Cortical changes in women with systemic lupus erythematosus with mild cognitive impairment: a voxel-based morphometry and surface-based morphometry study

Abstract

Background

The purpose of this study was to reveal the morphological changes of grey matter (GM) in women systemic lupus erythematosus (wSLE) patients with mild cognitive impairment (MCI) with normal conventional MRI.

Methods

The differences in brain morphological indicators among wSLE with MCI, wSLE without MCI and women healthy control (wHC) group were calculated and compared by voxel-based morphometry and surface-based morphometry. The GM volume (GMV), cortical thickness (CT), indicators of cortical complexity, including fractal dimension (FD), gyration index (GI), sulcus depth, the relationship between brain morphological indicators and clinical features, were analysed.

Results

In comparison to wSLE patients without MCI (n=36), wSLE with MCI (n=26) demonstrated a significant decrease in FD of the left lateral orbitofrontal gyrus. When compared with the wHC group (n=36), both wSLE patients with MCI and wSLE without MCI group exhibited a reduction in GMV in the medial of right superior frontal gyrus, a thinning of CT in the left paracentral and postcentral gyrus as well as in the right pars triangularis gyrus and superior frontal gyrus. Within the wSLE group, Mini-Mental State Examination scores were positively correlated with GMV in the middle of right superior frontal gyrus and with the FD of the left lateral orbitofrontal gyrus.

Conclusion

WSLE patients with MCI have brain morphological changes such as reduced GMV, thinning CT, reduced FD and increased GI. Cortical morphological changes may be involved in the pathological process of MCI in wSLE patients.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Systemic lupus erythematosus (SLE) disproportionately affects women and is frequently associated with cognitive impairment (CI). Existing research predominantly indicates that cognitive deficits in SLE patients are linked to grey matter damage.

WHAT THIS STUDY ADDS

• There are cortical morphological changes in women SLE patients with mild CI (MCI).

• The decrease of grey matter volume (GMV), cortical thickness (CT) and changed cortical complexity may underlie MCI in women SLE patients.

• The changes of GMV and cortical complexity in some brain regions are correlated with cognitive function.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The integration of voxel-based morphometry and surface-based morphometry enables the detection of early brain morphological alterations in women SLE patients with MCI. Specifically, GMV, CT, fractal dimension, gyration index and sulcal depth in certain brain regions may serve as potential biomarkers for this patient population.

Introduction

Systemic lupus erythematosus (SLE) is a systemic autoimmune disorder that is prone to occur in women of childbearing age. The pathogenesis is intricate and may involve multiple organs and systems.^{1 2} Central nervous system (CNS) involvement in patients with SLE is termed neuropsychiatric SLE (NPSLE). NPSLE is the second leading cause of death after lupus nephritis in SLE patients.³ Cognitive impairment (CI) is one of the 19 neuropsychiatric symptoms in the American College of Rheumatology (ACR) NPSLE classification criteria.⁴ Among them, attention and memory are the most frequently affected cognitive domains.⁵ CI often emerges in SLE among women with or without obvious CNS involvement, which can seriously impact the quality of life of patients, but its potential pathogenesis remains unclear.⁶ Additionally, some researchers have discovered that CI is an independent manifestation of nervous system involvement in SLE.⁷ Among them, mild CI (MCI) refers to a subtle yet measurable decline in cognitive function that can be objectively verified through neuropsychological assessment; moreover, MCI signifies an intermediate phase between normal ageing and dementia.⁸

The diagnosis of NPSLE is primarily based on clinical manifestations, and there is no single laboratory marker or imaging method that can serve as the ‘gold standard’.⁹ Studies have found that detailed neuropsychological measurements and brain MRI are effective approaches to assess the CNS involvement in SLE.¹⁰ MRI, encompassing structural MRI (sMRI) and functional MRI, is a crucial approach for assessing the alterations in brain structure and function in SLE patients. However, conventional MRI often fails to capture subtle neurostructural changes in SLE. Here, combining voxel-based morphometry (VBM) and surface-based morphometry (SBM) allows quantification of subclinical brain morphological changes, offering insights into early neurodegenerative mechanisms. VBM utilises 3D-T1-weighted image (T1WI) structural images to conduct voxel-by-voxel quantitative analysis of grey matter volume (GMV).¹¹ SBM analysis is based on brain tissue segmentation and cortical reconstruction to calculate cortical thickness (CT), gyrification index (GI), fractal dimension (FD) and sulcal depth (SD) and other parameters, which can effectively detect regional variations in brain morphology.¹² Objective evaluation of local differences in the brain has unveiled subtle morphological alterations in the brains of SLE patients without overt neuropsychiatric manifestations, who are still in the subclinical stage of brain damage.¹¹

A prior study has indicated that patients with NPSLE exhibit characteristic subcortical and regional grey matter (GM) atrophy in comparison to healthy controls (HCs).¹³ It was found that a smaller GMV was positively correlated with lower levels of perceptual speed and episodic memory in community residents.¹⁴ Previous research from our team has established that SLE patients have significantly diminished GMV relative to HCs. SLE patients among women have a notably reduced GMV compared with SLE patients among men.¹⁵ In addition, some studies have reported that, in contrast to non-NPSLE patients and HCs, NPSLE patients display structural and functional alterations in GM, particularly involving cognitive, and sensorimotor regions.¹⁶

CI is a common clinical manifestation in many neurological disorders, and the diagnosis of CI in SLE patients is often postponed.¹⁷ Moreover, previous studies have mainly centred on analysing GMV or CT as individual indicators and have seldom explored the changes in subtle brain structures such as FD, GI and SD.

As mentioned above, the prevalence of SLE in women is much higher than that in men, and studies have suggested that gender has a certain effect on cognitive function.¹⁸ Additionally, a meta-analysis showed that men outperformed women in most cognitive domains in the healthy population, and that women with Alzheimer’s disease (AD) worsened cognitive function faster than men in the early stages of the disease.¹⁹ SLE predominantly affects women of childbearing age, a phase associated with heightened ovarian endocrine activity. Elevated levels of estradiol and prolactin have been linked to SLE pathogenesis, particularly in active disease.²⁰ However, oestrogen has been found to have a protective effect on cognitive function in healthy elderly women and AD patients.²¹ Therefore, the pathogenesis of CI in women SLE patients (wSLE) is more complicated. Women account for a large proportion of SLE patients. Although MCI may have minimal impact on daily life initially,²² its progression can significantly affect academic performance, professional functioning and overall quality of life as cognitive deficits become more severe. Therefore, we need to pay attention to wSLE patients with MCI. Moreover, current research evidence suggests that the neural basis of general intelligence differs between the sexes.²³ Our study was conceived to investigate potential alterations in cortical volume and complexity in wSLE patients with MCI. We hypothesised that wSLE have structural damage to GM, which is more pronounced in wSLE patients with MCI. To comprehensively explore structural brain changes, we adopted a combination of VBM and SBM analyses. This comprehensive approach thoroughly examines the characteristics of brain structural changes underlying the development of MCI in wSLE patients with normal conventional MRI from a morphological perspective.

Participants and methods

Participants

WSLE patients from the Department of Rheumatology and Immunology of the First Affiliated Hospital of Kunming Medical University were recruited as the wSLE group. Women healthy volunteers during the same period were recruited as women healthy controls (wHCs). The research process of collecting clinical data and MRI imaging data was carried out following a standardised scheme, and the same researcher monitored the entire research process. Each subject was interviewed by a rheumatologist and psychiatrist with rich clinical experience, and a detailed and comprehensive physical examination was conducted, with a focus on neuropsychiatric lesions. The inclusion and exclusion criteria were as described in the next section.

Inclusion and exclusion criteria

Inclusion criteria for the SLE group: (1) WSLE patients fulfilling the SLE classification criteria set by the ACR in 1997, without a diagnosis of NPSLE, presenting no obvious neuropsychiatric symptoms, and having normal T1 and T2-weighted imaging scans of conventional head MRI, namely, wSLE patients without evident neuropsychiatric manifestations; (2) aged between 18 and 50 years; all participants were right-handed (assessed by the Edinburgh Handedness Inventory); (3) those who voluntarily take part in this study are informed and sign the informed consent and are capable of cooperating with MRI examination and questionnaire evaluation.

Exclusion criteria for the wSLE group: (1) having a history of severe traumatic brain injury; (2) patients suffering from organic encephalopathy or nervous system disorders that affect the brain structure (such as head trauma or surgery, Parkinson’s disease or epilepsy, stroke, etc); (3) having a history of depression, anxiety, schizophrenia or other serious mental illnesses; (4) patients with a history of alcoholism and drug abuse; (5) combined with other autoimmune disorders; (6) patients with severe clinical manifestations that might influence brain structure, such as severe hypertension, diabetes or renal insufficiency; (7) there are MRI scan contraindications: for instance, claustrophobia, metal foreign bodies like dentures; (8) patients who are unable to cooperate in completing the scale evaluation and questionnaire filling; (9) lactating and pregnant women.

Inclusion criteria for the wHCs group: (1) age comparable to that of the wSLE group; (2) judged as right-handed by the Edinburgh Hand Habit Scale (score >40 points); (3) absence of severe physical or mental disorders, and in good physical and mental health; (4) normal results in conventional head MRI T1 and T2-weighted imaging scans; (5) normal findings after comprehensive physical and neurological examinations conducted respectively by experienced rheumatologists and neurologists and further screened by experienced psychiatrists using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) non-patient structured clinical interview; (6) non-lactating and non-pregnant women; (7) subjects who voluntarily participated in this study, signed the informed consent and were able to cooperate with MRI examination and questionnaire evaluation.

Information collection

Demographic information, clinical indicators and relevant mental scales: (1) demographic information: gender, age, educational attainment, etc. (2) clinical-related indicators: duration of disease, disease activity index; the duration of disease was determined from the onset of the initial clinical manifestations that could be clearly attributed to SLE to the date of MRI acquisition. All clinical manifestations and laboratory tests were recorded in accordance with ACR criteria. (3) SLE disease activity index 2000 (SLEDAI-2k) score: it was employed to assess the disease activity of SLE patients and was accomplished on the day of MRI. (4) Evaluation of cognitive function: Mini-Mental State Examination (MMSE) was utilised to evaluate the cognitive function of the participants in the study. The maximum MMSE score was 30 points. The scores were judged as normal at 27–30 points, 21–26 points indicated MCI.²⁴ All scales were evaluated by an experienced psychiatrist within 2 days before and after the MRI examination.

Participants were divided into three groups: wSLE with MCI (MMSE scores 21–26), wSLE without MCI (MMSE scores 27–30) and wHCs (MMSE scores 27–30).

Imaging data acquisition: all image acquisition was carried out by an experienced radiologist. All subjects utilised the same 1.5T MRI scanner (Twinspeed; GE Medical Systems) along with a bird cage head coil to accomplish the data acquisition, with the scanning range encompassing the entire brain. The subjects were positioned in a supine posture, and a foam support pad was employed to minimise head movement. During the scanning process, the eyes were closed, and the body was relaxed and remained stationary. During the magnetic resonance scanning, blood pressure, pulse and respiration were monitored, and video surveillance was conducted in the scanning chamber. At the conclusion of each sequence, the subjects were reminded to stay awake. Conventional T1WI and T2WI scans were conducted to rule out obvious structural abnormalities. The sMRI data employed 3D-T1-weighted fast spoiled gradient echo sequence (3D-T1-FSPGR sequence). The parameters were as follows: Repetition Time (TR)=10.5 ms, Echo Time (TE)=2.0 ms, Inversion Time (TI)=350 ms, layer thickness=1.8 mm, no layer interval, imaging matrix=256×256, Field of View (FOV)=24cm×18cm, Flip Angle (FA)=15°, spatial resolution=0.94×0.94×0.94 mm³, layer number=172.

Image processing

VBM and SBM analyses were performed using the Computational Anatomy Toolbox 12 (CAT12) (V.12.8.2; http://dbm.neuro.unijena.de/cat) toolbox on the SPM12 (V.6906; http://www.fil.ion.ucl.ac.uk/spm) platform in MATLAB 2016b, which effectively calculate and compare group differences in indicators such as GMV, CT and cortical complexity in structural imaging data. The technical details of its preprocessing procedures have been described in previous literature. To facilitate understanding, we hereby provide a detailed overview.

VBM preprocessing

1. Denoising and bias correction: the VBM process begins with spatially adaptive non-local means denoising filters to reduce noise in the raw 3D-T1 images. This step is crucial for improving the signal-to-noise ratio and enhancing the quality of subsequent analyses.

2. Bias correction: intensity inhomogeneity correction is applied to normalise the intensity distribution across the brain, ensuring that variations in image intensity are not due to scanner artefacts but rather to true anatomical differences.

3. Segmentation and registration: the images are then segmented into GM, white matter (WM) and cerebrospinal fluid (CSF) compartments using a probabilistic approach. The segmentation is followed by affine registration to align the images to a common space.

4. Normalisation and modulation: the segmented images are normalised to the Montreal Neurological Institute (MNI) 1.5 mm template using the Diffeomorphic Anatomical Registration through Exponentiated Lie algebra algorithm. This step ensures that the anatomical structures are aligned across subjects, facilitating group comparisons. Modulated GMV maps were generated by multiplying the Jacobian determinants from deformation fields to preserve absolute tissue volume postnormalisation.

5. Volume calculation: total intracranial volume (TIV) is calculated for each subject to account for differences in head size as covariates in subsequent analyses to control for potential confounding effects. Additionally, GMV, WMV and CSF volumes are computed for each subject.

SBM preprocessing

1. CT estimation: the SBM process involves CT estimation using the projection-based thickness method. This method provides a robust estimation of CT by projecting vertices from the WM surface to the pial surface.

2. Surface reconstruction and refinement: central surface reconstruction is performed to create a mid-cortical surface that represents the average position between the GM/CSF and GM/WM boundaries. Topology correction and surface refinement are applied to ensure accurate cortical surface representation.

3. Template transformation and uniformity check: the local thickness values are transferred to the ‘fsaverage’ template, and the uniformity of the obtained images is checked to ensure consistency across subjects.

4. Cortical complexity measures: The FD was gauged through the spherical harmonic reconstruction approach, enabling the preservation of an identical quantity of vertices for all reconstructed surfaces. This helped to mitigate the influence of individual vertices alignment, resampling and interpolation, leading to a more precise reconstruction. The GI was computed in accordance with the absolute mean curvature method, and a grid of the central surface (ie, the surface between the GM/CSF and the GM/WM boundary) was formed by cortical surface extraction. Subsequently, the local absolute mean curvature of the ventral surface was determined by averaging the mean curvature values of each vertex within 3 mm of a specific location. SD was obtained from the Euclidean distance between the central surface and its convex hull. The square root transformation was employed to make the data exhibit a more normal distribution.

After preprocessing, Gaussian kernel spatial smoothing was performed on structural index maps to reduce individual variance and ensure normal distribution of data, so as to improve the effectiveness of parameter statistical test. Grey-matter segments were generated using the default option with an absolute masking threshold of 0.1, resliced to an isotropic voxel size of 1.5 mm³ and then smoothed as follows: the GMV maps were smoothed with a Gaussian kernel with a half width of 8 mm, CT and complexity maps with 15 and 20 mm, Gaussian kernels, respectively.

Data quality control involves two steps: first, visually inspect all original Digital Imaging and Communications in Medicine (DICOM) images before processing to remove any obvious artefacts. Second, after processing, CAT12 assigns each subject an image quality rate based on resolution, deviation and noise. Subjects with quality below ‘B-’ (good quality) are excluded. In this study, no subjects were excluded after applying these quality control steps.

Statistical analysis

IBM SPSS Statistics V.27.0 (SPSS) was used for demographic and clinical data statistics. Shapiro-Wilk normality test was performed on all data. The data from the three groups that were normally distributed were reported in the form of mean±SD, while non-normally distributed data were expressed as medians (lower quartiles and upper quartiles). Then Kruskal-Wallis H test was used to compare the differences of age, education level and MMSE among the three groups. Mann-Whitney U test was used to compare the differences in course of disease, SLEDAI-2k between wSLE with MCI group and wSLE without MCI group. χ² test was used to compare the antibody positive rate of wSLE patients with MCI and wSLE patients without MCI. Categorical variables were compared with the χ² test; when any expected cell count was <5, Fisher’s exact test was used instead. The statistical significance threshold was set to p<0.05.

Cat12/SPM12, DPABI (a toolbox for Data Processing & Analysis for (Resting-State) Brain Imaging, http://rfmri.org/dpabi)²⁵ and SPSS V.27.0 were used to analyse the smoothed GMV and cortical complexity maps. First, covariance analysis (ANCOVA) was performed in the GM mask using the statistical module of DPABI software to analyse the differences in GMV among the three groups at the voxel level. Age, years of education and TIV were used as covariates, and the results were corrected using Gaussian Random Field (GRF) correction (p voxel <0.001, p cluster <0.005). ANCOVA was performed on CT and cortical complexity indicators (FD, GI, SD) using Cat12/SPM12 at the vertex level, with age and years of education as covariates, and using the alpha statistical threshold p<0.05 (Family Wise Error correction) or p<0.001 without correction. Next, the raw values of the significantly different clusters were extracted and Bonferroni correction in SPSS V.27.0 was used for post hoc comparison (p<0.05/3=0.017). Finally, Spearman correlation coefficient was used to evaluate the correlation between MMSE score and morphological changes in SLE with MCI and SLE without MCI patients.

Result

Demographic and clinical data

According to the above exclusion and inclusion criteria, 98 participants were enrolled: 26 wSLE patients with MCI, 36 wSLE patients without MCI and 36 matched wHCs. No subjects were excluded due to image quality issues. Demographic and clinical data of all subjects are shown in table 1. There was no significant difference in age and education level between wSLE with MCI group, wSLE without MCI group and wHCs group. There was no difference in SLEDAI-2k and disease duration between wSLE with MCI group and wSLE without MCI group. In the wSLE with MCI group, anti-dsDNA positivity was higher in the MCI group (76% vs 50 %, p=0.043).

Table 1. Demographic and clinical data of subjects.

	wSLE with MCI (n=26)	wSLE without MCI (n=36)	wHCs (n=36)	Statistic H(K)/U/χ2	P value
Age (year)	31.50 (25.00, 37.00)	30.50 (24.25, 37.50)	32.0 (19.25, 38.50)	0.03	0.99*
Education (year)	9.00 (9.00, 12.00)	12.00 (9.00, 14.00)	12.00 (9.00, 12.00)	3.93	0.14*
Disease duration (month)	21.00 (4.75, 38.25)	11.00 (2.25, 24.00)	–	371.50	0.168†
MMSE	25.00 (23.00, 26.00)	29.00 (28.00, 29.00)	30.00 (30.00, 30.00)	76.63	<0.001*
SLEDAI-2k	9.00 (6.50, 13.25)	8.50 (4.25, 13.50)	–	432.00	0.61†
Anti-dsDNA antibody (%)	19/25 (76.00)	17/34 (50.00)	–	4.09	0.043‡
Anti-Sm antibody (%)	11/25 (44.00)	10/34 (29.40)	–	1.34	0.247‡
Anti-P0 antibody (%)	13/25 (52.00)	15/34 (44.10)	–	0.36	0.549‡
Anti-ACL-Ig antibody (%)	3/15 (20.00)	3/17 (17.60)	–	–	0.999‡
Anti-ACL-IgG antibody (%)	4/19 (21.10)	3/26 (11.50)	–	–	0.433‡
Anti-ACL-IgM antibody (%)	8/19 (42.10)	7/26 (26.90)	–	1.14	0.286‡
LAC (%)	7/15 (46.70)	3/18 (16.70)	–	–	0.126‡
β2GPⅠ antibody (%)	5/15 (33.30)	3/18 (16.70)	–	–	0.418‡
Anti-U1RNP antibody (%)	4/25 (16.00)	8/34 (23.50)	–	0.50	0.478‡
Anti-SSA 52KD antibody (%)	17/25 (68.00)	19/34 (55.90)	–	0.89	0.346‡
Anti-SSA 60KD antibody (%)	18/25 (72.00)	22/34 (64.70)	–	0.35	0.554‡
Anti-SSB antibody (%)	12/25 (48.00)	11/34 (32.40)	–	1.48	0.223‡
Anti-histones antibody (%)	15/25 (60.00)	14/34 (41.20)	–	2.04	0.153‡
Anti-nucleosome antibody (%)	10/25 (40.00)	13/34 (38.20)	–	0.02	0.891‡
Anti-centromere antibody (%)	1/25 (4.00)	1/34 (2.90)	–	–	0.999‡
Anti-DNP antibody (%)	2/25 (12.00)	4/30 (13.3)	–	–	0.999‡

^*The Kruskal-Wallis H test was used to compare the differences in age, education level and MMSE among the three groups of wSLE with MCI, wSLE without MCI and wHCs.

^†The Mann-Whitney U test was used to compare the differences in course of disease and SLEDAI-2k between wSLE with MCI group and wSLE without MCI group.

^‡χ2 test was used to compare the composition ratio of autoantibodies between wSLE with MCI group and wSLE without MCI group.

Anti-ACL-Ig antibody, anti-cardiolipin-Ig antibody; Anti-ACL-IgG antibody, anti-cardiolipin-IgG antibody; Anti-ACL-IgM antibody, anti-cardiolipin-IgM antibody; Anti-DNP antibody, anti deoxyribose nucleoprotein antibody; Anti-dsDNA antibody, anti-double-stranded DNA; Anti-P0 antibody, anti-ribosomal P0-protein autoantibody; Anti-Sm antibody, anti-Smith antibody; Anti-SSA 52KD antibody, anti-Sjögren's syndrome-related antigen A 52-kilodalton antibody; Anti-SSA 60KD antibody, anti-Sjögren's syndrome-related antigen A 60-kilodalton antibody; Anti-SSB antibody, anti-Sjögren's syndrome type B antibody; Anti-U1RNP antibody, anti-U1 ribonucleoprotein antibody; LAC, lupus anticoagulant; MCI, mild cognitive impairment; MMSE, Mini-Mental State Examination; SLEDAI-2k, SLE disease activity index 2000; wHCs, women healthy controls; wSLE, women systemic lupus erythematosus; β2GPI antibody, anti-β2 glycoprotein I antibody.

GMV analysis

Whole-brain voxel-wise analysis revealed a single cluster (peak MNI: x=4.5, y=31.5, z=61.5; cluster volume=845 mm³; F=15.49, p-voxel <0.001, p cluster <0.005, GRF-corrected) in the medial right superior frontal gyrus where GMV differed among the three groups. Post hoc comparisons showed that both wSLE with MCI (p<0.0001, Bonferroni correction) and wSLE without MCI (p=0.0013, Bonferroni correction, η²=0.257, power (1-β)=1) had lower GMV in this region than wHCs, whereas the difference between the two patient subgroups did not reach significance (p=0.062, Bonferroni correction) (figure 1). Thus, the reduced GMV in the medial right superior frontal gyrus distinguished both SLE groups from controls, but not the MCI from the non-MCI patients.

Figure 1. Brain regions with significant differences in GMV in wSLE with MCI group (n=26), wSLE without MCI group (n=36) and wHCs group (n=36). Specifically, regions of reduced GMV in the three groups (A, B) (p voxel<0.001, p cluster<0.005, age, years of education and TIV as covariates, GRF correction), (A) the total GMV in wSLE with MCI group was lower than that in wHCs (^*p < 0.05). (B) The peak point is located in the cluster of the medial of right superior frontal gyrus. (C) GMV distribution map in the cluster with peak point located in the medial of right superior frontal gyrus. Post-hoc analysis showed that GMV was significantly reduced in the wSLE with MCI group (^****p < 0.0001) and wSLE without MCI group (^**p < 0.01) compared with wHCs. Red represents wSLE with MCI group, orange represents wSLE without MCI group, blue represents wHCs. GMV, grey matter volume; GRF, Gaussian Random Field; MCI, mild cognitive impairment; TIV, total intracranial volume; wHCs, women healthy controls; wSLE, women systemic lupus erythematosus.

Analysis of CT and complexity

SBM revealed four regions—left pars triangularis, right pars opercularis, left middle temporal gyrus and right lateral occipital cortex—where CT was reduced in both SLE groups relative to controls (p<0.05, Bonferroni correction). In addition, orbitofrontal FD was decreased and temporal GI increased exclusively in the SLE-MCI group (p<0.05, Bonferroni correction). The specific results are as follows:

Analysis of CT showed that there was a significant difference in the CT of four regions in the left and right cerebral hemispheres between the three groups, which were located in the left paracentral gyrus (F=12.62, p_uncorr<0.001), the left postcentral gyrus (F=10.71, p_uncorr<0.001), the right pars triangularis gyrus (F=13.09, p_uncorr<0.001) and the right superior frontal gyrus (F=11.01, p<0.001, p_uncorr<0.001) (table 2 and figure 2A). Post hoc one-way analysis of variance found: compared with the wHCs group, the CT of the left paracentral gyrus (p=0.0060, Bonferroni correction), the left postcentral gyrus (p=0.0018, Bonferroni correction), the right pars triangularis gyrus (p<0.0001, Bonferroni correction) and the right superior frontal gyrus (p=0.0051, Bonferroni correction) decreased in the wSLE with MCI group. However, wSLE without MCI group showed thinner CT of the left paracentral gyrus (p<0.0001, Bonferroni correction), the left postcentral gyrus (p=0.0002, Bonferroni correction), the right pars triangularis gyrus (p=0.0002, Bonferroni correction) and the right superior frontal gyrus (p<0.0001, Bonferroni correction) compared with wHCs. There was no statistically significant difference in the above four brain regions between wSLE with MCI group and wSLE without MCI group (all p>0.05) (figure 2B–E).

Table 2. Names, locations, sizes, statistics, coordinates of brain regions with thinner CT and differences in FD in wSLE with MCI group (n = 26) compared with wHCs (n = 36) and wSLE without MCI (n = 36).

Morphology index	Hemisphere	Brain region	Cluster size	F-value	P value	Peak MNI coordinates			η2	Power (1−β)
						X	Y	Z
CT
Region 1	LH	100% paracentral	91	12.62	<0.001	−6	−27	51	0.255	>0.999
Region 2	LH	88% postcentral 12% precentral	101	10.71	<0.001	−54	−10	22	0.164	0.991
Region 3	RH	54% pars triangularis 34% pars opercularis 12% rostral middle frontal	101	13.09	<0.001	53	27	15	0.210	>0.999
Region 4	RH	91% superior frontal 7% caudal middle frontal 2% rostral middle frontal	97	11.01	<0.001	22	23	43	0.007	0.104
FD
Region 1	LH	100% lateral orbitofrontal	7	7.77	0.001	−17	46	−18	0.070	0.697
GI region 1	LH	74% inferior temporal 26% fusiform	20	8.55	<0.001	−43	−23	−23	0.086	0.803

CT, cortical thickness; FD, fractal dimension; GI, gyration index; LH, left hemisphere; MCI, mild cognitive impairment; MNI, Montreal Neurological Institute; RH, right hemisphere; wHC, women healthy control; wSLE, women systemic lupus erythematosus.

Figure 2. Brain areas with thinning CT and differences in FD and GI in wSLE with MCI group (n=26), wSLE without MCI group (n=36) and wHCs group (n=36). Specifically, (A) represents the CT reduction area in the three groups (cortical complexity index for covariance analysis, age, years of education and TIV as covariate and α statistical threshold p<0.05 (FWE correction) or p_uncorr<0.001). The peak point is located in the cluster diagram of the left paracentral gyrus, the left postcentral gyrus, the right pars triangularis gyrus and the right superior frontal gyrus. (B–E) wSLE with MCI group and wSLE without MCI group compared with wHCs group CT thinning area (p_uncorr<0.001). (B) The peak point was located in the left paracentral gyrus (F=12.62, p_uncorr<0.001). Post hoc analysis showed that the CT of wSLE with MCI group (**p<0.01, Bonferroni correction) and wSLE without MCI group (****p<0.0001, Bonferroni correction) were significantly thinner than that of wHCs group. (C) The peak point was located in the left postcentral gyrus (F=10.71, p_uncorr<0.001). Post hoc analysis showed that the CT of wSLE with MCI group (**p<0.01, Bonferroni correction) and wSLE without MCI group (***p<0.001, Bonferroni correction) were significantly thinner than that of wHCs group. (D) The peak point was located in the right pars triangularis gyrus (F=13.09, p_uncorr<0.001). Post hoc analysis showed that the CT of wSLE with MCI group (****p<0.0001) and wSLE without MCI group (***p<0.001, Bonferroni correction) were significantly thinner than that of wHCs group. (E) The peak point was located in the right superior frontal gyrus (F=11.01, p<0.001, p_uncorr<0.001). Post hoc analysis showed that the CT of wSLE with MCI group (**p<0.01) and wSLE without MCI group (****p<0.0001, Bonferroni correction) were significantly thinner than that of wHCs group. (F) represents differences in FD values among the three groups of wSLE with MCI, wSLE without MCI and wHCs. The brain area was located in the lateral orbitofrontal gyrus of the left cerebral hemisphere (F=7.77, p_uncorr=0.001). (G) Post hoc analysis found that the FD of the left lateral orbitofrontal gyrus was decreased in the wSLE with MCI group compared with the wSLE without MCI group (***p<0.001, Bonferroni correction) and wHCs group (*p<0.05, Bonferroni correction). (H) represents the brain regions with different GI in the three groups, and the peak point is located in the left inferior temporal gyrus (F=8.55, p_uncorr<0.001). (I) Post hoc analysis showed that there was statistically significant increase in the GI of the left inferior temporal gyrus (**p<0.01, Bonferroni correction) in the wSLE without MCI group compared with wHCs group. CT, cortical thickness; FD, fractal dimension; FWE, family-wise error; GI, gyrification index; MCI, mild cognitive impairment; TIV, total intracranial volume; wHCs, women healthy controls; wSLE, women systemic lupus erythematosus.

Among the three groups, the brain region with FD difference was located in the lateral orbitofrontal gyrus of the left cerebral hemisphere (F=7.77, p_uncorr=0.001) (table 2 and figure 2F). Post hoc one-way analysis of variance found: there was significantly decreased in FD of the left lateral orbitofrontal gyrus in wSLE with MCI group compared with wSLE without MCI group (p=0.0002, Bonferroni correction) and wHCs group (p=0.0218, Bonferroni correction) (figure 2G).

The GI of the left inferior temporal gyrus (F=8.55, p_uncorr<0.001) was significantly different among the three groups (table 2 and figure 2H). Post hoc one-way analysis of variance found: there was statistically significant increase in the GI of the left inferior temporal gyrus (p=0.0022, Bonferroni correction) in the wSLE with MCI group compared with wHCs group (figure 2I).

No brain regions with significant differences in SD were found in the three groups of wSLE with MCI, wSLE without MCI and wHCs.

Correlation analysis

In wSLE with MCI group (n=26), there was no significant correlation between MMSE, SLEDAI-2k, course of disease and GMV, CT, FD, GI (all p>0.05).

In wSLE without MCI group (n=36), there was no significant correlation between MMSE, SLEDAI-2k, course of disease, and GMV, CT, FD, GI (all p>0.05).

In wSLE group (wSLE with and without MCI group, n=62), there was a positive correlation between the GMV value of the medial of right superior frontal gyrus (r=0.2881, p=0.0232) and the FD of the left lateral orbitofrontal gyrus (r=0.4953, p<0.0001) with MMSE. At the same time, there was a negative correlation between the total GMV value (r=−0.3063, p=0.0155) with period of disease, and between the CT of the left postcentral gyrus (r=−0.2675, p=0.0356) with SLEDAI-2k (figure 3).

Figure 3. Correlation analysis between GMV value, FD, CT and MMSE, period of disease, and SLEDAI-2k in wSLE group (n=62). (A) The GMV value of the medial of right superior frontal gyrus in wSLE group was positively correlated with MMSE (r=0.2881, p=0.0232). (B) The figure shows that the FD of the left lateral orbitofrontal gyrus in the wSLE group was positively correlated with MMSE (r=0.4953, p<0.0001). (C) There was a negative correlation between the total GMV value (r=−0.3063, p=0.0155) with period of disease. (D) The CT of the left postcentral gyrus (r=−0.2675, p=0.0356) was negatively correlated with SLEDAI-2k. CT, cortical thickness; FD, fractal dimension; GMV, grey matter volume; MCI, mild cognitive impairment; MMSE, Mini-Mental State Examination; SLEDAI-2k, SLE disease activity index; wHC, women healthy control; wSLE, women systemic lupus erythematosus.

Discussion

Because of the complexity of its pathogenesis, the universality of the affected brain regions and the diversity of clinical manifestations, NPSLE has always been a hot and difficult point in research, especially in the study of pathogenesis and biomarkers. Studies have suggested that GM is particularly affected in NPSLE patients.²⁶ The pathogenesis of wSLE patients is related to oestrogen, and the cognitive function is also affected by oestrogen, and the neuroanatomical differences in cognition have been confirmed to be gender differences.²⁷ Therefore, the pathogenesis of CI in wSLE patients may be different from that of CI in man SLE patients.

Previous researches have confirmed that NPSLE patients have reduced GMV.¹³ Our team’s previous studies have also found that structural brain atrophy may occur even before obvious neuropsychiatric symptoms and signs appear and is associated with subclinical symptoms such as CI.¹⁵ Consistent with the previous findings, we found a reduction in total GMV in wSLE with MCI compared with wHCs. Further studies have also found that NPSLE patients showed significant subcortical GM atrophy in the same area without significant brain dysfunction compared with the HCs.¹⁶ Studies have found that temporal lobe structure may be the most reduced structure in wSLE patients with CI compared with wSLE patients without CI, but there is a lack of HCs for comparison.²⁸ Similar to the results of a previous longitudinal study on SLE patients, the study found that patients with severe CI had more extensive GM atrophy, especially in the frontal and temporal lobes, but no difference was observed between SLE patients without neuropsychiatric involvement and HCs.²⁹ Previous studies have found that frontal lobe³⁰ plays an important role in the relationship between various brain regions and cognitive function. In healthy people, it was found that 84% of GM voxels related to cognition were located in the frontal lobe in women, while 45% in men.²³ Our study further substantiates robust quantitative evidence. The study revealed that, relative to HCs, patients with wSLE combined with MCI exhibited a significant reduction in overall GMV and in specific frontal lobe regions. The decrease in GMV within the frontal lobe may play a crucial role in the pathogenesis of MCI associated with wSLE. Additionally, we observed that compared with wHCs, wSLE patients without MCI also demonstrated a significant reduction in GMV in corresponding frontal lobe regions. The GMV of the medial of right superior frontal gyrus in wSLE patients with MCI was significantly smaller than that in wSLE patients without MCI, and early studies have found that selective GM tissue-specific atrophy mainly occurs in SLE patients with diffuse NP syndrome.³¹ Therefore, we speculate that the above-mentioned brain regions may be among those that are prone to early damage in patients with wSLE. However, further studies with a larger sample size are needed to confirm these findings and determine whether these areas are indeed the first to be affected.

The study of CT in NPSLE patients also found that compared with HCs, the CT of left frontal lobe and parietal lobe cluster, right parietal lobe and occipital lobe in NPSLE group decreased.¹³ Compared with SLE patients without episodic memory deficits and HCs, SLE patients with episodic memory deficits showed significant cortical thinning in the clusters of the left marginal cortex and superior temporal gyrus.³² The above studies suggest that the reduction of CT in specific brain regions may be the morphological cause of disease-related CT. Consistent with our previous research results,³³ our study also found that wSLE patients with MCI had thinner CT in specific regions. However, previous studies did not consider the gender-induced inconsistency. In this study, CT of parts of the frontal and parietal lobe differences was found in wSLE patients with MCI. Compared with wHCs, wSLE patients with MCI and wSLE without MCI showed significantly thinner CT in the frontal lobe and parietal lobes, but there was no significant difference between wSLE patients without MCI and patients with MCI.

The brain is a hierarchical organisation with a high degree of topological and functional complexity, and FD can analyse and quantify the geometric complexity of the brain.³⁴ The higher the FD and the lower the porosity, the more complex and healthy the brain is.³⁵ Many neurological diseases including Alzheimer’s disease, frontotemporal dementia, Parkinson’s disease and multiple sclerosis have found a decrease in FD.³⁶ Studies on the association between FD and consciousness level have found that regardless of the damage mechanism, higher FD networks are associated with higher levels of consciousness, confirming the close relationship between human brain structure and its ability to support cognitive function.³⁷ Another study also compared the cortical FD between Alzheimer ’s patients with amnestic mild CI (aMCI) and HCs. It was found that the FD of the bilateral temporal lobe, right marginal lobe and right parietal lobe in the aMCI group was significantly lower.³⁸ In a longitudinal cohort study of the transition from MCI to dementia, cortical GM-FD was found to be the primary predictive feature of the transition period.³⁹ In patients with frontotemporal dementia, the correlation analysis between FD and cognition showed that the orbitofrontal cortex and paracentral gyrus were particularly vulnerable to memory and language impairments.⁴⁰ This study found that wSLE with MCI exhibited significantly lower FD compared with those wSLE without MCI and wHCs. No correlation between MMSE and FD was observed in the wSLE with MCI group, but a positive correlation between MMSE scores and FD was found in the wSLE group, regardless of the presence or absence of CI. Therefore, we conclude that the reduction of FD in the orbitofrontal gyrus may be one of the causes of wSLE with MCI. In addition to the complex geometric structure, GI was be used to characterise the complex morphological structure of the brain’s folded or smooth cortical surface.⁴¹ Previous studies have found that GI is associated with cognitive function.⁴² In Parkinson’s patients, frontal-parietal and temporal lobe regions were found to be associated with cognitive-related cortical thinning and decreased GI.⁴³ FD and GI of CI have been studied in other diseases, but there is no study on the GI of wSLE patients with MCI. Orbitofrontal FD reduction and temporal GI elevation were seen only in wSLE-MCI, signifying early cortical reorganisation: loss of FD coupled with GI preceding atrophy. Comparable transient ‘hypercomplexity’ has been described in presymptomatic Alzheimer’s⁴⁴ and type-2 diabetes⁴⁵ without CI, underscoring FD and GI as early sensitive markers of maladaptive plasticity.

Another complex morphological structure of the brain is the sulcus. Previous studies have shown that the SD and GI based on surface cortical morphology measurement show higher feature importance than GMV.⁴⁶ Earlier findings suggest that the changes in SD are correlated with CI in various diseases.⁴⁷ Our study focuses on the depth of cortical sulcus. There was no previous study on the depth of sulcus in wSLE with MCI patients. Although there was no significant difference between the three groups. More research is needed in the future to confirm the changes in SD in patients with SLE and their correlation with CI.

The study is aimed to investigate the impact of SLE and MCI on brain structure in women patients. Our results demonstrated that both wSLE groups (with and without MCI) showed extensive changes in cortical structure compared with wHCs, which indicates that SLE itself affects brain structure. Notably, compared with the wSLE without MCI group, the wSLE with MCI group exhibited a significant difference in FD in one cluster. This suggests that cognitive function differences may also be related to structural brain changes. However, it is important to note that the causal links between structural changes, SLE disease status and SLE-related MCI require further study with larger samples and cohort research.

This is the first comprehensive study of GM microstructure in wSLE patients with MCI using SBM combined with VBM, aiming to find early brain injury in wSLE patients with MCI, and our results also found significant differences in GMV, CT, FD and GI in some brain regions. There was no significant difference in SD, which may be due to the fact that our patient group (wSLE with MCI) was still in the early stage of brain damage. With the progression of the disease, the brain damage is aggravated, resulting in a decrease in GMV. Additionally, changes in cortical complexity in specific brain regions may underlie the morphological alterations associated with wSLE-related CI.

By analysing the correlation between brain morphological changes and clinical indicators, we found that the GMV values of the medial of right superior frontal gyrus and the FD of the left lateral orbitofrontal gyrus in the wSLE group (including wSLE with and without MCI group, n=62) were positively correlated with the MMSE. In other words, the reduction in GMV and FD in specific brain regions among wSLE patients may be associated with the development of MCI. This result is also consistent with the previous research results.^{40 48} We found that there was a negative correlation between the total GMV value with period of disease, and the CT of the left postcentral gyrus with SLEDAI-2k, this is also consistent with the results of previous studies.⁴⁹

However, our study also has certain limitations. First, as an exploratory investigation, no a-priori power analysis was performed; post hoc sensitivity analyses indicate, the CT in the right superior frontal gyrus, the reduction of FD in the left lateral orbitofrontal cortex, and the elevation of GI in the left inferior temporal gyrus were detected with only modest statistical power (≈ 0.10–0.80). These findings likely represent genuine yet subtle alterations that may be underestimated in the present sample, thereby increasing the risk of type II error; larger cohorts are needed to confirm their reliability. Second, the moderate sample size (n=98) and single-centre design restrict generalisability. Third, MCI was defined solely by MMSE, which is insensitive to subtle executive deficits; future work should incorporate comprehensive neuropsychological batteries.²⁴ Then, unmeasured factors—including cumulative corticosteroid dose, antimalarial or immunosuppressant exposure, and cardiovascular comorbidities—may further confound the observed structural alterations. Moreover, the cross-sectional design precludes causal inferences and prognostic conclusions; longitudinal studies are required to clarify the temporal evolution of neuroanatomical and cognitive changes. Finally, because most of the previous studies did not consider the cognitive-related neuroanatomical differences caused by gender differences, this is also a big reason for the inconsistency of the research results. We have observed that prior studies suggest that occupational status can impact cognitive reserve.⁵⁰ Unfortunately, occupational data were not collected in the present study, and we, therefore, cannot control for this potential confounder. In the future, more stratified and multicentre cohort studies with larger samples are needed to supplement and clarify the pathogenesis of these factors on wSLE with MCI.

In summary, wSLE exhibited region-specific reductions in GMV and CT, together with orbitofrontal FD decline and temporal GI elevation, consistent with early cortical reorganisation secondary to systemic inflammation. These metrics could be incorporated into routine cognitive monitoring of wSLE patients to trigger early anti-inflammatory or neuroprotective interventions. Longitudinal studies are required, however, to establish their prognostic value for incident CI.

Data availability statement

Data are available upon reasonable request.