The concept of vascular cognitive impairment (VCI) was proposed by Hachinski et al. in 1993. VCI is a cognitive impairment syndrome caused by cerebrovascular disease and risk factors for cerebrovascular disease, which manifests as single or multiple functional impairments; for example, in language, memory, calculation, and executive ability, and the symptoms generally worsen over time (1). The incidence of cognitive decline after stroke is much higher than the stroke recurrence rate, with as many as 30% of stroke patients showing varying degrees of cognitive decline (2). VCI is currently considered the second leading cause of dementia after Alzheimer's disease (AD) (3), and has a negative impact on patient quality of life, as well as representing a considerable burden to families and society. Therefore, it is particularly important to identify VCI occurrence at an early stage, and specific examination methods for early detection of VCI are currently lacking.

Diffusion tensor imaging (DTI) was initially assumed to function based on the principle of water molecule diffusion by Gaussian motion; however, experiments have demonstrated that, in biological tissues, complex cell components and structures hinder and limit water molecule diffusion; the imaging gradient strength, direction, and temporal profile affect sensitivity to diffusion and are commonly reduced to a single simplified parameter referred to as the b-value [unit: s/mm²]. Therefore, at high b values, this assumption is not applicable in brain tissue (4). Diffusion kurtosis imaging (DKI) is a recently developed method for measuring non-Gaussian diffusion of water molecules. DKI parameters include both diffusion parameters [fractional anisotropy (FA), mean diffusion (MD), axial diffusion (Da), and radial diffusion (Dr)] and kurtosis parameters [fractional anisotropy kurtosis (FAK), mean kurtosis (MK), axial kurtosis (Ka) and radial kurtosis (Kr)] (5). To obtain diffusional kurtosis maps, the logarithm of the signal decay following a diffusion-sensitized sequence is described, as by the following equation (6): ln[S(b)] = ln[S(0)]-bD+1/6(b²D²K)+0(b³), where S(b) is the signal intensity, D is the diffusion coefficient and K is the diffusional kurtosis. When K = 0, Gaussian formalism is recovered. This equation represents a cumulant expansion for the diffusion nuclear magnetic resonance signal.

DKI can reflect the high heterogeneity in the limited diffusion of water particles in a brain tissue injury area; hence, compared with diffusion-weighted imaging (DWI) and DTI, it better reflects the heterogeneity and complexity of tissues, and is more conducive to detecting subtle changes in tissue structure under pathological conditions (7). DKI parameters are very sensitive and can identify characteristic changes in many neurological diseases (8, 9). Further, changes in DKI parameters can be detected even before they are visible using conventional imaging, and the performance of DKI is better than that of conventional DTI (10, 11). Therefore, compared with the traditional DTI imaging, DKI has a wider scope for application to detect microstructural changes (9). To date, there have been few studies on the application of DKI in nervous system diseases, and relevant studies using DKI to investigate VCI are very scarce.

Here, we aimed to detect microstructural changes in patients with cognitive impairment after acute cerebral infarction using DKI, and to further evaluate the relationship between these parameters and Mini-Mental State Examination (MMSE) score, to provide neuroimaging evidence to assist in diagnosis of patients with VCI.

In this study, no significant differences in age, sex, years of education, or vascular risk factors were detected among the N-VCI, VCI, and NORM groups, and there were no significant differences in the size, location, or severity of cerebral infarction between VCI and N-VCI groups; hence, the effects of possible confounding factors on cognitive assessment were largely eliminated. In addition, there were no significant differences in DKI parameters among the three groups in the corona radiate, subcortical white matter, or cerebral cortex, which largely eliminated the influence of acute cerebral infarction on DKI parameters, increasing confidence in our findings. Therefore, in a situation where there were no significant differences in potential confounding factors among the three groups, we found that FA values in the genu of corpus callosum, the anterior limb of bilateral internal capsule, and bilateral medial temporal lobe were lower in the VCI group than the N-VCI and healthy control groups, while MD values in most ROIs were higher. In previous studies, Chen et al. (14), Zhou et al. (15), and Wei et al. (16) reported that mean FA from DTI was significantly lower and MD from DTI significantly higher in the VCI group than in N-VCI and healthy control groups. Our results are consistent with these previous studies, as we discovered that FA values from DKI were lower in some ROIs, and the MD values from DKI higher, in the VCI group than the N-VCI and healthy control groups. FA reflects the degree of arrangement and structural integrity of cell structures in a fiber bundle, while MD reflects the overall diffusion level and diffusion resistance of whole molecules (9, 15). Although the specific cellular mechanisms underlying differences in these two parameters in VCI are unclear (17, 18), based on our findings we speculate that the abnormal diffusion coefficient indices observed in patients with VCI can be explained by axon demyelination, loss of neurons, and increased water channel activity, leading to reduced water diffusion. This hypothesis is supported by earlier studies in which diffuse white matter changes, with myelin loss and axonal abnormalities, have been recorded in almost all types of vascular dementia (19).

Previous studies on VCI focused only on DTI, and highlighted diffusion indicators, such as FA, MD, DA, and Dr, while related studies (20) demonstrated that a combination of kurtosis and diffusion indices can better reveal brain tissue signal heterogeneity, facilitating detection of brain tissue microstructure changes. Therefore, compared with previous studies, we added kurtosis indices, including FAK, MK, Ka, and Kr, and further expanded the ROI, with the aim of revealing more comprehensive and concrete changes in brain microstructure in patients with VCI. We found that the MK values of some ROIs in the VCI group were lower than those in N-VCI and healthy control groups, while in bilateral thalamus, MK values were lower in VCI group than those in healthy control group, but higher than those in the N-VCI group. To date, there have been very few DKI studies on VCI, as most previous DKI studies have focused on mild cognitive impairment (MCI) and AD. Falangolael et al. (21), Wang et al. (22), and Liu et al. (23) found that, relative to healthy controls, kurtosis indices were lower in patients with MCI and AD. In our study, we discovered that MK values in some ROIs of patients with VCI were lower than those of the N-VCI and healthy control groups, while in others ROIs, MK values were higher in patients with VCI than those of the N-VCI group. While the exact meaning of and pathological basis for kurtosis have yet to be fully elucidated (24, 25). To date, the general reason for kurtosis has been considered microstructure complexity (26), where the more complex and intricate a structure is, the higher the degree of non-Gaussian water molecule diffusion, and the greater the MK value (27, 28). Some scholars have postulated that increases and decreases in kurtosis indicate potential disease changes or microstructure, and hence that kurtosis can provide rich microstructure information, with extensive analysis of pathological kurtosis potentially providing more information than diffusion coefficient data (29, 30). The advantages of DKI have been supported by studies for other clinical applications, where it detected more abnormal areas than DTI (31, 32). In this study, we hypothesized that kurtosis indices could provide more specific indications of VCI pathology. Neuroinflammation, neuronal swelling, and glial proliferation may all influence brain tissue microstructural complexity in VCI, in addition to demyelination and axonal degeneration. Therefore, we predicted that important changes in kurtosis would be able to distinguish early changes involved in VCI.

Previous studies had identified certain correlations between DKI parameters and cognitive function (30, 31); Bozzali et al. (33) and Morikawa et al. (34) reported a significant correlation between FA and MMSE score, while Yoshiura et al. (35) demonstrated a correlation of MMSE score with MD, rather than FA. Our findings demonstrate that FA, FAK, MK, Ka, and Kr are positively correlated with MMSE score in most ROIs, while MD, Da, and Dr were negatively correlated with MMSE score. The strong associations between MMSE score and DKI parameters suggest that patient cognitive impairment in VCI may be due to brain tissue axonal loss or demyelination, as well as other microstructural alterations; however, interestingly, our study revealed bilateral inconsistency in the microstructural changes of the frontal lobes and their correlations with MMSE score. Microstructural changes in the left frontal lobe are more obvious than those in the right in patients with VCI, which may reflect the frontal lobe asymmetry routinely detected in healthy people. A study of asymmetry of the human brain showed a larger total surface area in the right than the left cerebral hemisphere (36); therefore, we speculate that the left frontal lobe is more vulnerable to microstructural changes in patients with VCI.

In addition, we found that splenium of corpus callosum MD had the strongest correlation with MMSE score (correlation coefficient, −0.652), further confirming that MD has better diagnostic value for VCI than other DKI parameters. Moreover, we conducted ROC curve analysis, which showed that, when the threshold MD value of the splenium of corpus callosum was set at 0.9915, the diagnostic AUC value of VCI was 0.964, with sensitivity and specificity values of 91.4 and 100%, respectively. No previous report has identified splenium of corpus callosum MD as a potential diagnostic indicator of VCI. It is established that acute cerebral infarction can lead to ischemia and hypoxia of tissue across the whole brain, leading to dysfunctional myelinogenesis and nerve fiber damage. The corpus callosum is made up of fibers that originate from large pyramidal neurons and is one of the most extensively myelinated areas of the brain. Therefore, pathological changes often occur in the corpus callosum, even when ischemia or hypoxia is mild. Previous studies reported that the strongest associations between abnormal DTI parameters and cognitive decline were in the corpus callosum and corona radiata (14, 37–41). In our study, we also found that abnormalities of the corpus callosum and DKI parameters were strongly represented.

Of all DKI parameters in all ROIs, we observed that MD was significantly negatively correlated with MK, Da with Ka, Dr with Kr, and that FA was significantly positively correlated with FAK, which led us to hypothesize that changes in diffusivity were accompanied by alterations in diffusional non-Gaussianity. Liu et al. (23) found that, in addition to a correlation between Da and Ka, there were correlations between Dr and Kr, MD and MK, and FA and FAK in patients with cerebrovascular disease complicated with MCI. Gong et al. (42) found that MD and MK, Da and Ka, and Dr and Kr were significantly negatively correlated when observing changes in white matter and gray matter DKI parameters in patients with MCI and AD. The combination of kurtosis and diffusion in the DKI coefficient matrix provides two perspectives on microstructure by integrating non-Gaussian and Gaussian processes into one model. Hence, combination of these two features can represent the whole microstructure. Although variations in kurtosis parameters can be partly attributed to radial or axial variation of the absolute diffusion coefficient, they cannot be completely explained by these variations. Therefore, while kurtosis parameters represent the geometric distribution of diffusion, the neurobiological mechanisms leading to changes in kurtosis remain unclear.

This study has certain limitations. First, our preliminary findings were primarily used to determine whether DKI is useful as a new imaging tool for assessing changes in brain microstructure in patients with VCI; however, our sample size requires expansion and long-term longitudinal follow-up studies are needed to verify the long-term diagnostic significance of DKI parameters in patients with VCI. Second, because of the limited number of patients with VCI included in this study, we did not conduct further analysis according to VCI classification; therefore, we need to expand the sample size to allow VCI grading and more accurate evaluation of the significance of DKI parameters in patients with VCI. Third, although interesting changes in kurtosis parameters were observed in the frontal lobe, parietal lobe, internal capsule, corpus callosum, head of caudate nucleus, thalamus, and medial temporal lobe, the potential pathophysiological significance of these parameters requires further study. Finally, previous studies (43, 44) had combined electroencephalogram (EEG) with DTI, and provided a methodology to access specific networks of subcortical fiber tracts subserving the maintenance of interhemispheric resting state coherence in the human brain. Therefore, we can further combine DKI with EEG, that may be a more accurate method to evaluate patients with VCI.

In conclusion, pathological changes in some brain regions may underlie cognitive impairment after acute cerebral infarction, especially the splenium of corpus callosum. These preliminary results suggest that, in patients with VCI, DKI may be useful for assessing microstructural tissue damage. For early diagnosis of changes in white matter microstructure in patients with VCI, DKI provides a comprehensive imaging reference, and DKI-derived parameters may be a feasible means of evaluating patients with VCI. In addition, some DKI parameters are significantly correlated with patient clinical evaluation scores. Therefore, DKI-derived parameters may be a viable method to evaluate patients with VCI, and to predict potential brain structure imaging biomarkers of VCI progress.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Second Hospital of Dalian Medical University. The patients/participants provided their written informed consent to participate in this study.

LF, FEI, and YF contributed to the preparation of the manuscript. LF, XC, HY, contributed to the data collection. LF, ZW, CT, XC, YM, and YG contributed to data analysis and interpretation. LF, YD, CY, and YG contributed to the experimental design and manuscript revision. All authors contributed substantially to this work and approved the final manuscript.

This study was supported by the National Natural Science Foundation of China 81800823, Natural Science Foundation of Liaoning Province 20180530015, United Fund of the Second Hospital of Dalian Medical University and Dalian Institute of Chemical Physics, Chinese Academy of Sciences (UF-ZD-202012).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.