Association Between Cognitive Impairment and Multimodal MRI Markers in Patients with Cerebral Microbleeds: A Retrospective Case–Control Study

Introduction

Cerebral microbleeds (CMBs) are small hemosiderin deposits caused by cerebral small vessel disease (CSVD), typically defined as round or ovoid hypointense lesions with a diameter of 2–10 mm on susceptibility-weighted imaging (SWI).^1,2 They are predominantly distributed in the subcortical white matter, basal ganglia, thalamus, and brainstem, and are increasingly recognized as one of the core neuroimaging markers of CSVD. CMBs are frequently observed in patients with hypertension, ischemic stroke, intracerebral hemorrhage, and neurodegenerative disorders, as well as in the aging population.³ With the widespread application of high-resolution MRI, the detection rate of CMBs has risen steadily, and their broader clinical implications have attracted growing attention. While early research primarily focused on the association between CMBs and hemorrhagic risk, accumulating evidence suggests that CMBs may also contribute to cognitive impairment. Proposed mechanisms include disruption of white matter microstructure, impairment of structural brain networks, and secondary effects on functional connectivity.^4,5 Several studies have demonstrated that a higher CMB burden and wider anatomical distribution are associated with poorer global cognition and deficits in executive function, attention, and processing speed.^6,7 These findings support the concept that CMBs may represent not merely focal vascular lesions but markers of diffuse network-level brain injury.

Advances in multimodal MRI, particularly diffusion tensor imaging (DTI), have provided new opportunities to investigate the microstructural basis of CMB-related cognitive dysfunction. DTI-derived parameters, including fractional anisotropy (FA) and apparent diffusion coefficient (ADC), enable quantitative assessment of white matter integrity and microstructural damage beyond what can be detected by conventional MRI.⁸ Prior DTI-based studies have shown that decreased FA and increased ADC are associated with cognitive decline in CSVD and mild cognitive impairment, indicating potential predictive value.^9,10 However, most existing studies have examined DTI abnormalities or CMB burden in isolation, with limited integration of multimodal imaging parameters and standardized cognitive assessments. Importantly, it remains unclear whether regional white matter microstructural alterations mediate the relationship between CMB burden and cognitive impairment, and which brain regions are most strongly associated with cognitive decline. Moreover, the independent predictive value of combined CMB characteristics and DTI parameters has not been fully elucidated. From a network disruption perspective, CMB-related microstructural damage may impair information transfer efficiency across key cognitive networks, thereby contributing to global cognitive dysfunction. Therefore, we hypothesized that increased CMB burden and region-specific white matter microstructural damage, as quantified by multimodal MRI-derived FA and ADC parameters, are independently associated with cognitive impairment. Accordingly, this retrospective case–control study aimed to compare cognitive performance and multimodal MRI parameters between patients with CMBs and healthy controls, to analyze the associations between CMB burden, regional FA/ADC values, and cognitive scores, and to identify independent predictors of cognitive dysfunction, thereby providing imaging-based evidence for early risk stratification and intervention.

Materials and Methods

Study Design and Participants

This study was a single-center, retrospective case–control study. All MRI examinations and clinical data were obtained from routine clinical practice, and cognitive assessments were completed within 7 days after MRI as part of standardized clinical evaluation, without any prospective intervention. A total of 71 consecutive patients diagnosed with CMBs by MRI and hospitalized in the Neurology Department of our hospital between August 2022 and February 2024 were enrolled as the observation group. Meanwhile, 65 age- and sex-matched healthy individuals undergoing routine health examinations during the same period were recruited as the control group. No subject was enrolled or assessed based on future outcomes, ensuring the retrospective nature of the study.

The study protocol was approved by the Medical Ethics Committee of Quanzhou First Hospital (Approval No.: SNYX24-19) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants or their legal representatives. The research overview flowchart is shown in Figure 1.

Figure 1 Research Overview Flowchart.

Sample Size Consideration

Given the retrospective design, no a priori sample size calculation was performed. However, the final sample size was comparable to or larger than that of previous multimodal MRI studies investigating CMBs-related cognitive impairment, and was considered adequate to detect moderate-to-large effect sizes in correlation and multivariate regression analyses.

Inclusion and Exclusion Criteria

Inclusion criteria: Observation group: (1) Age ≥50 years; (2) Diagnosis of CMBs confirmed by cranial MRI including SWI; (3) Ability to complete cognitive assessment and MRI examination. Control group: (1) Age- and sex-matched with the observation group; (2) No history of neurological or psychiatric disorders; (3) No structural abnormalities on cranial MRI.

Exclusion criteria: (1) Neurodegenerative diseases (eg, Alzheimer’s disease, Parkinson’s disease); (2) Severe sensory, psychiatric, or communication disorders; (3) Recent stroke, traumatic brain injury, or intracranial surgery; (4) Poor image quality or severe motion artifacts; (5) Severe systemic diseases or acute metabolic disorders affecting cognition.

Cognitive Function Assessment

Cognitive function was assessed using the Montreal Cognitive Assessment (MoCA) by trained neurologists within 7 days after MRI. The MoCA evaluates executive function, attention, memory, language, abstraction, and orientation (total score: 30). Scores ≤26 indicated cognitive impairment, with one additional point added for individuals with ≤12 years of education. All assessments were performed by the same evaluator in a quiet environment.

MRI Acquisition and Post-Processing

MRI scans were performed on a 3.0-T scanner (GE SIGNA EXCITE II). Sequences included T1WI, T2WI, FLAIR, SWI, and DTI. DTI was acquired with 32 diffusion directions (b = 1000 s/mm²). DTI data were processed using FSL (version 6.0). Preprocessing steps included eddy current and head motion correction, brain extraction, and tensor fitting. Fractional anisotropy (FA) and apparent diffusion coefficient (ADC) maps were generated. Regions of interest (ROIs) including the frontal–temporal lobe, parietal lobe, occipital lobe, and basal ganglia were defined using a standard neuroanatomical atlas. Automated extraction was followed by manual verification to ensure anatomical accuracy and consistency across subjects.

Assessment of CMBs

CMBs were identified on SWI as round or ovoid hypointense lesions with diameters of 2–10 mm. Two experienced neuroradiologists independently evaluated all images. Discrepancies were resolved by consensus with a third reviewer.

Data Preprocessing and Variable Transformation

Prior to regression modeling, continuous variables were assessed for normality. The number of CMBs was dichotomized using a threshold of ≥10 lesions, based on prior literature and observed distribution, to reflect high lesion burden. FA and ADC values were entered as continuous variables after z-score normalization. ROI-specific rules were applied uniformly across all regions.

Covariates and Multicollinearity Assessment

Potential vascular confounders, including hypertension and diabetes mellitus, were collected and entered as covariates. Multicollinearity was assessed using variance inflation factors (VIF), with VIF < 5 considered acceptable.

Statistical Analysis

Statistical analyses were conducted using SPSS 26.0. Group comparisons were performed using independent-sample t-tests or χ²-tests as appropriate. Spearman correlation analysis assessed associations between imaging parameters and cognitive scores. Multivariate linear regression was used to identify independent predictors of cognitive impairment. Model assumptions were checked, and missing data (<5%) were handled by complete-case analysis. A two-tailed P < 0.05 was considered statistically significant.

Results

Comparison of General Data

The observation group consisted of 71 patients with CMBs, including 38 males and 33 females; the average age was (66.42 ± 8.75) years. The control group included 65 individuals, 35 males and 30 females; the average age was (65.83 ± 9.12) years. There were no statistically significant differences between the two groups in terms of gender, age, body mass index (BMI), or years of education (P > 0.05), indicating good baseline comparability. In addition, the prevalence of hypertension and diabetes mellitus did not differ significantly between the two groups (both P > 0.05). See Table 1.

Table 1 Comparison of General Data Between the Two Groups

Cognitive Function Score Results

The MoCA score in the observation group was (21.63 ± 1.59), whereas that in the control group was (26.58 ± 1.97). The MoCA score in the observation group was significantly lower than that in the control group (mean difference: −4.95 points; 95% CI: −5.56 to −4.34; P < 0.001), as shown in Figure 2.

Figure 2 Comparison of MoCA Scores Between the Two Groups. Note: Intergroup comparison, *P < 0.05.

Comparison of CMBs Count and Imaging Indicators

The total number of CMBs in the observation group was (12.41 ± 6.85), which was significantly higher than that in the control group (0.53 ± 0.26) (P < 0.001; Cohen’s d = 2.41), as shown in Figure 3. In the observation group, CMBs were mainly distributed in the frontal–temporal lobe, parietal lobe, and basal ganglia, with basal ganglia involvement accounting for 54.0%, representing the most common site.

Figure 3 Comparison of the number of CMBs between the two groups. Note: Intergroup comparison, *P < 0.05.

Comparison of FA and ADC Values in Brain Regions

In the multimodal MRI analysis, FA values in the frontotemporal lobe, parietal lobe, and basal ganglia in the observation group were significantly lower than those in the control group, while ADC values were significantly higher (all P < 0.001; effect sizes ranged from Cohen’s d = 0.82 to 1.46). There was no statistically significant difference in FA or ADC values in the occipital lobe between the two groups (P > 0.05). The lack of significance in the occipital lobe was accompanied by small effect sizes (FA: d = 0.12; ADC: d = 0.10), suggesting limited clinical relevance, as shown in Table 2.

Table 2 Comparison of FA and ADC Values Between the Two Groups by Brain Region

Spearman Correlation Analysis Results

Spearman correlation analysis showed that the number of CMBs was significantly negatively correlated with the MoCA score (r = −0.643, 95% CI: −0.76 to −0.50; P < 0.001). FA values in the frontal–temporal lobe were positively correlated with the MoCA score (r = 0.529, 95% CI: 0.32–0.69; P = 0.002), whereas ADC values were negatively correlated with the MoCA score (r = −0.597, 95% CI: −0.73 to −0.41; P < 0.001). FA values in the basal ganglia were positively correlated with the MoCA score (r = 0.563, 95% CI: 0.36–0.71; P < 0.001), while ADC values were negatively correlated with the MoCA score (r = −0.626, 95% CI: −0.75 to −0.46; P < 0.001), as shown in Table 3 and Figures 4–6. No significant correlations were observed between parietal lobe FA/ADC values and MoCA scores (both P > 0.10), which may partly reflect limited statistical power for region-specific analyses.

Table 3 Correlation Analysis of CMBs Count, FA/ADC Values, and Cognitive Scores

Figure 4 Scatter plot of correlation between CMBs count and cognitive scores.

Figure 5 Scatter plot of correlation between FA/ADC values in frontotemporal lobe and cognitive scores.

Figure 6 Scatter plot of correlation between FA/ADC values in basal ganglia and cognitive scores.

Multiple Linear Regression Analysis Results

Using MoCA scores as the dependent variable, and gender, age, years of education, hypertension, diabetes mellitus, CMBs count (≥10), and FA/ADC values in the frontotemporal lobe and basal ganglia as independent variables, a multiple linear regression model was constructed. Results indicated that having ≥10 CMBs lesions, reduced FA values, increased ADC values, and involvement of the basal ganglia were independent risk factors for cognitive impairment (all P < 0.05), whereas hypertension and diabetes mellitus were not independently associated with MoCA scores (both P > 0.05). The final model explained 58.4% of the variance in cognitive performance (adjusted R² = 0.584). Model diagnostics demonstrated acceptable residual normality and no evidence of multicollinearity (all VIFs < 2.5), as shown in Table 4.

Table 4 Multiple Linear Regression Analysis: Independent Predictors of Cognitive Function (MoCA Score)

Discussion

This study aimed to explore the characteristics of cognitive impairment in patients with CMBs and its association with multimodal MRI parameters, further identifying independent risk factors influencing cognitive function. The results showed that patients with CMBs scored significantly lower on the MoCa scale compared to healthy controls, indicating notable cognitive impairment. Moreover, this study confirmed that the number, distribution, and microstructural changes in white matter (decreased FA, increased ADC) were closely related to the severity of cognitive deficits. A CMBs lesion count ≥10, basal ganglia involvement, and abnormal changes in FA/ADC values were identified as significant predictors of cognitive dysfunction.

Association Between CMBs and Cognitive Impairment

CMBs are key neuroimaging manifestations within the spectrum of CSVD, typically representing hemosiderin deposition from chronic hemorrhage in the brain parenchyma, commonly observed in SWI sequences.¹¹ The pathological basis includes microarteriolosclerosis, fibrinoid necrosis, microaneurysm formation, and blood-brain barrier dysfunction.¹² Persistent hypertension, cerebral arteriosclerosis, and cerebral amyloid angiopathy are major contributing factors for the formation of CMBs, which in turn may affect cerebral functional homeostasis through multiple mechanisms.¹³ Although our study accounted for hypertension and diabetes in the regression model, further assessment of cerebrovascular comorbidities, including amyloid burden, is warranted to fully clarify the independent impact of CMBs.

Recent neuroimaging studies^14–16 have indicated that CMBs not only cause local microstructural damage but also disrupt the integrity of neural network connectivity, leading to dysfunction in remote brain areas. This is particularly evident in the cognitive systems of the elderly. Notably, DTI-based connectome studies have shown that CMBs can disrupt large-scale networks, including the default mode network and executive control network, contributing to domain-specific deficits such as impaired executive function and memory retrieval.^17,18 In this study, a significant negative correlation was observed between total CMBs count and cognitive scores (MoCA), suggesting a progressive decline in cognitive levels with increasing microbleed burden. This supports the currently proposed “lesion burden-related damage model”,¹⁹ where the number and cumulative impact of CMBs are positively associated with the extent of cognitive impairment.

Furthermore, the anatomical distribution of microbleeds may play a crucial role in the mechanism of cognitive decline. This study highlighted that the basal ganglia region was among the most common sites for microbleed occurrence, and its involvement was strongly associated with cognitive dysfunction. The basal ganglia are not only involved in motor control but also form cognitive function-related loops through connections with the prefrontal cortex, thalamus, and limbic system, governing executive control, motivation, working memory, and behavioral flexibility.²⁰ Therefore, when CMBs affect the basal ganglia, they may impair signal transmission within these circuits, resulting in inefficient information processing and executive dysfunction, manifested clinically as inattention, reduced planning ability, and poor adaptability.

Relationship Between White Matter Microstructure and Cognitive Function

White matter serves as the critical infrastructure for high-speed communication between brain regions, and its microstructural integrity is fundamental to higher-order cognitive activities.²¹ DTI, a sophisticated technique for assessing white matter microstructure, has been widely applied in various neurological disorders. FA reflects the directional preference of water molecule diffusion along axons, representing axonal alignment and myelin integrity,²² while ADC measures the overall diffusion level of water molecules and correlates with cell density and tissue complexity.²³ In this study, quantitative DTI analysis revealed significantly decreased FA values and increased ADC values in the frontal-temporal lobes, parietal lobes, and basal ganglia of CMBs patients compared to healthy controls, indicating microstructural damage in these brain regions. These alterations may reflect underlying pathologies such as axonal disruption, demyelination, or glial activation, suggesting that CMBs may induce not only focal deposition of blood components but also secondary damage to white matter structures, impairing the speed and accuracy of information transmission. From a network perspective, the observed regional FA/ADC changes likely reflect disruptions in large-scale networks, such as the default mode and executive control networks, which may underlie the domain-specific cognitive deficits observed in these patients. From a functional localization perspective, the frontal-temporal lobes are central to executive functions, working memory, and language processing. Damage to these areas can directly lead to impairments in task-switching, contextual integration, and strategy formulation.²⁴ The parietal lobes are involved in spatial orientation and visual attention; white matter damage in these areas often results in reduced directional sense and diminished visual search efficiency.²⁵ The basal ganglia modulate executive control, emotional regulation, and reward mechanisms through loops connecting to the prefrontal cortex, thalamus, and limbic system, functioning as a core hub of the cognitive network. Structural abnormalities in this region can disrupt subcortical circuitry and lead to reduced behavioral motivation and delayed response speed.²⁶

Further correlation analysis revealed that FA values in the frontal-temporal lobes and basal ganglia were positively correlated with MoCA scores, while ADC values were negatively correlated, supporting their role as sensitive imaging markers for cognitive impairment. These findings indicate a clear linear relationship between the degree of white matter damage in these regions and overall cognitive performance, suggesting potential utility for early risk identification and quantitative assessment of cognitive decline. Notably, this study did not find statistically significant correlations between FA/ADC values in the parietal lobes and cognitive scores, which may be attributed to several factors: First, damage in the parietal region may affect specific cognitive domains (eg, visuospatial abilities, sustained attention),²⁷ which are not fully captured by global cognitive measures such as MoCA. Second, the degree of parietal white matter damage in this study cohort might have been relatively mild, not reaching the threshold to impact overall cognitive function. Lastly, the parietal lobes may exhibit greater neuroplasticity and compensatory capacity, masking early damage in standard cognitive assessments.²⁸

These findings are consistent with studies by Hernández-Díaz et al²⁹ and Hristovska et al³⁰ on CSVD-related white matter lesions and cognitive decline, reinforcing the concept that microstructural white matter damage, particularly within key functional hub regions, acts as a critical intermediary in CMBs-related cognitive impairment. Therefore, DTI not only offers superior sensitivity in detecting microstructural brain changes but also holds promise as a frontline imaging tool for cognitive risk assessment.

Clinical Application Value of Multivariate Regression Analysis in Cognitive Risk Assessment

By constructing a multivariate linear regression model, this study incorporated multiple variables that may affect cognitive function, including demographic factors (such as gender, age, and years of education), imaging characteristics (such as the number and distribution of CMBs), and microstructural white matter indicators (FA, ADC values, etc). The results showed that having ≥10 CMB lesions, involvement of the basal ganglia, decreased FA values, and increased ADC values were significant independent predictors of cognitive decline. These findings not only deepen the understanding of the mechanisms linking CMBs and cognitive impairment but also provide quantitative evidence for clinically identifying high-risk populations.

Unlike traditional methods that rely solely on cognitive scale scores or subjective clinical judgment, this study emphasized the value of quantitative imaging parameters in cognitive function assessment. Particularly in the early stages when some patients have not yet shown obvious symptoms of cognitive impairment, structural burden indicators based on MRI combined with DTI microstructural parameters can objectively predict the risk of potential cognitive damage, thereby improving the sensitivity and foresight of disease screening. Moreover, this integrative multifactor analysis model supports the formulation of individualized intervention strategies and helps in accurately identifying patient groups that require enhanced management and intervention. In addition, the progression of cognitive impairment in CMBs patients is often insidious and diverse,³¹ and the independent predictive capabilities of the indicators in the regression model suggest that cognitive decline is not caused by a single factor but is the result of multiple brain structural damage mechanisms acting in concert.³²

Strengths and Limitations of This Study

The main strength of this study lies in the use of multimodal MRI technology for a systematic assessment of patients with cerebral microbleeds. It not only focused on lesion count and distribution but, for the first time, incorporated microstructural indicators such as FA and ADC to quantitatively explore their relationship with cognitive impairment, enhancing both precision and scientific rigor. Importantly, the findings suggest that patients with a high CMB burden or abnormal FA/ADC values in critical regions could be prioritized for enhanced cognitive monitoring. As a preliminary guideline, clinicians may consider follow-up intervals of 6–12 months and pay special attention when FA decreases by ≥0.05 or ADC increases by ≥0.1 × 10⁻³ mm²/s relative to baseline, although these thresholds require validation in larger cohorts.

In addition, future studies could leverage artificial intelligence and advanced computational modeling to integrate multimodal MRI parameters, cognitive assessments, and clinical features to develop early predictive models for cognitive decline. For example, recent computational neuroimaging frameworks demonstrate the feasibility of combining structural and functional MRI data to simulate network-level brain dynamics and predict domain-specific cognitive deficits.

There are also several limitations. First, this was a single-center retrospective study, and subtle differences in scanner hardware or protocols may introduce systematic bias, highlighting the need for multicenter validation. Second, the study did not conduct domain-specific cognitive analyses, which could help refine patterns of impairment across executive function, memory, attention, and visuospatial processing. Third, other CSVD markers (eg, white matter hyperintensities, lacunar infarctions) were not included, possibly underestimating cumulative pathological burden. Future work should integrate multimodal neuroimaging, electrophysiology, biomarkers, and longitudinal follow-up to construct multifactorial predictive models and evaluate whether interventions—such as neuroprotection, anti-inflammatory therapy, or vascular remodeling—can mitigate white matter damage and cognitive decline in CMBs patients.

Conclusion

This study systematically investigated the associations between CMBs, white matter microstructure, and cognitive function using cognitive assessments combined with multimodal MRI. The findings indicate that higher CMB burden, basal ganglia involvement, and region-specific alterations in FA and ADC are independently associated with cognitive impairment. These neuroimaging markers may serve as potential indicators for early risk stratification and for identifying patients who could benefit from closer cognitive monitoring.

The study highlights the clinical relevance of combining conventional MRI and DTI to detect microstructural white matter damage and to quantify alterations in brain integrity in CMBs patients. From a mechanistic perspective, the results support the role of CMB-related microstructural disruption in cognitive decline, providing a foundation for future longitudinal studies and interventional trials aimed at mitigating cognitive deterioration.