IgG, immunoglobulin G; IQR, interquartile range; N/A, not applicable; NPSLE, neuropsychiatric systemic lupus erythematosus; SLEDAI, systemic lupus erythematosus disease activity index.

^a Continuous variables are presented as means ± SDs.

^b Any cardiovascular risk factor included at least one of the following: hypertension, hyperlipidemia, and diabetes mellitus.

^c P‐value not applicable due to identical proportions (0%) in both groups.

TABLE 2.

Details of the subcategories of NPSLE in this cohort (total of 47 subjects).

Diffuse syndrome (n = 11)
Mood disorder	6 (13%)
Acute confusion state	3 (7%)
Psychosis	2 (4%)
Focal syndrome (n = 32)
Cerebrovascular disease	16 (34%)
Headache	7 (15%)
Seizure disorders	6 (13%)
Aseptic meningitis	3 (7%)
Peripheral nerve syndrome (n = 4)
Polyneuropathy	2 (4%)
Cranial neuropathy	2 (4%)

Note: Each number represents the actual patient count (percentage).

Abbreviation: NPSLE, neuropsychiatric systemic lupus erythematosus.

2.5. MRI Acquisition

All examinations were performed on a 3‐T MRI system (Signa Premier 3 T, GE Healthcare, Milwaukee, WI, USA) using a dedicated 48‐channel head coil (48ch Head Coil 3 T, GE Healthcare, Milwaukee, WI, USA). Our institution's standard protocol for brain MRI included diffusion‐weighted imaging (DWI), T1‐weighted imaging (T1WI), T2‐weighted imaging (T2WI), T2*‐weighted imaging (T2*WI), fluid‐attenuated inversion recovery (FLAIR) imaging, and MR angiography (MRA). The VWI parameters (3D‐T1‐weighted‐CUBE) were as follows: field of view (FOV) 24 cm, spatial resolution 0.7 × 0.8 × 0.9 mm, slice thickness 1 mm, repetition time (TR) 383 ms, echo time (TE) 10 ms, matrix 288 × 256, number of excitations (NEX) 1, bandwidth 90.1 kHz, acceleration factor (ARC) 2*1 (with an acceleration factor of 2 in the phase‐encoding direction and 1 in the slice‐encoding direction), compressed sensing (HyperSense) 1.25, and scan time 4 min. The imaging data were acquired in the sagittal plane, with the superior–inferior (S/I) coverage determined by the FOV, set to 240 mm, ensuring full brain coverage. All VWI acquisitions were performed after intravenous administration of contrast agents; no pre‐contrast VWI were obtained. At our institution, CE‐VWI is not performed routinely but is reserved for patients clinically suspected of having NPSLE. The CE‐VWI scans were performed using either 0.1 mL/kg of Gadavist (gadobutrol, Bayer Yakuhin Ltd., Osaka, Osaka, Japan) or 0.2 mL/kg of ProHance (gadoteridol, Bracco Japan Co. Ltd., Toshima, Tokyo, Japan.) or Magnescope (gadoterate meglumine, Guerbet Japan K.K., Chiyoda, Tokyo, Japan). Contrast agent selection followed a weekday‐based institutional protocol, resulting in random assignment. All contrast agents were injected slowly, and CE‐VWI was performed 5 min later to optimize vessel wall enhancement. Source images were reconstructed into 1‐mm coronal and axial datasets using the “Reformat/MIP” console application on the MR imaging system.

2.6. MRI Analyzes

All MRI scans were evaluated by three independent board‐certified radiologists: observer 1 (S.I.), with 14 years of experience; observer 2 (Y.M.), with 15 years of experience; and observer 3 (Y.Y.), with 7 years of experience. All three radiologists evaluated all cases and were blinded to the clinical data. For the purposes of systematic grading, each reader assessed the entire dataset without consulting the others.

2.7. CE‐VWI Assessments

The intracranial arteries were segmented into 15 regions ranging from the supraclinoid internal carotid arteries and basilar arteries to the second segments of each cerebral artery (A1–A2 for anterior cerebral artery, M1–M2 for middle cerebral artery, and P1–P2 for posterior cerebral artery). The segmentation was based on MRA and applied to CE‐VWI to help distinguish arteries from adjacent veins. Excluded segments were the anterior and posterior communicating arteries, omitted due to their small diameters and assessment challenges. Hypoplastic A1 or P1 segments, undetectable on MRA, were also excluded. Peripheral vessels, including the third segments and beyond of each cerebral artery, were omitted due to non‐diagnostic image quality, and the vertebral artery was excluded due to wall enhancement from the venous plexus [21].

VWLs were evaluated in a structured three‐step process. First, the presence or absence of VWLs was assessed. VWLs were defined as sites of distinct vessel wall thickening compared to the contralateral healthy side or to the proximal and distal segments under evaluation, according to the definition by van der Kolk et al. [29]. Second, for segments with wall thickening, the thickening pattern was classified as either concentric or eccentric, according to the criteria described by Swartz et al. [30]. Third, vessel wall enhancement was assessed in segments with wall thickening, following the method by Lindenholz et al. [31] and Larson et al. [32]. The pituitary stalk and brain parenchyma served as internal references. Enhancement was classified as absent (signal intensity similar to brain parenchyma), low‐grade (higher than parenchyma but lower than the pituitary stalk), or high‐grade (equal to or higher than the pituitary stalk). Representative examples of CE‐VWI findings are shown in Figures 2, 3, 4. Additionally, Figure 5 presents cases in which interobserver disagreement occurred, illustrating the variability in interpretation of VWLs. A full spectrum of VWLs—including absent, low‐grade, and high‐grade enhancements—across all intracranial arterial segments is shown in Figure S1.

FIGURE 2.

Representative case of a high‐grade enhancing vessel wall lesion. A 28‐year‐old female with diffuse syndrome (mood disorder) with NPSLE. (a) Sagittal and (b) coronal CE‐VWI demonstrate concentric wall thickening with high‐grade enhancement in the M1 segment of the left middle cerebral artery (arrows in a and c). The enhancement was equal to the pituitary stalk in signal intensity. An inset image of the pituitary stalk is included in the upper right corner of (b) as the internal reference. (c) MRA reveals no stenosis corresponding to the wall‐thickening site (arrow). NPSLE, neuropsychiatric systemic lupus erythematosus; CE‐VWI, contrast‐enhanced vessel wall imaging; MRA, MR angiography.

FIGURE 3.

Representative case of a non‐enhancing vessel wall lesion. A 40‐year‐old male with diffuse syndrome (acute disorientation) with NPSLE. (a) Sagittal and (b) coronal CE‐VWI demonstrate concentric wall thickening without enhancement (signal intensity similar to brain parenchyma) in the P1 portion of the right posterior cerebral artery (arrows in a and b). An inset image of the pituitary stalk is included in the upper right corner of (b) as the internal reference. (c) MRA reveals no stenosis corresponding to the wall‐thickening site (arrow). NPSLE, neuropsychiatric systemic lupus erythematosus; CE‐VWI, contrast‐enhanced vessel wall imaging; MRA, MR angiography.

FIGURE 4.

Representative case of a low‐grade enhancing vessel wall lesion with associated stenosis on MRA. A 46‐year‐old woman with NPSLE presenting with diffuse syndrome (acute confusion state). (a) Sagittal and (b) coronal CE‐VWI images demonstrate eccentric wall thickening with low‐grade enhancement in the right M1 segment (arrows). The enhancement was lower in signal intensity than the pituitary stalk but higher than brain parenchyma. An inset image of the pituitary stalk is included in the upper right corner of (b) as the internal reference. (c) MRA shows mild stenosis at the corresponding site of the vessel wall lesion (arrow). NPSLE, neuropsychiatric systemic lupus erythematosus; CE‐VWI, contrast‐enhanced vessel wall imaging; MRA, MR angiography.

FIGURE 5.

Representative cases of vessel wall lesions with interobserver disagreement. (a) A 61‐year‐old female with focal syndrome (cerebrovascular disease) with NPSLE. (a) Coronal CE‐VWI shows a moderately prominent enhancing lesion in the left supraclinoid ICA (arrow), which was assessed as low‐grade enhancement by two readers and high‐grade by one reader, illustrating interobserver variability in CE‐VWI grading. An inset in the upper right displays the pituitary stalk, which served as the internal reference for signal comparison. The arrowhead indicates a high‐grade enhancing lesion in the right supraclinoid ICA, on which all three readers agreed. (b and c) A 27‐year‐old male with focal syndrome (aseptic meningitis) with NPSLE. (b) Sagittal and (c) coronal CE‐VWI show a subtle vessel wall lesion in the right distal MCA M1 segment (arrow). Two readers classified the lesion as non‐enhancing, while one reader assessed it as low‐grade enhancement, highlighting the subjective nature of CE‐VWI interpretation, particularly in borderline cases. An inset in the upper right of (c) shows the pituitary stalk, which was used as the internal reference for signal intensity. CE‐VWI, Contrast‐Enhanced Vessel Wall Imaging; ICA, Internal Carotid Artery; MCA, Middle Cerebral Artery.

To minimize interpretation bias, MRI data from NPSLE and non‐NPSLE patients were reviewed in a randomized order. Radiologists focused exclusively on assessing intracranial VWLs, blinded to brain parenchymal lesions, by digitally masking the surrounding brain tissue on VWI using Advantage Workstation (GE Healthcare, Milwaukee, WI, USA). A representative example of the masked image used for blinded assessment is provided in Figure S2.

2.8. MRA Assessments

Stenosis calculations for each segment were performed using the criteria from the Warfarin–Aspirin Symptomatic Intracranial Disease Study [33]. Stenoses identified by the criteria were categorized according to the following grading scale: normal (0%–9%), mild (10%–29%), moderate (30%–69%), severe (70%–99%), or occluded (no flow detected).

2.9. Brain Lesion Assessments

Brain MRI abnormalities other than VWLs were assessed according to established radiological criteria. Cerebral infarctions were classified into five subtypes: lacunar infarctions (small < 15 mm subcortical infarcts due to single perforating artery occlusion) [34]; borderzone infarctions (infarcts at the junctions of major arterial territories) [35]; localized cortical infarctions (focal cortical lesions not involving an entire vascular territory) [36]; large perforator infarctions (deep infarcts caused by occlusion of proximal perforating arteries) [37]; and large territorial infarctions (infarctions affecting most or all of a major arterial territory) [38]. White matter lesions (WMLs) were graded on FLAIR images using the Fazekas scale, with grade 2 or 3 considered abnormal [39]. Cerebral hemorrhages were identified based on characteristic hypointense signal on T2*‐weighted imaging [40]. Brain atrophy was assessed using the Pasquier scale for global cortical atrophy, with a grade of 2 or higher interpreted as atrophy [41]. In addition, other findings, including brain swelling, abnormal signals or enhancement of brain parenchyma or cranial nerves, and dural thickening, were also recorded. Those brain lesions were assessed in a separate session using only conventional MRI sequences (DWI, T1WI, T2WI, T2*WI, and FLAIR), independently and blinded to VWI findings.

2.10. Subgroup Analysis of Non‐Cerebrovascular NPSLE

To further evaluate the diagnostic value of CE‐VWI in non‐cerebrovascular NPSLE, we conducted a subgroup analysis excluding patients diagnosed with cerebrovascular disease within the focal syndrome category. As a result, 16 patients were excluded from the NPSLE cohort, yielding a non‐cerebrovascular NPSLE subgroup of 31 patients. Comparisons of CE‐VWLs, VWLs, and arterial stenosis on MRA were performed between this subgroup and the non‐NPSLE group. Additionally, the diagnostic accuracy of CE‐VWLs in this subgroup was analyzed to evaluate its ability to distinguish non‐cerebrovascular NPSLE from non‐NPSLE.

2.11. Statistical Analyzes

All statistical analyzes were performed using the Bell Curve for Excel software program (version 4.08; Social Survey Research Information Co. Ltd., Tokyo, Japan). The results of observer 1 were used for statistical analysis. Paired comparisons of the MRI and clinical data between the two groups were conducted. Categorical variables, presented as numbers and percentages, were compared using the χ² test, while continuous variables, shown as means and standard deviations (SDs), were analyzed using the independent samples t‐test. A multivariate logistic regression analysis identified factors (clinical data, CE‐VWI and MRA findings, and brain lesions) associated with NPSLE among the variables with significant differences in paired comparisons. The diagnostic accuracy for NPSLE using CE‐VWI was evaluated using a receiver operating characteristic (ROC) analysis. The cutoff value for the parameters in the ROC analysis was defined using the Youden index. The χ ² test was used to compare brain lesion prevalence, NPSLE symptoms, and Neuro BILAG index disease activity [42] between cases with and without CE‐VWLs. A multiple linear regression analysis was used to identify confounding factors for CE‐VWLs, including age, disease duration, corticosteroid dosage, immunosuppressant use, smoking history, and cardiovascular risk factors.

Interobserver agreement was assessed using the complete dataset, which included 1513 arterial segments from 102 patients. All three readers independently evaluated the same set of segments during the first reading session. Fleiss' kappa values were calculated for both VWL patterns (no lesion, concentric, or eccentric wall thickening) and enhancement patterns (absent, low‐grade, or high‐grade). Intraobserver agreement was additionally assessed by repeating the evaluations in a second reading session conducted after a 2‐month interval, and Cohen's kappa values were calculated for each reader. The strength of agreement was considered fair for values of 0.21–0.40, moderate for κ values of 0.41–0.60, good for κ values of 0.61–0.80, and excellent for κ values of ≥ 0.81. Statistical significance was set at a p‐value < 0.05.

3. Results

3.1. Paired Comparisons of Brain Lesions of NPSLE/Non‐NPSLE

Large perforator infarctions were significantly more frequent in patients with NPSLE than in those without NPSLE (11% vs. 0%). In contrast, other types of cerebral infarctions showed no significant differences between groups: lacunar infarctions (p = 0.29), borderzone infarctions (p = 0.29), localized cortical infarctions (p = 0.55), and large territorial infarctions (p = 0.29). Overall, cerebral infarctions were significantly more common in NPSLE patients than in non‐NPSLE patients (26% vs. 3%). There were no significant differences between the groups in the WMLs of Fazekas grade 2 or 3 (p = 0.68), cerebral hemorrhage (p = 0.29), or brain atrophy (p = 0.30). Brain swelling, parenchymal or cranial nerve abnormalities, and dural thickening were infrequent but observed more often in the NPSLE group; however, statistical analysis was not performed due to the low number of cases. Consequently, abnormal findings on conventional MRI were significantly more common in the NPSLE group (53%) than in the non‐NPSLE group (27%) (Table 3).

TABLE 3.

Paired comparisons of brain/vessel wall/stenotic lesions between NPSLE and non‐NPSLE.

	NPSLE (n = 47)	Non‐NPSLE (n = 55)	p
Abnormal MRI	25 (53%)	15 (27%)	0.004
Any cerebral infarctions	12 (26%)	2 (3%)	< 0.001
Lacunar infarctions	3 (6%)	1 (2%)	0.29
Borderzone infarctions	3 (6%)	1 (2%)	0.29
Localized cortical infarctions	2 (4%)	1 (2%)	0.55
Large perforator infarctions	5 (11%)	0 (0%)	0.01
Large territorial infarctions	1 (2%)	0 (0%)	0.29
WMLs (>Fazekas grade 2)	7 (15%)	10 (18%)	0.68
Cerebral hemorrhage	3 (6%)	1 (2%)	0.29
Brain atrophy	4 (9%)	2 (4%)	0.30
Other brain lesions a	11 (23%)	2 (4%)	NA
VWLs (median IQR)	11 [8–12]	6 [3–9]	< 0.001
Total segments with VWLs	465/696 (67%)	349/817 (42%)	< 0.001
Percentage of concentric nature	351/465 (75%)	295/349 (85%)	NA
CE‐VWLs (median IQR)	2 (0.5–4)	0 (0–1)	< 0.001
Total segments with CE‐VWLs	141/465 (30%)	27/349 (8%)	< 0.001
Patients with ≥ 1 ce‐VWL	35 (74%)	17 (31%)	< 0.001
Patients with ≥ 1 CE‐VWL in Normal MRI	14/22 (64%)	8/40 (20%)	< 0.001
Patients with stenotic MRA	19 (40%)	5 (9%)	< 0.001
Total segments of stenosis	39/696 (6%)	7/817 (1%)	< 0.001
Percentage of mild stenosis	36/39 (93%)	7/7 (100%)	0.44
Percentage of moderate stenosis	3/39 (7%)	0/7 (0%)	0.47

Abbreviations: CE‐VWL, contrast‐enhanced vessel wall lesion; IQR, interquartile range; MRA, MR angiography; NA, not analyzed; NPSLE, neuropsychiatric systemic lupus erythematosus; VWLs, vessel wall lesions; WMLs, white matter lesions.

^a Other brain lesions included brain swelling/abnormal signals, abnormal enhancement of the brain parenchyma, and cranial nerves, and dural thickening.

3.2. Paired Comparisons of CE‐VWI and MRA of NPSLE/Non‐NPSLE

Among the 47 NPSLE patients, 696 segments were evaluated; among the 55 non‐NPSLE patients, 817 segments were assessed. Both groups had a median of 15 evaluable segments per patient (interquartile range [IQR], 15–15), with minimal variability. There was no significant difference in the distribution of evaluated arterial segments between the NPSLE and non‐NPSLE groups (Table S2). The NPSLE group had significantly more VWLs than the non‐NPSLE group (median, 11; IQR, 8–12 vs. median, 6; IQR, 3–9), with 75% showing a concentric pattern. Furthermore, CE‐VWLs were also significantly more frequent in the NPSLE group than in the non‐NPSLE group (median, 2, IQR 0.5–4 vs. median 0; IQR, 0–1). Additionally, MRA stenosis was significantly more common in the NPSLE group (40%, 19 cases) than in the non‐NPSLE group (9%, 5 cases) and involved significantly more stenotic segments. In both groups, > 90% of stenotic lesions on MRA were mild (Table 3). In assessing the distribution of VWLs, wall thickening was observed throughout the major cerebral arteries in both groups, but the NPSLE group showed a significantly higher frequency than the non‐NPSLE group in most segments (Figure 6a). Furthermore, CE‐VWLs were significantly more prevalent in the NPSLE group than in the non‐NPSLE group, with higher frequencies across almost all segments (Figure 6b).

FIGURE 6.

Differences in the frequency of (a) VWLs and (b) CE‐VWLs between NPSLE and non‐NPSLE. VWLs, vessel wall lesions; CE‐VWLs, contrast‐enhanced vessel wall lesions; NPSLE, neuropsychiatric systemic lupus erythematosus.

Among the 47 patients with NPSLE, 22 (47%) showed no abnormalities on conventional MRI, of whom 14 (64%) exhibited contrast‐enhancing vessel wall lesions (CE‐VWLs). In contrast, among the 55 non‐NPSLE patients, 40 had normal conventional MRI findings, and CE‐VWLs were present in 8 of these 40 cases (20%).

The interobserver agreement among the three readers for VWI was as follows: VWL patterns had a Fleiss' kappa value of 0.84, and enhancement patterns had a value of 0.89. The intraobserver agreement (Cohen's κ, 2‐month interval) was: Observer 1, 0.83 (VWL), 0.88 (enhancement); Observer 2, 0.87 (VWL), 0.86 (enhancement); Observer 3, 0.80 (VWL), 0.82 (enhancement).

3.3. Identifying Clinical and Imaging Factors Associated With NPSLE

A multivariate logistic regression analysis indicated that total CE‐VWLs remained the only significant factor associated with NPSLE (odds ratio [OR], 1.97; 95% confidence interval [CI], 1.23–3.16) (Table 4).

TABLE 4.

Multivariate logistic regression analysis of factors associated with NPSLE.

Variables	Univariate analysis		Multivariate analysis
	OR (95% CI)	p	OR (95% CI)	p
APS	0.36 (0.14–0.92)	0.03	0.30 (0.09–1.03)	0.55
Any abnormal MRI	3.03 (1.33–6.91)	0.008	1.11 (0.36–3.50)	0.85
Any cerebral infarctions	9.09 (1.92–43.1)	0.006	3.20 (0.50–20.3)	0.22
VWLs	1.29 (1.14–1.45)	< 0.001	1.10 (0.94–1.29)	0.24
CE‐VWLs	2.29 (1.53–3.43)	< 0.001	1.97 (1.23–3.16)	0.005
Stenotic arterial lesions	3.08 (1.51–6.28)	0.002	3.36 (0.91–12.5)	0.07

Abbreviations: APS, antiphospholipid antibody syndrome; CE‐VWL, contrast‐enhanced vessel wall lesion; CI, confidence interval; NPSLE, neuropsychiatric systemic lupus erythematosus; OR, odds ratio; VWLs, vessel wall lesions.

3.4. Diagnostic Accuracy of CE‐VWI for Differentiating NPSLE From Non‐NPSLE

The ROC analysis showed that the area under the curve (AUC) for CE‐VWLs was 0.78, with a sensitivity of 60%, specificity of 87%, and an OR of 10.1 at the optimal cutoff value of 2. Notably, high‐grade enhancement was exclusively observed in NPSLE patients, resulting in a specificity of 100%. When VWLs were used as an indicator, the AUC was 0.75, and the OR was 6.00 (Figure 7).

FIGURE 7.

Diagnostic accuracy of vessel wall imaging for differentiating NPSLE from non‐NPSLE using a receiver operating characteristic analysis. CE‐VWLs, contrast‐enhanced vessel wall lesions; VWLs, vessel wall lesions; NPSLE, neuropsychiatric systemic lupus erythematosus.

3.5. Paired Comparisons of CE‐VWI and MRA of Non‐Cerebrovascular NPSLE/Non‐NPSLE

The non‐cerebrovascular NPSLE group (n = 31) showed a significantly higher median number of CE‐VWLs (2 [IQR, 0.5–3.5] vs. 0 [0–1]), total VWLs (10 [7–11.5] vs. 6 [3–9]), and a greater proportion of patients with stenotic MRA lesions (14/31 [45%] vs. 5/55 [9%]) compared to the non‐NPSLE group. ROC analysis using CE‐VWL count to differentiate non‐cerebrovascular NPSLE from non‐NPSLE demonstrated comparable diagnostic performance to that of the full cohort (AUC = 0.77; specificity, 87%; sensitivity, 58%) (Figure S3).

3.6. The Comparison of Imaging and Clinical Features Based on CE‐VWL Status

On dividing the cohort (102 patients) by the CE‐VWL status, 52 patients had CE‐VWLs, while 50 did not. In the comparison of imaging and clinical features based on CE‐VWL status, patients with CE‐VWLs had a significantly higher prevalence of cerebral infarction (21% vs. 6%) and WMLs (35% vs. 4%), while no significant difference was observed for acute cerebral infarctions (10% vs. 6%, p = 0.22). The frequency of NPSLE symptoms and high disease activity scores according to Neuro BILAG were also significantly higher in patients with CE‐VWLs (67% vs. 24% and 32% vs. 10%; Table 5). The distribution of symptoms differed, with focal and diffuse involvement being more prevalent in patients with CE‐VWLs, whereas non‐NPSLE symptoms were dominant in those without CE‐VWLs (Figure 8).

TABLE 5.

A comparison of imaging and clinical features in cases with and without CE‐VWL.

Variables	With CE‐VWL (n = 52)	Without CE‐VWL (n = 50)	p
Any cerebral infarctions	11 (21%)	3 (6%)	0.01
Acute cerebral infarctions	5 (10%)	3 (6%)	0.22
WMLs (>Fazekas grade 2)	18 (35%)	2 (4%)	< 0.001
Cerebral hemorrhage	3 (6%)	1 (2%)	0.15
Brain atrophy	3 (6%)	3 (6%)	0.48
CSF‐IL‐6 (pg/mL) a	19.5 ± 50.7	3.0 ± 2.2	0.06
CSF‐protein (mg/dL) a	48.9 ± 35.5	40.0 ± 21.3	0.21
CSF‐IgG index a	0.56 ± 0.2	0.51 ± 0.1	0.11
Frequency of NPSLE symptoms b	35 (67%)	12 (24%)	< 0.001
High disease activity of NPSLE c	17 (32%)	5 (10%)	0.003

Abbreviations: CE‐VWL, contrast‐enhanced vessel wall lesion; CSF, cerebrospinal fluid; IgG, immunoglobulin G; IL‐6, interleukin‐6; NPSLE, neuropsychiatric systemic lupus erythematosus; WMLs, white matter lesions.

^a Continuous variables are presented as means ± SDs.

^b NPSLE symptoms include diffuse, focal, and peripheral nerve syndromes.

^c A high disease activity is defined as a score of A or B in the neuro British Isles Lupus assessment group criteria.

FIGURE 8.

Distribution of NPSLE symptoms based on with or without CE‐VWLs. Panel (a) represents the group with CE‐VWLs (n = 52) and Panel (b) represents the group without CE‐VWLs (n = 50). NPSLE, neuropsychiatric systemic lupus erythematosus; CE‐VWLs, contrast‐enhanced vessel wall lesions; Diffuse, diffuse syndrome; Focal, focal syndrome; PNS, peripheral nerve syndrome; Non‐NP, non‐neuropsychiatric systemic lupus erythematosus.

3.7. Identifying Confounding Factors for CE‐VWLs in Clinical Data

The analysis revealed that the age and NPSLE diagnosis were significant predictors. Specifically, both age (β = 0.05, t = 3.038) and NPSLE diagnosis (β = 2.44, t = 5.228) had significant β values, with NPSLE demonstrating a stronger and substantial association with the number of CE‐VWLs (Table S1).

4. Discussion

A severe complication of SLE is NPSLE, which contributes to morbidity and mortality. Although intracranial vessel wall MRI may help diagnose NPSLE, the effectiveness of CE‐VWI in larger cohorts remains unclear. We therefore conducted a retrospective analysis to evaluate the diagnostic value of CE‐VWI in distinguishing NPSLE from non‐NPSLE patients. Overall, CE‐VWI allowed us to detect significantly more CE‐VWLs in NPSLE patients than in non‐NPSLE patients. A multivariate analysis identified CE‐VWLs as the sole significant factor for NPSLE. ROC analysis demonstrated that CE‐VWLs had good diagnostic performance for distinguishing NPSLE from non‐NPSLE, with high specificity at the optimal cutoff value. In addition, CE‐VWLs were significantly associated with more frequent cerebral infarctions, WMLs, NPSLE symptoms, and a higher disease activity.

Two main mechanisms are involved in the pathogenesis of NPSLE: ischemic and autoimmune‐inflammatory processes [8]. Ischemic processes are driven by vasculopathy or vasculitis, which can activate the immune system via anti‐endothelial cell antibodies, leading to endothelial damage associated with immune complex deposition and complement activation [43]. This causes luminal narrowing, thrombosis, and subsequent infarction or hemorrhage [44]. Autoimmune‐inflammatory processes are similarly exacerbated by endothelial damage caused by vasculopathy or vasculitis, increasing the permeability of neuroimmune barriers, such as the BBB, blood‐cerebrospinal fluid barrier (choroid plexus), glymphatic system, and meningeal barrier [8]. This allows harmful autoantibodies to enter the central nervous system, causing neurotoxicity, demyelination, and inflammation by targeting neurons, astrocytes, and myelin [45]. Both mechanisms can lead to focal or diffuse syndromes [8]. A key point is that cerebral vessel wall damage contributes to the onset and progression of NPSLE, with intracranial VWLs potentially reflecting these changes [8].

Common cerebral arterial lesions in NPSLE include vasculopathy, vasculitis, vasospasm [46], atherosclerosis [47], and arterial dissections [48]. In the present study, most VWLs were concentric, a pattern typically associated with vasculitis [21, 49], which involves inflammatory infiltrates and vessel wall destruction [8]. CE‐VWLs generally indicate active inflammation leading to luminal narrowing, infarction, or hemorrhage [26]. However, the predominance of mild stenosis and the absence of group differences in acute infarction suggest that active vasculitis was unlikely. By contrast, vasculopathy is characterized by endothelial proliferation, vessel wall thickening, and lumen narrowing without inflammation [8]. While limited, one pathological study in SLE demonstrated concentric wall thickening in the superficial temporal artery [50]. CE‐VWLs may reflect vasculopathy due to BBB dysfunction and contrast leakage. This is supported by CE‐VWI findings in cerebral amyloid angiopathy, a non‐inflammatory vasculopathy with BBB disruption [51]. In our cohort, cerebrospinal fluid protein, interleukin‐6, and immunoglobulin G index, markers associated with active neuroinflammation, showed no significant differences between the two groups. Thus, the CE‐VWLs in our study likely reflected vasculopathy rather than vasculitis.

To explore the clinical relevance of CE‐VWLs, we compared the imaging and clinical findings between patients with and without CE‐VWLs. The group with CE‐VWLs had a significantly higher prevalence of any cerebral infarctions and WMLs, a higher frequency of NPSLE symptoms, and greater disease activity of the Neuro BILAG index than the group without CE‐VWLs. These findings suggest that the presence of CE‐VWLs reflects active vasculopathy and may be associated with NPSLE pathophysiology. In the present study, high‐grade enhancement was observed exclusively in NPSLE cases, suggesting that such cases may involve severe endothelial damage, increased BBB permeability, and extensive vasa vasorum expansion.

The diagnostic performance of CE‐VWI for NPSLE was characterized by high specificity but only moderate sensitivity. This suggests that while CE‐VWLs are highly specific markers for NPSLE, their absence does not rule out the diagnosis. Importantly, CE‐VWLs were frequently observed in patients who showed no abnormalities on conventional MRI, indicating that CE‐VWI may be particularly valuable in cases with clinical suspicion of NPSLE but unremarkable conventional imaging findings. In such diagnostically challenging situations, the identification of CE‐VWLs may increase diagnostic confidence and support clinical decision making. These findings suggest that CE‐VWI could be integrated as an adjunctive tool in the diagnostic imaging protocol for SLE patients with suspected NPSLE, especially when conventional imaging is inconclusive. Thus, although CE‐VWI should not replace conventional diagnostic methods, it may serve as a useful adjunct in the comprehensive evaluation of suspected NPSLE.

Although CE‐VWLs were predominantly associated with NPSLE, a subset of non‐NPSLE patients also exhibited such lesions. This raises the possibility that some of these individuals may represent early or subclinical stages of neuropsychiatric involvement. Therefore, the presence of CE‐VWLs in non‐NPSLE patients may warrant closer clinical follow‐up and monitoring for potential neuropsychiatric manifestations. Future studies are needed to clarify whether CE‐VWLs in non‐NPSLE patients can serve as early imaging biomarkers for NPSLE.

The number of VWLs per patient in our study (mean 9.9 in NPSLE, 6.3 in non‐NPSLE) was relatively higher than that reported in a previous study involving stroke or TIA patients (mean 3.2 lesions per patient) [29]. This discrepancy may reflect the unique vascular pathophysiology of SLE, including immune‐mediated vasculopathy, as well as the potential influence of immunomodulatory therapies and corticosteroid use. Notably, a previous study using non‐contrast VWI in 60 SLE patients reported a mean of 9.6 VWLs per patient, which is comparable to our findings [19]. These findings suggest that numerous VWLs can be identified in patients with SLE, underscoring the potential importance of VWI in this population.

4.1. Limitations

First, as this was a retrospective, single‐center study with a moderate sample size, there may be selection biases that could impact its external validity. Although we applied predefined inclusion and exclusion criteria, NPSLE is known for its highly heterogeneous clinical presentation. Our inclusion criteria were designed to capture a wide range of suspected NPSLE cases. Nevertheless, atypical cases—such as patients with mild or non‐specific symptoms—that did not meet these criteria might have been inadvertently excluded. This may have influenced the generalizability of our findings to the broader SLE population. Second, the limited spatial resolution of the VWI in this study prevented an evaluation beyond the third segment. Incorporating noise reduction techniques, such as deep learning reconstruction, might enable the performance of high‐resolution VWI more quickly, thereby allowing the assessment of distal arteries—an area for future research. Third, without a histopathological examination, whether the detected VWLs truly represent vasculopathies remains uncertain. Fourth, as a retrospective cross‐sectional analysis, this study lacked longitudinal imaging data, which limits the ability to assess temporal changes in VWLs or to evaluate CE‐VWI as a biomarker for disease activity and treatment response. Moreover, since CE‐VWI was not incorporated into the clinical diagnostic process, its practical contribution to real‐world NPSLE diagnostics remains uncertain. To address these limitations, future research should include prospective, multicenter studies with longitudinal follow‐up to clarify the clinical utility of CE‐VWI in monitoring disease progression, predicting outcomes, and guiding therapeutic decisions in NPSLE. Fifth, this study did not include a quantitative assessment of vessel wall enhancement. Lesions were evaluated visually by three board‐certified radiologists, which may introduce subjectivity, especially in subtle cases, although both inter‐ and intra‐observer agreement were high. Future studies using quantitative metrics are needed for standardization. Sixth, the use of multiple contrast agents with different relaxivity profilesmay have introduced variability in enhancement grading—particularly for subtle lesions—despite random assignment. Lastly, the lack of pre‐contrast VWI—due to longer scan times—limits the ability to clearly distinguish true enhancement from intrinsic T1 hyperintensity, such as that seen in intraplaque hemorrhage. Although pre‐ and post‐contrast comparisons would be ideal, we applied a standardized grading system using post‐contrast VWI with internal references [31, 32]. Notably, most VWLs in this study showed concentric thickening, in contrast to the eccentric appearance typically associated with atherosclerotic plaques. Nonetheless, cautious interpretation is warranted, especially in subtle cases.

5. Conclusions

Imaging by CE‐VWI may demonstrate high specificity and good diagnostic performance for differentiating NPSLE from non‐NPSLE, with CE‐VWLs in major intracranial arteries being more frequently observed in NPSLE patients. Furthermore, CE‐VWLs may be associated with cerebral infarctions, WMLs, neuropsychiatric symptoms, and high disease activity in NPSLE.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1. Supplementary Information.

Ide S., Fujita Y., Murakami Y., et al., “The Diagnostic Value of Contrast‐Enhanced Vessel Wall MRI for Diagnosing Neuropsychiatric Systemic Lupus Erythematosus,” Journal of Magnetic Resonance Imaging 62, no. 4 (2025): 1021–1034, 10.1002/jmri.29810.