Brain functional alterations associated with visuospatial working memory impairment in gouty arthritis patients

Introduction

Gouty arthritis (GA) is a disease caused by the deposition of monosodium urate crystals in articular and non-articular structures, typically presenting as acute arthritis accompanied by severe pain (1). The physiological basis of GA is hyperuricemia (HUA), and there exists a profound dependency between serum uric acid (UA) levels and the risk of developing GA. Compared to asymptomatic HUA, GA is accompanied by local and systemic inflammatory responses caused by monosodium urate crystal deposition (2). These inflammatory mediators can cross the blood-brain barrier and activate microglia, potentially affecting neurological function through multiple pathways including oxidative stress, neuronal damage, and disruption of synaptic plasticity (3,4). Current research has demonstrated that the production of UA can induce oxidative stress, causing damage to neuronal cells and vascular endothelium (5). Similarly, studies utilizing cognitive function behavioral scales have found that elevated serum UA levels are independently associated with cognitive decline in patients (6). GA represents the clinical endpoint and important complication of HUA, with the number of patients showing an annually increasing trend. Therefore, neuroimaging studies on GA patients are beneficial for further exploring the complex relationship between GA and cognitive impairment.

Functional magnetic resonance imaging (fMRI) can effectively capture spontaneous neural activity and subtle changes in brain function (7). Among them, fMRI based on blood oxygenation level-dependent (BOLD) imaging is a more extensive one used to observe the functional activities of the brain, helping researchers to determine the activities of specific brain regions under different tasks or stimuli. Recent resting-state fMRI (rs-fMRI) studies have identified cognitive impairments in HUA patients across disease progression (8). Visuospatial working memory (VSWM), as an important cognitive function, has the ability to temporarily store and manipulate visuospatial information and plays a critical role in environmental navigation and spatial reasoning (9,10). The mental rotation task (MRT) is a classic experimental paradigm used to assess VSWM, which can be conducted simultaneously with MRI scans to observe the mental rotation manifestations and underlying neural correlation of the participants (11). Task-state fMRI (ts-fMRI) focuses on brain activity under specific task conditions and activates specific functional areas through the execution of tasks. rs-fMRI focuses on the spontaneous neural activities of the brain in a relaxed state, studying the functional connections and network patterns of the brain. Therefore, the combination of rs- and ts-fMRI provides different perspectives on brain function, thereby better revealing the characteristics of brain activity and cognitive function connection patterns in different aspects under GA disease.

In recent years, the integration of machine learning (ML) into the field of neuroimaging has transformed our ability to analyze complex brain data and uncover patterns related to cognitive function (12). ML algorithms, particularly in the realm of supervised learning, have been applied to classify and predict cognitive states based on neuroimaging data (13). In classification and prediction tasks, algorithms such as support vector machine (SVM) and linear regression are widely favored due to their excellent performance. Researchers utilize these algorithms to construct predictive models that can accurately identify patients’ cognitive states, such as working memory and attention, based on brain imaging data. This provides a new research perspective for exploring the neural mechanisms underlying cognitive functions (14).

In this study, we extracted the resting-state whole-brain functional connectivity of GA patients and healthy controls (HC) based on regions of interest (ROIs) as features, trained a better-performing classification model using supervised learning, and analyzed the brain functional regions that significantly influence the classification accuracy. Simultaneously, we employed ts-fMRI using the MRT paradigm to explore abnormal brain activation in GA patients during MRT execution, aiming to provide neuroimaging evidence for elucidating the neural mechanisms of cognitive impairment in GA. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-81/rc) (15).

Methods

Participants

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Ethics Committee of the First Affiliated Hospital of Shantou University Medical College (No. B-2021-234). All the participants provided written informed consent.

Participants were recruited into two distinct cohorts: individuals diagnosed with GA and HC. The data collection period spanned from April 2021 through June 2024. The GA patients were recruited from the outpatient and inpatient departments of the First Affiliated Hospital of Shantou University Medical College, who were in the acute attack phase of GA on the day of examination and had not been on stable long-term oral UA-lowering drugs. Patient enrollment in this investigation required adherence to specific inclusion criteria (Figure 1): (I) confirmed GA diagnosis according to the 2015 classification guidelines established collaboratively by the European League Against Rheumatism and the American College of Rheumatology (EULAR/ACR) (16); (II) age range of 18–50 years, minimum educational attainment of six years, right-hand dominance, and normal or corrected visual acuity; and (III) have complete clinical examination, neuropsychological scale assessment data, and MRI data available. Patients were excluded from the study if they met any of the following exclusion criteria: (I) coexistence of other chronic diseases such as hypertension, hyperlipidemia, or hyperglycemia; (II) previous central nervous system disorders or conditions potentially impacting cognitive function; (III) prior substance abuse or alcohol dependency; and/or (IV) contraindications for MRI scanning. As GA predominantly affects males in clinical practice, and no female patients meeting the inclusion and exclusion criteria were collected during the study period, all GA patients included in this study were male. A total of 25 male GA patients were initially recruited for this investigation. Exclusions occurred in four cases: three participants had inadequate MRI datasets owing to scan cooperation issues, whereas one participant exhibited motion artifacts exceeding the 2° threshold during fMRI acquisition. The remaining 21 GA patients constituted the final study sample. Figure 1 provides a comprehensive flowchart detailing the inclusion and exclusion procedures. The HC group consisted of 21 HCs matched for age, gender, and education level.

Figure 1 Flow diagram of patients included and excluded in the study. fMRI, functional magnetic resonance imaging; GA, gouty arthritis; MRI, magnetic resonance imaging.

Questionnaires

The Montreal Cognitive Assessment (MoCA) is a cognitive screening instrument developed by Nasreddine et al. in 2005 (17), which has been widely utilized for screening mild cognitive impairment and provides a comprehensive evaluation of multiple cognitive domains. In clinical practice, scores below 26 indicate the need for further assessment to determine the potential presence of cognitive dysfunction. The Chinese adaptation of the MoCA scale, previously translated and adapted by Peking Union Medical College Hospital (18), was administered to all study participants. Testing occurred at the same venue and followed identical temporal protocols.

MRI examination

The imaging protocol employed a Signa HDxt 1.5T superconducting MRI scanner (GE Healthcare, Chicago, IL, USA). Scanning procedures were executed by certified MRI operators at the First Affiliated Hospital of Shantou University Medical College, with radiological assessment by specialist physicians to eliminate intracranial organic pathology. Anatomical imaging utilized three-dimensional (3D) fast gradient-echo T1-weighted imaging (T1WI) for high-resolution structural data acquisition. The imaging parameters for T1WI were as follows: echo time (TE) =1.6 ms, repetition time (TR) =5.1 ms, field of view (FOV) =256 mm × 256 mm, flip angle (FA) =10°, slices =244, and slice thickness =1.4 mm. The fMRI data were obtained using gradient echo-echo planar imaging (19), which effectively improves the sensitivity and temporal resolution of BOLD contrast. The imaging parameters for fMRI were as follows: TE =45 ms, TR =3,000 ms, FOV =250 mm × 200 mm, FA =90°, slices =20, and slice thickness =6 mm.

Experimental paradigm and presentation of MRT

The MRT paradigm adapted from a previous study (20), which was programmed and presented using E-Prime 3.0 software (Psychology Software Tools Inc; PST, Pittsburgh, PA, USA). In the MRT paradigm task, we employed 3D models designed by Shepard and Metzler (11), with the 3D images rotated at angles of 0° and 100°. We paired these images into sets, which included two conditions: rotated-overlapping and mirror-imaged. In the ts-fMRI scan, there were eight alternating blocks consisting of two rotation angles (0° and 100°) for paired 3D object judgment. In each block, after fixating on a cross for 0.5 seconds, participants were required to determine whether a pair of randomly presented 3D objects were identical or mirror images of each other and respond by pressing keys within 6.5 seconds. Each rotation angle block contained 12 trials and lasted for 84 seconds. To prevent visual fatigue, a 30-second rest period was inserted between every two blocks. The entire experimental paradigm presented a total of 96 pairs of rotated and non-rotated cube images, with a cumulative duration of 774 seconds. The overall process of MRT is shown in Figure 2. Prior to the fMRI scan, all participants received training and testing on the MRT. During the fMRI scan, the task paradigm was synchronized and presented through an audio-visual presentation system (SA-9939, Shenzhen Sinorad Medical Electronics Co., Ltd., Shenzhen, China). Participants’ reaction times and accuracy were recorded using E-prime software for subsequent analysis.

Figure 2 Schematic diagram of the mental rotation task experimental procedure.

Data processing

Processing of rs-fMRI datasets was conducted using MATLAB software (MathWorks, Inc., Natick, MA, USA) in conjunction with the GRETNA analysis package (21) (version 1.2.1) (http://www.nitrc.org/projects/gretna/). All the selected MRI scans were exported in Digital Imaging and Communications in Medicine (DICOM) format and then converted to Neuroimaging Informatics Technology Initiative format (NIfTI). Subsequently, image preprocessing involved sequential correction steps: (I) slice timing, (II) realignment, (III) normalization, (IV) spatial smoothing, (V) temporal detrending, (VI) regression, and (VII) filtering. Following these preprocessing procedures, brain parcellation was performed using the automated anatomical labeling (AAL) atlas (22) to define 90 distinct ROIs, enabling extraction of mean BOLD signal time courses from each ROI. Pearson correlation analysis was then conducted pairwise between the ROIs. Functional connectivity (FC) data were subjected to Fisher-z transformation to enhance distributional normality.

The ts-fMRI data were analyzed using the Analysis of Functional NeuroImages software package (http://afni.nimh.nih.gov/afni/), which is designed for preprocessing, analyzing, and displaying fMRI data. The preprocessing steps included slice timing correction, head motion correction, spatial smoothing, and spatial normalization to the standard coordinates of the Talairach and Tournoux atlas. A general linear model with six motion covariates was used to model task-related effects in the first-level individual analysis. The fixation conditions were treated as the baseline, and two task-related regressors were deconvolved with the canonical hemodynamic response function. The preprocessed ts-fMRI data were analyzed using the analysis of variance (ANOVA) method to identify brain regions exhibiting activation differences between the GA group and the HC group during the execution of the MRT. In ts-fMRI data validation, only when the threshold is P<0.05 and the voxel value of the activated brain area is greater than 20, will it be included in the effective activation results.

ML

ML classification of GA patients versus HC was performed using MATLAB-based algorithms, including linear discriminant analysis (LDA), logistic regression (LR), K-nearest neighbors (KNN), and SVM approaches. These supervised learning methods establish input-output relationships to enable predictive classification of unlabeled datasets. The SVM implementation specifically utilized the LIBSVM computational library (23) for model training and validation. Feature extraction involved deriving functional connectivity matrices encompassing all ROIs for dataset construction. Data normalization procedures ensured feature consistency, whereas leave-one-out cross-validation methodology optimized model hyperparameters. Additionally, we employed F-score for feature filtering to enhance the generalization ability of the models.

A single subset served as the validation dataset in each iteration, with remaining subsets designated for model training. This methodology minimizes overfitting risks in small sample datasets while providing robust performance estimates across various data partitions. Additionally, it guarantees that all data points contribute to both training and validation phases, thereby strengthening evaluation reliability. Model classification performance was assessed using accuracy, sensitivity, specificity, and area under the curve (AUC) metrics. The DeLong statistical test (24) enabled AUC comparisons across different algorithms to identify the superior performing model. Subsequently, permutation analysis validated whether model performance exceeded chance levels. This procedure involved randomly shuffling dataset labels 1,000 times and retraining models with permuted data. Trained model performance was then benchmarked against the distribution derived from permuted datasets. Significantly superior performance compared to randomly permuted models indicated genuine pattern recognition rather than stochastic effects. Finally, feature weight analysis identified neuroimaging regions with substantial contributions to model architecture. The brain areas predominantly involved in the classification model’s weight were visualized using the BrainNet Viewer software (25) (www.nitrc.org/projects/bnv).

Statistical analysis

The data of this study were analyzed using statistical software SPSS 26.0 (IBM Corp., Armonk, NY, USA). The general information (age, years of education), clinical data, and behavioral data (MoCA scores, MRT reaction times, and accuracy rates) of the GA patients and HC groups were analyzed as quantitative data. For variables following a normal distribution, the two independent samples t-test was employed. Non-parametric Mann-Whitney U testing was applied for variables lacking normal distribution characteristics. Continuous data were presented as mean ± standard deviation values. Two-tailed statistical testing was employed throughout, with P<0.05 establishing the threshold for statistical significance.

Results

Demographic, clinical data, and neuropsychological characteristics

The demographics, clinical, and cognitive neuropsychological characteristics of the GA patients and HC are shown in Table 1. No significant differences were found between the two groups in terms of age, education level, glutamic oxaloacetic transaminase, glutamic pyruvic transaminase, creatinine, and hemoglobin (P>0.05). Among the clinical indicators, serum UA levels in the GA group were significantly higher than the normal range and markedly elevated compared to the HC group (P<0.001). The average total score of MoCA in the GA patient group was 23.24±1.79 (mean ± standard deviation), which was significantly lower than that of the HC group (P<0.001). For the subscale scores of MoCA, the GA group exhibited significantly lower scores in Visuospatial/Executive function, Language, Abstraction, and Delayed Recall compared to the HC group (P<0.05).

Performance of ML based on whole-brain resting-state functional connectivity features

To evaluate the performance of different classification models, we plotted receiver operating characteristic (ROC) curves, with their respective AUC compared using the DeLong test (Figure 3). The classification accuracy of the LR, LDA, KNN, and SVM supervised learning models was 47.22%, 52.78%, 58.33%, and 77.78%, respectively. The corresponding AUC values were 0.4522 [95% confidence interval (CI): 0.2620–0.6423], 0.5062 (95% CI: 0.3593–0.7395), 0.5386 (95% CI: 0.2932–0.6753), and 0.7685 (95% CI: 0.6122–0.9248). The results demonstrated that the SVM model exhibited significantly higher classification accuracy and performance compared to other models (P<0.05). The classification accuracy of the SVM model was validated through permutation testing, showing significantly higher accuracy than models with randomly shuffled labels (P<0.01). We extracted the weight coefficient matrix of the support vectors with the highest classification accuracy from the constructed SVM model (Figure 4). The results revealed that FC involving nodes in the cuneus, superior occipital gyrus (SOG), inferior parietal lobule (IPL), superior parietal gyrus (SPG), and middle frontal gyrus (MFG) appeared frequently in the weight coefficient matrix, indicating their significant involvement in the model’s decision-making process. These key brain regions are visualized in Figure 5.

Figure 3 Analysis of receiver operating characteristic curves for different classification models. *, P<0.05. AUC, area under the curve; CI, confidence interval; KNN, K-nearest neighbor; LDA, linear discriminant analysis; LR, logistic regression; SVM, support vector machine.

Figure 4 Support vector weight coefficient matrix obtained in SVM model. The element grid tending towards red represents the functional connectivity with higher positive weights, while the element grid tending towards blue represents the functional connectivity with lower negative weights. AAL, automated anatomical labeling; ROIs, regions of interest; SVM, support vector machine.

Figure 5 Brain regions with significant influence on model construction in the weight matrix.

Behavioral performance of MRT and ts-fMRI results

The behavioral test results of the MRT for GA patients and HC are presented in Table 2. The results showed that compared to HC group, the GA patients performed worse in both the 0° control block task and the 100° rotation task. Their accuracy in making similarity judgments was significantly lower than the HC group, whereas their reaction times were significantly higher. The concurrent ts-fMRI analysis revealed significant activation differences in the group main effect, specifically in the right superior temporal gyrus, bilateral lingual gyrus, left anterior cingulate cortex, right middle temporal gyrus, bilateral cingulate cortex, and right precentral gyrus (Figure 6A, Table 3). For the main effect of task, we observed more activation differences in the bilateral cingulate cortex, cuneus, IPL, precuneus, MFG, and right middle temporal gyrus (Figure 6B, Table 3). The group × task interaction effect showed significant differences in the left precuneus, right posterior cingulate cortex, left IPL, left insula, left MFG, and right MFG (Figure 6C, Table 3). Finally, the simple effects analysis between groups revealed that during the 100° rotation task, GA patients exhibited reduced BOLD activation intensity in the left IPL compared to the HC group, while showing increased activity in the right parahippocampal gyrus (Figure 6D, Table 4).

Figure 6 Differences in task-state fMRI data acquired during the mental rotation task between GA patients and HC. (A) Main effect of group; (B) main effect of task; (C) interaction effect between group and task; (D) differences between GA patients and healthy subjects during the execution of the 100° rotation task. fMRI, functional magnetic resonance imaging; GA, gouty arthritis; HC, healthy controls; L, left; R, right.

Discussion

With the global economic development, GA has emerged as one of the most prevalent inflammatory arthropathies, with increasing incidence and prevalence rates worldwide, particularly in developed countries (26). Studies have suggested that GA patients may face an elevated risk of cognitive impairment and dementia (2,27). In this study, we initially employed the MoCA scale to evaluate the overall cognitive function of GA patients. Our analysis revealed that GA patients may exhibit mild cognitive impairment, which aligns with previous findings (6,28,29). Gout can induce increased secretion of pro-inflammatory cytokines such as interleukin (IL)-1β and IL-6, which may lead to cognitive impairment through mechanisms affecting synaptic plasticity and neurotransmitter function (30). To further elucidate the neuroimaging characteristics of cognitive function-related brain regions in GA patients, this study utilized rs-fMRI and ts-fMRI combined with MRT. We conducted a multifaceted exploration of the behavioral performance and brain activation changes in VSWM among GA patients. Additionally, we employed common small-sample ML models in an attempt to effectively differentiate GA patients from HC from a neuroimaging perspective.

Rs-fMRI is a neuroimaging technique used to study spontaneous neural activity in the brain. Analyzing FC in individuals during a resting state helps to elucidate activity patterns between static brain networks. Whole-brain FC not only reflects synchronicity between different functional brain regions but also provides abundant neuroimaging features for supervised learning (31). To date, no studies have employed ML methods to investigate the neuroimaging characteristics of GA patients. In the present study, we extracted whole-brain resting-state FC matrices for all participants based on ROIs and conducted classification training using common small-sample ML models. The results indicate that SVM is currently the optimal model for distinguishing GA patients from HC using whole-brain resting-state FC, achieving an accuracy of 77.78%, which was validated through permutation testing. SVM, a widely used ML algorithm for classification and regression analysis, has demonstrated excellent classification performance and generalization ability in recent years, further enhancing the efficiency of neuroimaging data analysis (32,33). Therefore, we posit that ML models optimized through hyperparameter tuning can aid in early screening and warning of cognitive function changes in GA patients. Furthermore, by utilizing SVM for ML classification, whole-brain static FC shows potential as a future neuroimaging biomarker.

After identifying the SVM model with superior classification performance, we extracted the weight coefficient matrix from the model based on the decision function to understand the contribution of each feature to the classification decision. Our results indicate that FC with nodes in the cuneus, SOG, IPL, SPG, and MFG significantly contributed to the model’s classification decision. These regions are likely to be areas of the brain where differences exist between GA patients and HC. From a functional perspective, the cuneus and SOG are primarily involved in early visual information processing and spatial feature encoding (34,35). Current research suggests that the cuneus exhibits enhanced activity during visual-spatial attention and MRTs, whereas the SOG plays a crucial role in spatial location encoding and visual search (36). Activation of the cuneus and SOG typically occurs during the encoding phase of VSWM tasks (37). Parietal regions play important roles in spatial attention and information manipulation (38). Within this area, the IPL is involved in the integration and maintenance of spatial information, particularly in tasks requiring attentional shifts (39), whereas the SPG is more involved in processing spatial relationships (40). Current research indicates that forebrain cortical regions demonstrate particular sensitivity to purine metabolism abnormalities during development, with the frontal cortex exhibiting greater dependence on the de novo purine synthesis pathway (41). Consequently, elevated UA levels may contribute to frontal lobe dysfunction in patients. Furthermore, as part of the prefrontal cortex, the MFG plays a central role in executive control and information maintenance in VSWM (42). The coordinated activity of these brain regions is crucial for successful completion of VSWM tasks. Our findings suggest that GA patients may differ from HC in this aspect of cognitive function.

To further elucidate abnormal activation in brain regions associated with VSWM function in GA patients, this study employed ts-fMRI techniques based on the MRT for subsequent analysis. The behavioral results of the MRT in this study demonstrated that GA patients performed poorly compared to HC when completing the MRT, indicating the presence of VSWM impairment in GA patients. This finding aligns with previous research suggesting that chronic pain can lead to impairment of multiple cognitive functions, including working memory (43). During the participants’ execution of the MRT, simultaneous collection of BOLD-fMRI signals allowed observation of brain region activation. These regions are neurologically associated with VSWM, a notion consistent with previous studies (44,45). Furthermore, this study found that compared to the HC group, GA patients exhibited significantly reduced activation in the IPL when completing the 100° rotation task. Previous research has shown that chronic pain patients demonstrate weakened IPL activation during working memory tasks such as spatial N-back, with activation intensity negatively correlated with pain severity (46). Pain may directly interfere with spatial information processing in the IPL, leading to decreased VSWM function. Moreover, the IPL is closely connected with regions such as the dorsolateral prefrontal cortex (DLPFC) in the fronto-parietal network, jointly participating in the maintenance and manipulation of VSWM (47). Therefore, reduced IPL activation may indirectly affect the function of frontal lobe regions such as the DLPFC, further exacerbating VSWM impairment. Additionally, the activity of GABA-ergic inhibitory interneurons in the parietal cortex is crucial for maintaining VSWM function (48). Thus, reduced IPL activation may indicate impaired inhibitory neuron function, making it difficult to suppress interference from irrelevant information, and ultimately resulting in poor MRT performance in GA patients.

The hippocampus plays a crucial role in the long-term storage and retrieval of spatial information, particularly in VSWM tasks that require the integration of new and old information (49). It also plays a significant role in visual working memory encoding, maintaining bidirectional connections with parietal regions (50). The compensatory hyperactivation of the hippocampus in GA patients during MRT execution reflects the functional decline of parietal regions, leading to inefficient encoding. Concurrently, inflammatory damage to the hippocampus itself may further exacerbate working memory dysfunction. Notably, the hippocampus is one of the most active brain regions for cellular renewal, with its neurogenesis process being highly sensitive to inflammatory factors. Pro-inflammatory cytokines can inhibit neural stem cell proliferation, differentiation, and synaptic plasticity, impairing hippocampus-dependent learning and memory (51). Additionally, research suggests that chronic pain can induce structural and functional changes in the hippocampus, such as volume reduction and weakened long-term potentiation, potentially affecting memory encoding efficiency (52). In conclusion, pain and inflammatory factor release induced by GA may influence VSWM processing at different stages of memory formation, maintenance, and retrieval through multiple mechanisms.

Although some studies suggest that UA possesses antioxidant properties and may protect the nervous system under certain circumstances (53,54), these findings appear to contrast with our current results. This apparent contradiction likely reflects the dual role of UA in the nervous system. Although UA may exert protective effects at physiological concentrations, the persistent hyperuricemic state in GA patients may impair cognitive function through multiple mechanisms. First, elevated UA levels can promote oxidative stress, endothelial dysfunction, and microvascular remodeling, thereby increasing the risk of cerebral microvascular disease (5,28). Second, gout-associated chronic inflammatory states may directly affect neuronal function and synaptic plasticity through inflammatory mediators [such as IL-1β, IL-6, and tumor necrosis factor (TNF)-α] that penetrate the blood-brain barrier (4). Recent research has introduced the emerging concept of rheumatic cognitive impairment (55), which characterizes the patterns of cognitive dysfunction observed across various rheumatic diseases. Although GA is primarily classified as a metabolic disorder from a pathophysiological perspective, our findings suggest that the cognitive impairment patterns observed in GA patients may share certain similarities with those seen in rheumatic diseases (20). This convergence may be attributed to common inflammatory mediator pathways that underlie both conditions. Based on the integrated analysis of our research findings and existing literature, we propose that cognitive impairment in GA patients likely results from the synergistic effect of HUA and recurrent inflammatory episodes. Our findings reveal VSWN damage in GA patients from a neuroimaging perspective, suggesting that clinical practice requires more individualized assessment of cognitive status in GA patients, and consideration of early interventions such as cognitive training and strict UA control to potentially improve patients’ cognitive prognosis and quality of life.

This study has several limitations. Firstly, due to the higher likelihood of comorbidities in female gout patients, only male patients were included. However, gender remains a factor to consider, and future studies will aim to recruit more female GA patients meeting the inclusion criteria. Secondly, given the relatively small sample size, only models suitable for small samples were used for ML, which may have led to overfitting. Additionally, we were unable to further explore the relationship between pain severity, disease duration, and cognitive function. Subsequent studies will recruit larger patient cohorts and implement comprehensive pain assessment scales, including a subgroup of asymptomatic HUA cases, to investigate the complex relationships between various influencing factors and cognitive impairment in GA patients. After collecting a sufficient sample size, we intend to analyze the temporal relationship between changes in inflammatory markers, UA levels, and cognitive performance to better understand the causal mechanisms involved. Furthermore, we acknowledge the potential limitations of using a 1.5 T MRI scanner. However, we believe that our imaging was stable and produced meaningful results. Future studies could utilize fMRI sequences with thinner scan layers and more precise atlases for segmentation, enabling more detailed functional brain analysis in GA patients.

Conclusions

This study comprehensively revealed VSWM impairment in GA patients and preliminarily elucidated its neuroimaging differences using multimodal fMRI techniques and ML methods, combined with behavioral assessments. We identified SVM as the optimal ML model for effective classification of GA patients and HC based on whole-brain resting-state FC features. Moreover, we found that cognitive dysfunction caused by GA may be associated with activation differences in VSWM-related brain regions, including the fronto-parietal network and hippocampus. This study provides a promising direction for future clinical diagnosis and personalized treatment of cognitive impairment in GA patients. Clinically, GA patients should undergo cognitive screening and assessment, and be provided with appropriate cognitive and pain management strategies.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-81/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-81/dss

Funding: This work was supported by the grants from the National Natural Science Foundation of China (grant numbers 82274657, 82004468); Natural Science Foundation of Guangdong Province of China (grant number 2024A1515010813); China Postdoctoral Science Foundation (grant number 2019M663021).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-81/coif). All authors report grants from the National Natural Science Foundation of China (grant numbers 82274657, 82004468), Natural Science Foundation of Guangdong Province of China (grant number 2024A1515010813), and China Postdoctoral Science Foundation (grant number 2019M663021). The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The trial was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of the First Affiliated Hospital of Shantou University Medical College (No. B-2021-234), and written informed consent was taken from all individual participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Dalbeth N, Gosling AL, Gaffo A, Abhishek A. Gout. Lancet 2021;397:1843-55. [Crossref] [PubMed]
2. Singh JA, Cleveland JD. Gout and dementia in the elderly: a cohort study of Medicare claims. BMC Geriatr 2018;18:281. [Crossref] [PubMed]
3. DiSabato DJ, Quan N, Godbout JP. Neuroinflammation: the devil is in the details. J Neurochem 2016;139:136-53. [Crossref] [PubMed]
4. Kono H, Chen CJ, Ontiveros F, Rock KL. Uric acid promotes an acute inflammatory response to sterile cell death in mice. J Clin Invest 2010;120:1939-49. [Crossref] [PubMed]
5. Amaro S, Llull L, Renú A, Laredo C, Perez B, Vila E, Torres F, Planas AM, Chamorro Á. Uric acid improves glucose-driven oxidative stress in human ischemic stroke. Ann Neurol 2015;77:775-83. [Crossref] [PubMed]
6. Suzuki K, Koide D, Fujii K, Yamazaki T, Tsuji S, Iwata A. Elevated Serum Uric Acid Levels Are Related to Cognitive Deterioration in an Elderly Japanese Population. Dement Geriatr Cogn Dis Extra 2016;6:580-8. [Crossref] [PubMed]
7. Buchbinder BR. Functional magnetic resonance imaging. Handb Clin Neurol 2016;135:61-92. [Crossref] [PubMed]
8. Lin L, Zheng LJ, Joseph Schoepf U, Varga-Szemes A, Savage RH, Wang YF, Zhang H, Zhang XY, Lu GM, Zhang LJ. Uric Acid Has Different Effects on Spontaneous Brain Activities of Males and Females: A Cross-Sectional Resting-State Functional MR Imaging Study. Front Neurosci 2019;13:763. [Crossref] [PubMed]
9. Baddeley A. Working memory: theories, models, and controversies. Annu Rev Psychol 2012;63:1-29. [Crossref] [PubMed]
10. Constantinidis C, Klingberg T. The neuroscience of working memory capacity and training. Nat Rev Neurosci 2016;17:438-49. [Crossref] [PubMed]
11. Shepard RN, Metzler J. Mental rotation of three-dimensional objects. Science 1971;171:701-3. [Crossref] [PubMed]
12. Emani PS, Warrell J, Anticevic A, Bekiranov S, Gandal M, McConnell MJ, Sapiro G, Aspuru-Guzik A, Baker JT, Bastiani M, Murray JD, Sotiropoulos SN, Taylor J, Senthil G, Lehner T, Gerstein MB, Harrow AW. Quantum computing at the frontiers of biological sciences. Nat Methods 2021;18:701-9. [Crossref] [PubMed]
13. Xin J, Zhang Y, Tang Y, Yang Y. Brain Differences Between Men and Women: Evidence From Deep Learning. Front Neurosci 2019;13:185. [Crossref] [PubMed]
14. Qiu S, Joshi PS, Miller MI, Xue C, Zhou X, Karjadi C, et al. Development and validation of an interpretable deep learning framework for Alzheimer’s disease classification. Brain 2020;143:1920-33. [Crossref] [PubMed]
15. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig L, Lijmer JG, Moher D, Rennie D, de Vet HC, Kressel HY, Rifai N, Golub RM, Altman DG, Hooft L, Korevaar DA, Cohen JF. STARD Group. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ 2015;351:h5527. [Crossref] [PubMed]
16. Neogi T, Jansen TL, Dalbeth N, Fransen J, Schumacher HR, Berendsen D, et al. 2015 Gout Classification Criteria: an American College of Rheumatology/European League Against Rheumatism collaborative initiative. Arthritis Rheumatol 2015;67:2557-68. [Crossref] [PubMed]
17. Nasreddine ZS, Phillips NA, Bédirian V, Charbonneau S, Whitehead V, Collin I, Cummings JL, Chertkow H. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J Am Geriatr Soc 2005;53:695-9. [Crossref] [PubMed]
18. Yu J, Li J, Huang X. The Beijing version of the Montreal Cognitive Assessment as a brief screening tool for mild cognitive impairment: a community-based study. BMC Psychiatry 2012;12:156. [Crossref] [PubMed]
19. Chang Y, Woo ST, Kim Y, Lee JJ, Song HJ, Lee HJ, Kim SH, Lee H, Kwon YJ, Ahn JH, Park SJ, Chung IS, Jeong KS. Pallidal index measured with three-dimensional T1-weighted gradient echo sequence is a good predictor of manganese exposure in welders. J Magn Reson Imaging 2010;31:1020-6. [Crossref] [PubMed]
20. Zheng Y, Xie L, Huang Z, Peng J, Huang S, Guo R, Huang J, Lin Z, Zhuang Z, Yin J, Hou Z, Ma S. Enhanced activity of the left precuneus as a predictor of visuospatial dysfunction correlates with disease activity in rheumatoid arthritis. Eur J Med Res 2023;28:276. [Crossref] [PubMed]
21. Wang J, Wang X, Xia M, Liao X, Evans A, He Y. GRETNA: a graph theoretical network analysis toolbox for imaging connectomics. Front Hum Neurosci 2015;9:386. [Crossref] [PubMed]
22. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O, Delcroix N, Mazoyer B, Joliot M. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 2002;15:273-89. [Crossref] [PubMed]
23. Chang CC, Lin CJ. LIBSVM: A library for support vector machines. ACM Transactions on Intelligent Systems and Technology 2007;2:27.
24. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837-45.
25. Xia M, Wang J, He Y. BrainNet Viewer: a network visualization tool for human brain connectomics. PLoS One 2013;8:e68910. [Crossref] [PubMed]
26. Qiu L, Cheng XQ, Wu J, Liu JT, Xu T, Ding HT, Liu YH, Ge ZM, Wang YJ, Han HJ, Liu J, Zhu GJ. Prevalence of hyperuricemia and its related risk factors in healthy adults from Northern and Northeastern Chinese provinces. BMC Public Health 2013;13:664. [Crossref] [PubMed]
27. Xu Y, Wang Q, Cui R, Lu K, Liu Y, Zhao Y. Uric acid is associated with vascular dementia in Chinese population. Brain Behav 2017;7:e00617. [Crossref] [PubMed]
28. Euser SM, Hofman A, Westendorp RG, Breteler MM. Serum uric acid and cognitive function and dementia. Brain 2009;132:377-82. [Crossref] [PubMed]
29. Kueider AM, An Y, Tanaka T, Kitner-Triolo MH, Studenski S, Ferrucci L, Thambisetty M. Sex-Dependent Associations of Serum Uric Acid with Brain Function During Aging. J Alzheimers Dis 2017;60:699-706. [Crossref] [PubMed]
30. Cao S, Fisher DW, Yu T, Dong H. The link between chronic pain and Alzheimer’s disease. J Neuroinflammation 2019;16:204. [Crossref] [PubMed]
31. Sporns O. Network attributes for segregation and integration in the human brain. Curr Opin Neurobiol 2013;23:162-71. [Crossref] [PubMed]
32. Jia YL, Yang BN, Yang YH, Zheng WM, Wang L, Huang CY, Lu J, Chen N. Application of machine learning techniques in the diagnostic approach of PTSD using MRI neuroimaging data: A systematic review. Heliyon 2024;10:e28559. [Crossref] [PubMed]
33. Cantor-Rivera D, Khan AR, Goubran M, Mirsattari SM, Peters TM. Detection of temporal lobe epilepsy using support vector machines in multi-parametric quantitative MR imaging. Comput Med Imaging Graph 2015;41:14-28. [Crossref] [PubMed]
34. Eriksson J, Vogel EK, Lansner A, Bergström F, Nyberg L. Neurocognitive Architecture of Working Memory. Neuron 2015;88:33-46. [Crossref] [PubMed]
35. Bettencourt KC, Xu Y. Decoding the content of visual short-term memory under distraction in occipital and parietal areas. Nat Neurosci 2016;19:150-7. [Crossref] [PubMed]
36. Booth JR, MacWhinney B, Thulborn KR, Sacco K, Voyvodic JT, Feldman HM. Developmental and lesion effects in brain activation during sentence comprehension and mental rotation. Dev Neuropsychol 2000;18:139-69. [Crossref] [PubMed]
37. Peelen MV, Kastner S. Attention in the real world: toward understanding its neural basis. Trends Cogn Sci 2014;18:242-50. [Crossref] [PubMed]
38. Culham JC, Kanwisher NG. Neuroimaging of cognitive functions in human parietal cortex. Curr Opin Neurobiol 2001;11:157-63. [Crossref] [PubMed]
39. Berryhill ME, Olson IR. The right parietal lobe is critical for visual working memory. Neuropsychologia 2008;46:1767-74. [Crossref] [PubMed]
40. Koenigs M, Barbey AK, Postle BR, Grafman J. Superior parietal cortex is critical for the manipulation of information in working memory. J Neurosci 2009;29:14980-6. [Crossref] [PubMed]
41. Mizukoshi T, Yamada S, Sakakibara SI. Spatiotemporal Regulation of De Novo and Salvage Purine Synthesis during Brain Development. eNeuro 2023;10. ENEURO. [Crossref] [PubMed]
42. Nee DE, Brown JW, Askren MK, Berman MG, Demiralp E, Krawitz A, Jonides J. A meta-analysis of executive components of working memory. Cereb Cortex 2013;23:264-82. [Crossref] [PubMed]
43. Moriarty O, McGuire BE, Finn DP. The effect of pain on cognitive function: a review of clinical and preclinical research. Prog Neurobiol 2011;93:385-404. [Crossref] [PubMed]
44. Wandell BA, Dumoulin SO, Brewer AA. Visual field maps in human cortex. Neuron 2007;56:366-83. [Crossref] [PubMed]
45. Jerath R, Crawford MW. Neural correlates of visuospatial consciousness in 3D default space: insights from contralateral neglect syndrome. Conscious Cogn 2014;28:81-93. [Crossref] [PubMed]
46. Čeko M, Shir Y, Ouellet JA, Ware MA, Stone LS, Seminowicz DA. Partial recovery of abnormal insula and dorsolateral prefrontal connectivity to cognitive networks in chronic low back pain after treatment. Hum Brain Mapp 2015;36:2075-92. [Crossref] [PubMed]
47. Lara AH, Wallis JD. The Role of Prefrontal Cortex in Working Memory: A Mini Review. Front Syst Neurosci 2015;9:173. [Crossref] [PubMed]
48. Wang M, Yang Y, Wang CJ, Gamo NJ, Jin LE, Mazer JA, Morrison JH, Wang XJ, Arnsten AF. NMDA receptors subserve persistent neuronal firing during working memory in dorsolateral prefrontal cortex. Neuron 2013;77:736-49. [Crossref] [PubMed]
49. Bird CM, Burgess N. The hippocampus and memory: insights from spatial processing. Nat Rev Neurosci 2008;9:182-94. [Crossref] [PubMed]
50. Sreenivasan KK, Curtis CE, D’Esposito M. Revisiting the role of persistent neural activity during working memory. Trends Cogn Sci 2014;18:82-9. [Crossref] [PubMed]
51. Rizzo FR, Musella A, De Vito F, Fresegna D, Bullitta S, Vanni V, Guadalupi L, Stampanoni Bassi M, Buttari F, Mandolesi G, Centonze D, Gentile A. Tumor Necrosis Factor and Interleukin-1β Modulate Synaptic Plasticity during Neuroinflammation. Neural Plast 2018;2018:8430123. [Crossref] [PubMed]
52. Vachon-Presseau E, Centeno MV, Ren W, Berger SE, Tétreault P, Ghantous M, Baria A, Farmer M, Baliki MN, Schnitzer TJ, Apkarian AV. The Emotional Brain as a Predictor and Amplifier of Chronic Pain. J Dent Res 2016;95:605-12. [Crossref] [PubMed]
53. Chaudhary NS, Bridges SL Jr, Saag KG, Rahn EJ, Curtis JR, Gaffo A, Limdi NA, Levitan EB, Singh JA, Colantonio LD, Howard G, Cushman M, Flaherty ML, Judd S, Irvin MR, Reynolds RJ. Severity of Hypertension Mediates the Association of Hyperuricemia With Stroke in the REGARDS Case Cohort Study. Hypertension 2020;75:246-56. [Crossref] [PubMed]
54. Li X, Huang L, Tang Y, Hu X, Wen C. Gout and risk of dementia, Alzheimer’s disease or vascular dementia: a meta-epidemiology study. Front Aging Neurosci 2023;15:1051809. [Crossref] [PubMed]
55. Vitturi BK. Rheumatic cognitive impairment: a new clinical entity? Lancet Rheumatol 2025;7:e80-1. [Crossref] [PubMed]