Altered dynamic functional connectivity associated with cognition in diverse severity of white matter hyperintensity Authors: Tianyuyi Feng, Yunfei Li, Chunxiao Wei, Xiaohu Zhao Published: 2025 ABSTRACT AbstractAimTo investigate the dynamic functional connectivity (DFC) pattern in diverse severity of white matter hyperintensity (WMH) and explore the relationship between DFC temporal properties and cognitive impairment in WMH severity.MethodsOur study enrolled 85 CSVD patients diagnosed by the Neurology Department of the Fifth People’s Hospital of Shanghai, Fudan University. All participants underwent demographic surveys, assessments of vascular risk factors, neuropsychological testing, both structural and resting-state functional MRI scans. To explore the influence of WMH severity on cognition, subjects were categorized into two groups based on their WMH Fazekas rating scale scores: (1)mild WMH group scored 1–2 (n = 55); (2)severe WMH group scored 3–6 (n = 30). We calculated DFC by using sliding window correlation approach. Subsequently, we employed k-means clustering to identify distinct DFC states and calculated DFC temporal properties (including mean dwell time, fractional windows and transition numbers).ResultsThe intrinsic brain functional connectivity of both mild and severe WMH subjects was clustered into four distinct connectivity states (state 1: intermediate pattern, state 2: a more frequent, sparsely connected pattern, state 3: intermediate pattern, state 4: a less frequent, strongly connected pattern). Compared to mild WMH group, severe WMH subjects dwelled relatively longer in hypoconnected state 2, and shorter in hyperconnected state 4. Across the whole WMH subjects, mean dwell time and fractional windows of hypoconnected state 2 showed negative exploratory correlations with executive function, whereas the corresponding metrics of hyperconnected state 4 showed positive exploratory correlations before multiple comparison correction. Additionally, transition numbers demonstrated a negative correlation with the WMH Fazekas rating scale scores.ConclusionThe severity of WMH affects DFC temporal properties to a certain extent, and exploratory correlations suggest a possible link to cognition. These exploratory findings need replication in larger, multicenter cohorts before firm conclusions can be drawn.Keywords: Cerebral small vessel disease, White matter hyperintensity, Cognition, Dynamic functional connectivity Introduction Cerebral small vessel disease (CSVD) is a clinical and imaging syndrome caused by a variety of etiology that affects small arteries, micro arterials, capillaries and small veins in the brain [1], which is a common cause of vascular cognitive dysfunction and vascular dementia. Notably, CSVD tends to manifest prominently on MRI scans among the elderly [2]. With the coming of aging society, the study of CSVD takes on paramount significance. The main imaging features of CSVD include white matter hyperintensity (WMH), lacunar infarcts, lacunes, perivascular spaces, microbleeds and brain atrophy [3].Among them, the most common marker is WMH, which was reported related to cognitive impairment closely, particularly within domains of executive function. Several studies have endeavored to uncover the intricate association between WMH and cognitive impairment [4, 5], while the relationship between WMH severity and cognition remains unclear. Most prior investigations exploring the link between WMH severity and cognition have leaned on structural magnetic resonance imaging and resting-state functional magnetic resonance imaging (rs-fMRI). Wang et al. have discerned that heightened WMH burden accentuated gray matter volume (GMV) atrophy in related brain regions, thereby contributing to cognitive decline [6]. Chen et al. found that patients with heavy WMH load could cause cognitive decline by affecting the structural connection between hippocampus and thalamic subregion [7]. On the other hand, rs-fMRI, which typically measures the temporal correlation of BOLD signals, often ignore the spontaneous fluctuations in brain information interaction [8–10]. Indeed, brain functional connectivity changes dynamically over timescales even during the resting state [11], and the dynamic fluctuations of brain connectivity analysis could provide more information that cannot be observed in the static state. Dynamic functional connectivity (DFC) is a kind of method of estimating time-varying functional connectivity by dividing the entire length of time into several shorter time scales to assess time-varying properties within brain. Generally, three main temporal parameters are used to characterize DFC temporal properties, including mean dwell time (the number of consecutive windows belonging to one state), fractional windows (the number of total windows belonging to one state) and transition numbers (the number of transitioning between states). Currently, DFC analysis had been widely used in study of several neurological diseases. For instance, research on Alzheimer’s disease (AD) has indicated that AD patients tend to exhibit prolonged mean dwell times in a weakly connected state [12]. Studies of Parkinson’s disease have suggested that increased mean dwell times in segregated states and reduced number of transitions between states may contribute to the onset of dementia [13]. These studies presented the valuable roles of DFC analysis in elucidating the underlying mechanisms of cognition and neurological diseases. As far as we know, DFC analysis has not been applied to diverse severity of WMH. In the present study, we applied DFC analysis to explore the relationship between WMH severity and cognition. Our findings hold the potential to illuminate the neural underpinnings that link WMH severity to cognitive impairment, thereby advancing our understanding of this critical issue. Materials and methods Participants 140 CSVD patients diagnosed by the Neurology Department of the Fifth People’s Hospital of Shanghai, Fudan University were enrolled. The inclusion criteria were as follows: (1)patients older than 55 years old, (2)no history of brain trauma or dementia, (3)MRI scan showed WMH imaging manifestations. The exclusion criteria were as follows: (1)history of hydrocephalus, cerebral tumor or space occupying lesions, (2)unable to cooperate with this study independently or suffering from serious physical and mental diseases, (3)MRI contraindications, (4)leukodystrophy caused by other causes (such as multiple sclerosis). Finally 85 CSVD patients with WMH were enrolled, including 21 male and 64 female whose average age was 68. According to Fazekas rating scale [14], WMH in the patient’s periventricular and deep white matter were graded separately, and then added the grade together for total score 1–6. Previous studies demonstrated Fazekas scores 1 and 2 barely exhibited any differences in any cognitive domains compared to the performance of individuals without WMH. Besides, Fazekas score 3 was an important watershed point, the participants began to show significant impairments in both cognitive functions and white matter microstructure [15]. Thus, in order to explore the relationship between WMH severity and cognition, we divided 85 WMH patients into two groups: (1)mild WMH group scored 1–2 (n = 55); (2)severe WMH group scored 3–6 (n = 30). In addition, all subjects underwent demographic surveys, vascular risk factor examination, neuropsychological test (including Montreal Cognitive Assessment Scale (MoCA) [16] and Mini-mental State Examination (MMSE) [17]. All participants signed informed consent form in accordance with the Declaration of Helsinki. The protocol was approved by the Fifth People’s Hospital of Shanghai, Fudan University. Imaging analysis Magnetic resonance imaging acquisition Structural and rs-fMRI images were obtained using the 3.0 T Siemens Skyra MRI scanner (Siemens, Germany) with twenty-channel standard head coil. The sequence parameters were as follows: 3D T1-weighted with TR/TE/TI = 2400/2.13/1100ms, FOV = 256*256mm2, number of slices = 144. 3D T2-FLAIR with TR/TE/TI = 5000/395/1800ms, FOV = 256*256mm2, number of slices = 192. Rs-fMRI with TR/TE = 2000/15 ms, FOV = 384*384mm2, voxel size = 3.5*3.5*4.0mm3, slice thickness = 4 mm, number of slices = 36, 2 40 volumes acquired, scan duration = 8 min. Image preprocessing Rs-fMRI data preprocessing was performed using the Data Processing Assistant for Resting-State fMRI (Matlab2013b) based on Statistical Parametric Mapping (SPM12, http://www.fil.ion.ucl.ac.uk/spm) and Resting-State fMRI Data Analysis Toolkit (REST). The steps used for preprocessing showed following: (1)Converged the DICOM format of the original data to NIfTI format. (2)Removed the first 10 time points to eliminate the influence of magnetic field inhomogeneity. (3)Slice timing for eliminating the time difference between the interlayer scans. (4)Head-motion correction; participants with > 2 mm translation or > 2° rotation were excluded. (5)Nuisance regression: Friston-24 head-motion parameters, white-matter (WM) and cerebrospinal-fluid (CSF) signals, plus linear trend were regressed out. (6)Temporal band-pass filtering (0.01–0.1 Hz) to retain low-frequency fluctuations. (7)Normalized individuals onto the EPI brain template of the Montreal Neuroscience Institute (MNI) standard space and spatially resamples in voxels of 3 mm×3 mm×3 mm. (8)Used half-height full width (FWHM) to smooth the image to 6 mm Gaussian smoothing to improve the signal-to-noise ratio of the image. Brian functional networks determination In order to obtain the functional networks, we performed ICA after data preprocessing. Resting state data of all participants were analyzed using spatial group independent composition analysis (sGICA) as implemented in the GIFT software [18, 19] to decompose the data into functional networks that exhibited a unique time course profile. The steps included: (1)Data dimension reduction: using the main component analysis methods to isolate the 120 main components. (2)Independent component estimation: data were reduced into 100 independent component (ICs) with the information maximization algorithm (Informax) [20]. In the ICASSO toolbox, set up the Informax algorithm to run repeatedly 20 times to ensure the stability and reliability of the algorithm. (3)Back-reconstruction approach: subject-specific spatial maps and time courses were obtained by using the back-reconstruction approach [21]. (4)Pick ICs: using the Sorting GUI module to show ICs, selecting ingredients that were similar to the previous researches [22–25]. We confirmed if the peak activation coordinates were located primarily in grey matter, showing low spatial overlap with vascular, ventricular, edge regions corresponding to artefacts [26]. Finally, 21 meaningful ICs were selected from 100 ICs, these ICs were organized into 11 functional networks, including: visual network (VIN (IC1, IC6)), right frontoparietal network (r-FPN (IC7)), left frontoparietal network (l-FPN (IC11)), cerebellar network (CEN (IC4, IC8, IC23)), executive control network (ECN (IC13)), ventral attention network (VAN (IC15)), sensorimotor network (SMN (IC3, IC14, IC20, IC21, IC25)), salience network (SAN (IC24)), dorsal-default mode network (d-DMN (IC9, IC22)), ventral-default mode network (v-DMN (IC2, IC5, IC18)), dorsal attention network (DAN (IC12)). Dynamic functional connectivity analysis A sliding-window correlation approach and k-mean clustering were used to calculate DFC. Using GIFT’s Temporal dFNC module for DFC analysis, including steps: (1)According to the above functional networks to add the ICNs. (2)Using the sliding time window technique to intercept time series. Shirer W et al. showed that cognitive states can be identified within a 30 - to 60 - second window [23]. Also, Hutchison et al. found that with a 30 - second window, brief negative correlations, alternating high/low correlation cycles, and temporary new network nodes emerged, which longer windows didn’t reveal [27]. Therefore, we selected a 30 - second window length, which balances temporal resolution and connection estimation stability, offering enough time points for reliable connection strength calculation while capturing fine grained functional connectivity fluctuations, using a Gaussian window (σ = 3 TRs) and sliding in 1TR (2 s) steps. Pearson r was Fisher-z transformed within each window before clustering. (3)All the dynamic functional networks connectivity windows across all subjects were used to estimate the FC states by using k-means clustering algorithm, and the optimal number of clusters was calculated by elbow method to be 4, represented by 4 FC states (state1,state2, state3, state4). (4)We calculated the following temporal properties of DFC states in each participant: mean dwell time, fractional windows and transition numbers. Statistical analysis The SPSS 22 was used to perform statistical analysis. Demographic and vascular risk factors between the two groups were compared by chi square test and Mann-Whitney U test. For neuropsychological characteristics and DFC temporal properties, ANCOVA was used to test the group difference with age, sex, and vascular risk factors as covariates. Pearson correlation analysis was used to explore the relationship between DFC temporal properties and cognitive scores in WMH patients. At the same time, we carried out spearman correlation analysis to explore the relationship between the DFC temporal properties and Fazekas rating scale score. Results Demographic, vascular risk factors and neuropsychological characteristics analysis The participants’ demographic, vascular risk factors and neuroimaging characteristics were provided in Table 1. Notably, severe WMH patients were significantly older than that of mild WMH patients (Z=-2.814, p = 0.005). Apart from age, other demographic factor had no effect on diverse severity of WMH. There were no statistically significant differences in vascular risk factors between the two groups. For the neuropsychological scale, mild WMH patients had higher naming scores than severe WMH patients (F = 4.589, p = 0.035). The significant difference in naming scores suggests that severe WMH may have a specific impact on naming ability. No statistically significant difference was found in the overall cognitive score and other sub-cognition scores between the two groups. Table 1Demographic, vascular risk factors and neuropsychological characteristics analysis Demographic, vascular risk factors and neuropsychological characteristics analysis Dynamic functional connectivity states clustering Based on the sliding time window analysis and k-means clustering analysis, the resting state network was hierarchically organized into multiple network clusters, finally four dynamic functional connectivity states that recurred during the scanning period were identified (Fig. 1). State 1 was an intermediate pattern; state 2 was a more frequent, sparsely connected pattern; state 3 was also an intermediate pattern; state 4 was a less frequent, strongly connected pattern. Four DFC states across the whole WMH patients. The k-means clustering method identified four DFC states. State 1: intermediate pattern, State 2: a more frequent, sparsely connected pattern, State 3: intermediate pattern, State 4: a less frequent, strongly connected pattern. The number of functional connection windows and the corresponding percentage of each state were displayed. The redder color bar indicated stronger positive correlation function connection within and between brain networks, and the bluer color bar indicated stronger negative correlation function connection within and between brain networks. Abbreviations: VIN, Visual Network; r-FPN, right Frontoparietal Network; l-FPN, left Frontoparietal Network; CEN, Cerebellar Network; ECN, Executive Control Network; VAN, Ventral Attention Network; SMN, Sensorimotor Network; SAN, Salience Network; D-DMN, Dorsal Default Mode Network; V-DMN, Ventral Default Mode Network; DAN, Dorsal Attention Network DFC temporal properties comparison between mild and severe WMH subjects We compared the temporal properties (mean dwell time, fractional windows and transition numbers) of DFC states between the two groups. The difference of mean dwell time did not reach significance in states 2 and 4, a trend was observed, with severe WMH patients showing relatively longer mean dwell time in hypoconnected state 2 and relatively shorter mean dwell time in hyperconnected state 4 than mild WMH patients (Fig. 2). No significant differences were found in fractional windows and transition numbers. Differences in the mean dwell time between the two groups. Compared to the mild WMH group, the severe WMH patients dwelled relatively longer in hypoconnected state 2 and relatively shorter in hyperconnected state 4. Abbreviations: s, second; S, sever WMH group; M, mild WMH group Correlation analysis Correlation between DFC temporal properties and cognitive function In the full cohort, mean dwell time and fractional window in hypoconnected state 2 showed negative associations with executive function scores (r = − 0.308, p = 0.002; r = − 0.361, p < 0.001), whereas the corresponding metrics in hyperconnected state 4 were positively associated (r = 0.183, p = 0.047; r = 0.211, p = 0.026) (Table 2). Moreover, mean dwell time and fractional windows in hypoconnected state 2 were negatively correlated with executive function scores in mild WMH subjects (r=-0.323, p = 0.008; r=-0.341, p = 0.005); fractional windows in hypoconnected state 2 was negatively correlated with executive function scores in severe WMH patients (r=-0.379, p = 0.020). After FDR correction, none of these correlations survived q < 0.05; they are therefore presented as exploratory trends requiring independent validation. There were no statistically significant differences in hyperconnected state 4 found in mean dwell time, fractional windows and executive function scores as shown in (Fig. 3a, b). Table 2Correlation between DFC temporal properties and executive function across the whole WMH patients Correlation between DFC temporal properties and executive function across the whole WMH patients Correlations between the DFC temporal properties and cognitive scores of the mild and severe WMH patients respectively. (a) The mean dwell time and fractional windows in hypoconnected state 2 was negatively correlated with executive function scores in mild WMH subjects (r=-0.323, P = 0.008; r=-0.341, P = 0.005). (b) The fractional windows in hypoconnected state 2 was negatively correlated with executive function scores in severe WMH patients (r=-0.379, P = 0.020). There were no statistically significant differences in hyperconnected state 4 found in mean dwell time, fractional windows and executive function scores Correlation between transition numbers and fazekas rating scale Transition numbers were negatively correlated with the Fazekas rating scores in the whole WMH subjects (r=-0.218, p = 0.046). Discussion WMH is the most common imaging marker of CSVD. At present, the precise impact of WMH severity on the intrinsic functional activity of the brain remains unclear. In our study, we employed DFC analysis to explore the effects of WMH severity on brain functional activity. The main findings were as follows. (1) Severity of WMH was mainly influenced by age, the subjects with older age showed severer WMH. (2) Severity of WMH may affect the naming function of WMH patients. (3) WMH severity impacted DFC temporal properties. Specifically, subjects with severer WMH spent relative more time in hypoconnected state, less time in hyperconnected state and exhibited fewer transition numbers between states compared to those with milder WMH. (4) Altered DFC temporal properties were associated with executive function among WMH participants, manifesting that an increased duration of sparse connectivity state and a reduced duration of strongly connected state were associated with poorer executive performance; however, these associations did not survive FDR correction and should be regarded as exploratory signals requiring independent validation. Our results provided preliminary insights into underlying neural mechanisms of WMH severity. Age had consistently been identified as a key factor associated with WMH severity in numerous studies. In a cross-sectional analysis of 655 WMH patients aged ≥ 60 years, Zheng et al. categorized WMH severity into mild, moderate and severe groups according to Fazekas rating scores, and found that patients’ age significantly increased along with higher WMH burden [28]. Another study including 116 WMH participants who underwent brain MRI discovered that older age participants were associated with severe WMH by rating their WMH lesions visible on MRI [7]. Consistent with these prior studies, our results implied that aging processed strongly influenced WMH progression. Moreover, we found that mild WMH subjects had higher naming scores than severe WMH subjects, suggesting that severe WMH may have a specific impact on naming ability. While, Yuan Chen et al. demonstrated that WMH is associated with impairments across multiple cognitive domains, such as naming, executive function, and language [29]. The observation that naming ability is relatively preserved in patients with mild WMH offers a more refined viewpoint on the interplay between WMH and cognitive function. This outcome implies that the effect of WMH on different cognitive domains is heterogeneous and may be contingent upon the severity of WMH. Based on DFC analysis, we employed k-means clustering analysis to categorize the intrinsic brain functional connectivity patterns among the entire cohort of WMH subjects. Depending on the varying strengths of the intrinsic brain functional connectivity, four distinct connectivity states of all subjects were identified a sparsely connected state (state 2), a strongly connected state (state 4) and the other two intermediate states. Several studies have proposed that sparsely connected state may indicate a widespread disruption of the whole brain network [30]. Other research suggested sparsely connected state may be related to low brain energy consumption [31]. Our study uncovered a noteworthy finding among patients with severe WMH, namely a prolonged mean dwell time in sparsely connected state, which correlated with worsened executive function before correction for multiple comparisons. Executive function encompasses the integrative mental and behavioral skills required for goal-directed behavior, reasoning, planning and synthesis, and constitutes the characteristic cognitive deficit profile associated with WMH [32]. The neural substrates underlying multi-dimensional executive functions are widely distributed and require coordinated integration across the whole brain. In addition, WMH are principally linked to neurofibrillary degeneration, demyelination, glial cell death, gliosis and axonal loss, which are driven by chronic ischemia, blood–brain barrier disruption and small vessel arteriosclerosis [33]. Although our research results can only be regarded as an exploratory indication, we postulate that these pathological changes underlie the brain network disorganization manifested as the sparsely connected state, impeding cerebral information transmission and consequently giving rise to executive dysfunction in patients with WMH. At present, there were few studies applied DFC analysis on WMH severity, but several studies on other neurological diseases. For instance, Fiorenzato et al. examined DFC in 118 patients with Parkinson’s disease, categorized into groups with normal cognition (n = 52), mild cognitive impairment (n = 46), and dementia (n = 20). Their findings indicated that patients with Parkinson’s dementia spent significantly longer periods in sparsely connected state compared to those with normal cognition and mild cognitive impairment [13]. In another study involving 19 individuals with subcortical ischemic vascular disease and 38 healthy controls who underwent DFC analysis, it revealed that subcortical ischemic vascular disease patients exhibited more extended periods of sparse connectivity states in comparison to the control [12]. These findings collectively suggested that prolonged periods in sparsely connected state may be associated with cognitive impairment. Besides, several studies have suggested that the strongly connected state plays a role in cognitive processing; a decreased mean dwell time in this state indicates that a key connectivity pattern has been disrupted during spontaneous neural activity [34]. We found that individuals with severe WMH spent less time in the hyperconnected state, and, at the uncorrected threshold, this reduction trended toward poorer executive performance. As an exploratory trend, we propose that shortening the strongly connected state could lower information-conveyance efficiency, potentially contributing to executive dysfunction in severe WMH. Moreover, DFC studies of other neurological disorders report similar reductions: Lewy body dementia patients [35] and subcortical ischemic vascular disease (SIVD) patients with amnestic mild cognitive impairment [36] both exhibited significantly shorter dwell times in hyperconnected states. Taken together with our preliminary observation, these results raise the hypothesis that WMH related structural damage perturbs the balance between global integration and local segregation. Firm evidence for a causal link, however, awaits replication in larger, multicenter cohorts before therapeutic modulation of dwell times can be advocated. In addition, WMH severity was found to exert an influence on transitions numbers. Transitions between states were thought to be a reflection of neuronal flexibility, which allowed different brain regions to flexibly engage without being locked into fixed interaction patterns [37]. Notably, Hellyer et al. put forth the notion that more frequent transitions between states were predictive of enhanced cognitive performance [38]. In the current work, we observed that WMH subjects with higher Fazekas rating scores exhibited fewer transitions between states. Consequently, we speculated that reduced number of transitions might contribute to the decline in cognitive function observed in individuals with more severe WMH. Intriguingly, previous research did not yield a significant association between the total load score of CSVD and transition numbers [39]. The possible reason for the discrepancy in findings between our study and prior research could potentially be attributed to our specific focus on WMH. Unlike the complex changes in brain functional activity seen in the multifaceted imaging manifestations of CSVD, WMH characteristics remained distinct and less obscured, possibly accounting for the observed differences in results. Several limitations should be considered when interpreting our results. Firstly, the cross - sectional design of our study limits our ability to draw definitive conclusions about the causal relationship and developmental trajectory between WMH severity and dynamic functional connectivity alterations, as well as their impacts on cognitive functions. Longitudinal studies are needed to track changes in WMH and brain functional activity over time. Secondly, our study used a 30 - second window length for dynamic functional connectivity analysis. Future studies should incorporate multiple window lengths to further test the robustness of our findings and to provide a more comprehensive understanding of dynamic functional connectivity patterns associated with white matter hyperintensities. Thirdly, while DFC analysis offers valuable insights into dynamic brain functional activity, it has certain limitations. DFC analysis primarily reflects the dynamic changes in brain activity at the macroscopic level and cannot directly reveal the underlying neurophysiological mechanisms or the specific neurotransmitters involved. Finally, the sample size was modest for detecting small-to-moderate effect sizes, especially after FDR correction for multiple comparisons; our findings therefore require validation in larger, multi-center cohorts. Conclusion In summary, our research explored the intricate relationship between WMH severity and intrinsic brain functional activity. Using DFC analysis, we discovered that increased WMH severity was linked to altered functional connectivity patterns, including prolonged durations in hypoconnected states, reduced time in hyperconnected states, and fewer state transitions. Exploratory trends hinted at a possible link between these altered temporal metrics and poorer executive performance, offering a potential mechanism whereby disrupted neural network dynamics could contribute to cognitive deficits; however, the associations did not reach statistical significance after adjustment for multiple comparisons and remain to be validated in larger, multicenter cohorts. These preliminary observations suggest that DFC-derived metrics, once replicated, could aid in understanding WMH related cognitive impairment and inform future intervention strategies. Acknowledgements Not applicable. Abbreviations Alzheimer’s disease Cerebellar network Cerebral small vessel disease Dorsal attention network Dynamic functional connectivity Dorsal-default mode network Executive control network Functional connections Gray matter volume High-density lipoprotein Independent component analysis Independent components Intrinsic connectivity networks Low-density lipoprotein Left frontoparietal network Mini-Mental State Examination Montreal Cognitive Assessment Right frontoparietal network Resting state functional magnetic resonance imaging Salience network Spatial group independent composition analysis Subcortical ischemic vascular disease Sensorimotor network Total cholesterol Triglycerides Ventral attention network Ventral-default mode network Visual network White matter hyperintensity Author contributions Conception and design were contributed Xiaohu Zhao and Chunxiao Wei. Data acquisition were contributed by Tianyuyi Feng. Data analysis and interpretation was contributed by Tianyuyi Feng and Yunfei LI. Drafting of the manuscript was contributed by Tianyuyi Feng and Chunxiao Wei. Critical revision of the manuscript was contributed by Xiaohu Zhao. Funding This study was funded by the Science and Technology Innovation Plan of Shanghai Science and Technology Commission [grant 22Y11910800]; the Minhang District Natural Science Foundation of Shanghai [No.2024MHZ024]; Shanghai Minhang District Health and Family Planning Commission [great discipline of Shanghai Minhang District No. 2024MWDXK03]. Data availability The datasets in the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The research adhered to the Declaration of Helsinki and its latest amendments. It was approved by the Medical Ethics Committee of the Fifth People’s Hospital of Shanghai, Fudan University, and the approval number is (2018) No. (172). Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Chunxiao Wei, Email: wcx123zhj@163.com. Xiaohu Zhao, Email: xhzhao999@263.net. References Associated Data Data Availability Statement The datasets in the current study are available from the corresponding author on reasonable request.