Mild cognitive impairment (MCI) is a transitional state between normal aging and dementia, characterized by objective evidence of cognitive decline, with a high likelihood of progressing to dementia. Therefore, it is essential to intervene and treat individuals in this transitional state to delay the progression of cognitive impairment and reduce the incidence of dementia (1). Diffusion tensor imaging (DTI) detects the diffusion of water molecules within tissues, revealing the microstructure and integrity of brain white matter fibers. Fractional anisotropy (FA) values are commonly used quantitative parameters in DTI examinations to analyze changes in white matter fibers. These values are crucial for identifying microstructural damage in the white matter of MCI patients (2, 3). Currently, numerous studies have been conducted globally on the use of DTI in diagnosing MCI. However, variations in results across these studies are due to various reasons. Therefore, this study systematically reviews the literature up to February 20, 2024, to evaluate the diagnostic value of DTI for MCI through meta-analysis. This analysis aims to provide objective evidence for selecting clinical examination methods and assessing diagnostic efficacy.

Mild cognitive impairment (MCI) is a transitional stage between normal aging and dementia, characterized by cognitive decline that exceeds normal age-related changes but does not meet the criteria for dementia (4). Alzheimer’s disease (AD) is the most common neurodegenerative disorder and the most prevalent type of dementia. The probability of MCI patients developing dementia within 1 year is 10–15%, and within 2 years, the probability is 40%, with the incidence rate increasing annually (5, 6). Studies have shown that some MCI patients experience cognitive improvement over time, and some even revert to normal cognitive function (7, 8). However, research by Roberts et al. (9) found that patients whose cognition returned to normal are more likely to progress to dementia than those who never had MCI. Therefore, early diagnosis and intervention in MCI patients are crucial.

Diffusion tensor imaging (DTI) is an advancement based on magnetic resonance diffusion-weighted imaging. It not only observes the movement speed of water molecules within tissues but also applies diffusion gradients in more than six directions to obtain anisotropic diffusion of water molecules within the plane. This forms images that trace the pathways of fibers, allowing for the non-invasive tracking of brain white matter fibers and reflecting their structural integrity and connectivity. DTI provides an objective basis for evaluating the pathophysiological changes in tissue structure, aiding in clinical diagnosis. In recent years, ex vivo micro-diffusion tensor imaging (micro-DTI) has demonstrated significant potential in neuroscience research. Studies have shown that micro-DTI can provide higher resolution and more precise information about microstructural changes, enabling fine quantification of FA changes in perforating pathways (10). This capability is particularly important for understanding the microstructural alterations of specific neural pathways. However, single DTI technology may not fully capture the complex pathophysiological processes of MCI. Quantitative Susceptibility Mapping (QSM) can quantify changes in iron deposition in the brain, particularly in the hippocampus and deep gray matter structures, where iron deposition is closely associated with the pathological features of MCI (11, 12). Additionally, QSM can assess myelin damage and venous oxygen saturation, providing supplementary information for the diagnosis of MCI (13, 14). The high resolution and quantitative capabilities of QSM enable it to offer more detailed tissue magnetic property information, thereby complementing DTI’s limitations in assessing the integrity of white matter fiber tracts. The combination of these two techniques can provide a more comprehensive understanding of brain structure and function, offering deeper insights into the underlying pathological processes of MCI and other neurodegenerative diseases, which is beneficial for the early diagnosis and intervention of MCI.

In studies on AD and MCI, the combination of DTI and functional magnetic resonance imaging (fMRI) with machine learning and deep learning techniques has significantly enhanced diagnostic accuracy. The diffusion-based graph contrast learning method (DGCL), through diffusion processes and graph contrastive learning, strengthens the consistency of brain networks, effectively mitigating the impact of individual differences on diagnostic outcomes, thereby improving the diagnostic accuracy for AD and MCI (15). The Decoupled Generative Adversarial Network (DecGAN) decomposes brain networks into sparse subgraphs and complementary graphs through a decoupling module. It utilizes an adversarial strategy to guide the decoupling module in extracting features more relevant to AD. By encoding the detected neural circuits using hypergraph data, DecGAN significantly enhances the diagnostic accuracy of AD (16). The model based on prior-guided adversarial learning and hypergraphs (PALH) guides multimodal representation learning by estimating the prior distribution of anatomical knowledge. It utilizes adversarial strategies to reduce the discrepancy between representation distributions. The hypergraph-aware network designed in this model effectively integrates the learned representations, establishing higher-order relationships both across and within modalities. This enhances the accuracy and reliability of abnormal connectivity prediction in AD (17). These studies are not only theoretically innovative but also provide significant practical insights, offering new perspectives and directions for the diagnosis and treatment of AD and MCI.

FA value is a key parameter used in DTI to quantify the directional diffusion of water molecules. It ranges from 0 to 1, with higher values indicating greater restriction of water molecule diffusion along a particular direction, typically reflecting the integrity of white matter tracts. This parameter is especially significant in many neurodegenerative diseases, particularly AD, as an early key biomarker. However, the interpretation of FA values must be considered within a complex biological context, as changes in FA may involve multiple mechanisms rather than just myelin damage. Firstly, a reduction in FA could be related to disruption of axonal membrane integrity, a decrease in axonal density, or dysfunction in axonal transport (18). Secondly, changes in the extracellular matrix due to glial cell proliferation or inflammatory responses might also reduce FA by increasing isotropic diffusion (19). Additionally, in areas of fiber crossing or branching (such as at the junction of the corpus callosum and the corona radiata), the natural multidirectionality of water molecule diffusion due to the presence of fibers in multiple directions can lead to a decrease in FA. This phenomenon is not indicative of pathological damage but rather reflects normal anatomical structures (20). It’s also worth noting that some studies suggest that a slight decrease in FA in specific brain regions may reflect fiber reorganization or compensatory repair processes, rather than purely white matter degeneration (21). Studies have shown that combining other diffusion metrics, such as mean diffusivity, radial diffusivity, and axial diffusivity, can provide a more comprehensive understanding of white matter integrity (22). Therefore, the clinical significance of the FA value should be interpreted in conjunction with the anatomical characteristics of the fiber tracts, the heterogeneity of the patient population, and multimodal imaging analyses.

The study found that patients with MCI exhibited significant reductions in the FA values of cortical-related brain regions, including the frontal lobe (WMD = −0.03, 95% CI: −0.04 to −0.02), hippocampus (WMD = −0.03, 95% CI: −0.04 to −0.01), and the splenium of the corpus callosum (WMD = −0.05, 95% CI: −0.05 to −0.02), suggesting impaired white matter integrity. These regions are closely associated with executive function, memory integration, and interhemispheric information transfer, potentially serving as biomarkers for early diagnosis. For instance, the reduced FA in the splenium of the corpus callosum is correlated with decreased functional connectivity in the bilateral frontal lobes, which may explain the executive control deficits observed in MCI patients. The WMD, 95% CI, heterogeneity (I² value), and clinical significance of each brain region are shown in Table 2.

The frontal and parietal lobes play significant roles in executive control functions. Zhao et al. (23) found that the functional connectivity of the frontal–parietal network is lower in AD patients, which is closely related to the decline in executive control functions observed in these patients (24). This may serve as a potential non-invasive biomarker for early diagnosis of AD. The superior and inferior longitudinal fasciculi connect the frontal and parietal lobes, transmitting sensory, visual, auditory, and proprioceptive information from the back of the brain to the front. These fiber tracts are crucial for memory, attention, and executive functions (25). The hippocampus is considered a critical brain region for memory and cognitive functions. The fornix, carrying most of its axons from the hippocampal output fibers, is an essential part of the hippocampal memory circuit (26). A reduction in the functional connectivity of the hippocampus is significantly associated with the disruption of fornix integrity (27). Amnestic MCI may predominantly affect brain regions associated with memory, such as the hippocampus and fornix, whereas non-amnestic MCI may show more pronounced reductions in FA values in brain areas related to executive function and language, such as the frontal and parietal lobes. The inferior fronto-occipital fasciculus connects the inferior dorsal frontal lobe, temporal lobe, and occipital lobe, playing a vital role in visual–spatial processing, object recognition, and memory. Compared to MCI patients, the effect size of the inferior fronto-occipital fasciculus is reduced in AD patients, marking it as an important site of white matter lesions (28). The corpus callosum, the main commissural fiber bundle connecting the two cerebral hemispheres, is crucial for integrating sensory and motor functions (29). Damage to the integrity of this part will reduce functional connectivity between the hemispheres, inducing cognitive impairment (25). Previous studies (30–32) have shown that cognitive functions are related to the cingulum and uncinate fasciculus. The cingulum is essential for episodic memory, while the uncinate fasciculus is associated with verbal memory, visual attention, verbal abstraction, and immediate recall cognitive functions.

In the studies included in this paper, most research focused on the left hemisphere of the brain and showed statistical significance. McHugh et al. (33) found that the dominant hemisphere in humans is often the left hemisphere, resulting in functional asymmetry between the brain hemispheres. The right hippocampus is more closely associated with spatial information, whereas most MCI patients primarily exhibit a decline in cognitive memory functions. Consequently, changes in the left hippocampus are more pronounced compared to the right, with the left hippocampus being smaller in volume than the right. Zhang et al. (34) reported that the FA values of fiber circuits in the left hemisphere decreased more in MCI patients compared to normal controls. Therefore, if data from both hemispheres were available in the included studies, the left hemisphere data were uniformly selected.

The following hypotheses are made regarding the sources of heterogeneity in the study: (1) The included studies primarily consist of retrospective studies and follow-up studies with varying sample sizes, which may lead to selection and information bias. (2) Different experimental conditions were used across studies, such as variations in equipment models, scanning parameters, and post-processing methods, resulting in high heterogeneity between the studies. Specifically, in terms of processing methods, most studies used region of interest (ROI) measurements for FA values, while some used voxel-based morphometry (VBM) techniques, and others applied Tract-Based Spatial Statistics (TBSS) methods. The latter two methods can automatically and objectively observe changes in brain white matter fiber tracts, minimizing the impact of human factors. These factors contribute to the high heterogeneity observed in this paper. To explore the impact of different scanning parameters on the results, we conducted a subgroup analysis based on magnetic field strength (1.5 T vs. 3.0 T), as shown in Figure 5. The subgroup analysis revealed that the effect sizes were consistent for both 1.5 T and 3.0 T field strengths, both showing an effect size of −0.03 with similar 95% confidence intervals. The levels of heterogeneity were also similar, indicating significant heterogeneity in both cases. Therefore, magnetic field strength does not appear to be the primary factor contributing to the heterogeneity. Other factors, such as study design, sample size, and participant characteristics, may be the main contributors to the heterogeneity. Although there is still significant heterogeneity, the funnel plot symmetry test indicates a low risk of publication bias. Moreover, the large sample size partially offsets such biases, and the combined results maintain both clinical and statistical significance, making the overall findings relatively robust and reliable.

A previous meta-analysis (35) discussed the diagnostic value of DTI for MCI patients. The findings of this study are generally consistent with previous meta-analyses; however, we additionally report that the analysis of the posterior limb of the internal capsule showed no statistical significance, which differs from earlier studies. This discrepancy may be related to the insufficient sample size. This study explores the diagnostic value of DTI for MCI based on a large sample size, including 76 studies. The results show that the overall pooled WMD and its 95% CI were −0.03 [−0.04, −0.02], with FA values being lower in the MCI group compared to the NC group. This difference is statistically significant (p < 0.05), indicating significant microstructural damage in the white matter of MCI patients compared to cognitively normal individuals.

For the occipital lobe, the pooled WMD and its 95% CI were −0.00 [−0.01, 0.00], with the 95% CI crossing the null line, indicating no statistically significant difference (p = 0.33 > 0.05). Similarly, for the posterior limb of the internal capsule, the pooled WMD and its 95% CI were −0.04 [−0.08, 0.01], with the 95% CI crossing the null line, indicating no statistically significant difference (p = 0.12 > 0.05). This is consistent with previous findings, which show that the distribution of white matter abnormalities in MCI is uneven, primarily concentrated in regions connected by associative cortices (such as the posterior cingulate fibers, corpus callosum, temporal lobe, frontal lobe, and parietal lobe white matter). The internal capsule, related to motor functions, and the occipital visual radiations are largely unaffected, appearing later in the progression of dementia (36). Moreover, the inclusion of the posterior limb of the internal capsule in the study was limited (only 8 studies), and the small sample size may reduce statistical power, highlighting the need for future studies to expand data for further validation in this region.

In summary, this study conducted a meta-analysis based on a large sample size and multiple regions to evaluate the diagnostic value of DTI for MCI. The increased sample size and enriched data from different brain regions have improved the reliability of the meta-analysis. By assessing the integrity of white matter fiber tracts in various brain regions, DTI technology can, to some extent, predict the progression of MCI to AD. This further confirms the impact of white matter damage in the progression from MCI to AD, providing an objective basis for early diagnosis and intervention in MCI.