Microstructural disruption of the right inferior fronto‐occipital and inferior longitudinal fasciculus contributes to WMH‐related cognitive impairment

Abstract Aims White matter hyperintensity (WMH) is the most common neuroimaging manifestation of cerebral small vessel disease and is related to cognitive dysfunction or dementia. This study aimed to investigate the mechanism and effective indicators to predict WMH‐related cognitive impairment. Methods We recruited 22 healthy controls (HC), 25 cases of WMH with normal cognition (WMH‐NC), and 23 cases of WMH with mild cognitive impairment (WMH‐MCI). All individuals underwent diffusion tensor imaging (DTI) and a standardized neuropsychological assessment. Automated Fiber Quantification was used to extract altered DTI metrics between groups, and partial correlation was performed to assess the associations between WM integrity and cognitive performance. Furthermore, machine learning analyses were performed to determine underlying imaging markers of WMH‐related cognitive impairment. Results Our study found that mean diffusivity (MD) values of several fiber bundles including the bilateral anterior thalamic radiation (ATR), the left inferior fronto‐occipital fasciculus (IFOF), the right inferior longitudinal fasciculus (ILF), and the right superior longitudinal fasciculus (SLF) were negatively correlated with memory function, while that of the anterior component of the right IFOF and the posterior and intermediate component of the right ILF showed significant negative correlation with MMSE and episodic memory, respectively. Furthermore, machine learning analyses showed that the accuracy of recognizing WMH‐MCI patients from the WMH populations was up to 80.5% and the intermediate and posterior components of the right ILF and the anterior component of the right IFOF contribute the most. Conclusions Changes in the properties of DTI may be the potential mechanism of WMH‐related MCI, especially the right IFOF and the right ILF, which may become imaging markers for predicting WMH‐related cognitive dysfunction.


| INTRODUC TI ON
Cerebral small vessel disease (CSVD) is an age-related clinical syndrome, manifesting as abnormal mood and gait, lacunar infarction, cognitive dysfunction, and Parkinson's disease. MRI manifestations include lacunar infarction, white matter hyperintensities (WMHs), perivascular spaces, microbleeds, and brain atrophy. 1 White matter hyperintensities is perceived as the most common neuroimaging manifestation because WMH is visible in 80% of healthy people over 60 years and almost all the people over 90 years. 2 Different degrees of demyelination, gliosis and loss of fibers and oligodendrocytes are shown in the pathological examination of WMH. 3,4 Recent studies have found that WMH can lead to vascular cognitive impairment, and in some, may ultimately progress to dementia. 5 The underlying mechanism is highly controversial. Vascular risk factor load, damage to neurotransmitter systems, interruption of prefrontal subcortical loops, and cerebral hypoperfusion were proposed to explain the correlation between WMHs and cognitive decline. [5][6][7][8][9][10][11] However, patients with similar visual extensive WMH always manifested a variable severity of cognitive dysfunction and affected different cognitive domains. 12 Longitudinal follow-up studies also showed further deterioration of cognitive function over time in WMH patients with mild cognitive impairment. 13 Furthermore, loss of microstructural integrity of normal-appearing white matter (NAWM) was associated with executive function, 14 which further suggested that WMH might be an extreme case of continuous spectrum of white matter (WM) damage. White matter tract disruption in diffusion tensor imaging (DTI) may lead to disconnections among cortico-cortical or cortico-subcortical pathways vital for some cognitive function. The socalled "disconnection hypothesis" may play a role in WMH-related cognitive impairment. 15,16 Diffusion magnetic resonance imaging (dMRI) is appropriate for WM microstructure study. 17 Diffusion tensor imaging (DTI) is currently the only noninvasive method that can effectively observe and track the WM fiber tracts in a living human brain. 18,19 Fractional anisotropy (FA) and mean diffusivity (MD) are two quantitative measures of DTI that, respectively, detect the anisotropy and overall displacement of water molecule diffusion. 20,21 One approach to analyze DTI is voxel-based analysis (VBA), the core of which is making measurements of FA or other diffusion metrics in specific regions of interest (ROIs), thus conducting group comparisons. For example, Della Nave reported that a large cluster of increased MD in the corpus callosum and pericallosal WM was associated with impaired motor function. 22 It is noticeable that VBA is extremely sensitive to registration errors but may fail to achieve sufficient precision because of the diversity of tract sizes and shapes among individuals. 23,24 Tract-based spatial statistics (TBSS) emerges then, which integrates selective voxels onto the nearest location on a pseudoanatomical WM skeleton and reduces the residual misalignment by 10% without spatial resolution loss. 25 By using TBSS, Otsuka found that reduced diffusion anisotropy of the corpus callosum in patients with extensive leukoaraiosis may explain global cognitive impairment. 12 However, it still fails to guarantee that any voxel corresponds to the same tract across individuals as mean FA equalizes the particularity by dispersing the original changes in specific sections of specific tracts to the whole bundle, with the average voxel value significantly influenced by the artificial split of anatomical locations. 25 29 The effect of hypertension on WM integrity and the correlation between the destruction of structural architecture of WM and cognitive impairment have also been revealed by AFQ.
In view of its several advantages, we also consider AFQ as an ideal analytic method in our study. We hypothesized that DTI parameters vary among different positions on the same fiber bundle and the localization-specific properties in WM integrity may contribute to WMH-related cognitive impairment. Considering that vascular cognitive impairment is recognized as a progressive condition from normal cognitive status to frank dementia, 30 early prediction and intervention play a pivotal role in the prevention of dementia, thus granting our study high significance.  The p value was obtained by The p value was obtained by one-way ANOVA. *Indicates a statistical difference between groups, P < .05.

| Neuropsychological measurement
All subjects underwent a standardized neuropsychological assessment protocol performed by an experienced neuropsychologist. General cognition was evaluated by the Beijing version of the Montreal Cognitive Assessment (MoCA-BJ) and MMSE. 33 As described in our previous study, MoCA-BJ was used to detect WMH-MCI according to the education level. 33  was used to evaluate the multiple cognitive domains of episodic memory, language, executive function, and information processing speed. The raw examination scores were Z-transformed to calculate each cognitive domain score.

| Magnetic resonance image preprocessing
For diffusion images, the data preprocessing was carried out by FSL 5.0.9 software (Oxford Centre for Functional Magnetic Resonance Imaging of the Brain, University of Oxford; https ://www.fmrib.ox.ac. uk/fsl/). The preprocessing included the following steps: DICOM-to-NIfTI format conversion, registering DWI images (b = 1000 s/mm 2 ) to the non-DWI image (B0), eddy current and head motion correction, and then nonbrain tissue exclusion. After preprocessing, wholebrain images of diffusion metrics, including FA, MD, axial diffusivity (AD), and radial diffusivity (RD), were obtained via DTIFIT command of FSL.

| Automated fiber quantification procedure
We identified whole-brain WM fiber tracts (20 major fiber tracts) and further quantified the diffusion metrics along the tract trajectory by applying the AFQ package. 28 Table 1.
Because AFQ uses strict criterion for tract identification, it did not successfully identify all 20 WM tracts in each participant. 34 We excluded 4 fiber tracts, the bilateral cingulum hippocampus (CH) and bilateral arcuate fasciculus (AF) which did not identify in large portion of subject (Table 1).

| Statistical and machine learning analyses
Demographic characteristics (including age, years of education, and gender distribution) and cognitive assessment were compared using one-way analysis of variance (ANOVA) or the chi-squared (χ2) test in SPSS software (Version 22). Significance was set at P < .05.
To examine group difference of WM tracts, we calculated DTI metrics (FA, MD, AD, and RD) of each fiber tract by averaging diffusion values of 100 nodes along each WM tract and performed the ANOVA to determine the between-group differences. Age, years of education, and gender were controlled for as confounding covariates.
Next, the point-wise analyses were based on the "Randomize" program of FSL software. Age, years of education, and gender were included as covariates in the general linear model (GLM). Familywise error (FWE) correction was applied to the 1600 points (16 fibers × 100 points) and a corrected significant level at 0.05. 35 Then, within each fiber, only more than or equal to three adjacent nodes were reported. 36 To assess the associations between WM integrity and cognitive performance, partial correlation was performed using

| Demographic and clinical characteristics
Demographic and clinical characteristics for the HC and WMH subgroups (WMH-NC and WMH-MCI) are provided in Table 2. There was no significant difference for age, years of education, and gender distribution between the subgroups (P > .05). The WMH-MCI subgroup showed poorer performances on MMSE (P < .001), MoCA-BJ (P < .001), episodic memory (P < .001), information processing speed (P < .011), and executive function (P < .029) than the other subgroups (the detailed cognitive assessment is shown in Table 2).

| Group difference in WM tract and pointwise levels
Between-group difference of WM tract and point-wise alterations was determined by mean diffusion metrics (FA, MD, AD, and RD) with AFQ.

| FA
We found that mean FA values of WM tracts significantly changed between the three groups in the bilateral anterior thalamic radiation (ATR), forceps minor, bilateral IFOF, and left ILF (Table S1 and Figure   S1).     Table S4.  Table S3.   Table S5.  Table S3.

| Correlations between altered diffusion metrics and cognition
For the altered WM fiber tracts in Table S1, we found no significant correlation between FA and cognitive performance in WMH-MCI.
Next, we estimated the relationships between the other three diffusion metrics (MD, AD, and RD) and the cognition in WMH-MCI.
We summarized the correlations between MD values (the entire WM fiber tract in Table S2 and point-wise level in Table S3 and Table 3 shows the results of the RF analyses for identifying WMH- are presented in Figure 3A and Figure S7.

| D ISCUSS I ON
Diffusion tensor imaging is the only noninvasive method to measure the microstructural integrity of brain tissue by detecting the diffusion of water molecules in it. Fractional anisotropy, MD, AD, and RD are four common indicators. White matter hyperintensities is the most common neuroimaging manifestation of CSVD, which is related to cognitive dysfunction or dementia. Studies that focused on NAWM have also confirmed that the change in DTI metrics of NAWM always precedes the development of WMH and is associated with MCI. 14,38 Combined with the emergence of WMH penumbra theory, 39 we propose that direct visual WMH may be a manifestation of more extensive and subtle WM microstructural degeneration and those changes of the properties in DTI may be the potential mechanism of WMH-related MCI, as stated in the theory of cortical "disconnection" hypothesis. 15,16 To further explore and verify the mechanism, many efforts have been made using VBA or TBSS. 12,22 However, their results are inconsistent and controversial and were obtained from a perspective of brain region or the entire fiber tract only. So we conducted the research, and in this study, we Furthermore, four bundles showed a statistically significant difference between the WMH-MCI and WMH-NC, that is, the right ATR, the right IFOF, the right ILF, and the left SLF. This suggests that MD is more sensitive than FA in detecting the degree of WM damage, consistent with the study on Alzheimer's disease. 40,41 The development of WMH is partly caused by focal ischemia, which may result in a decrease in tissue density and an increase in water diffusivity while maintaining underlying directional structure, and those outcomes cause an increase in MD when FA remains unchanged. 41,42 Partial correlation analysis showed that the MD values of several fiber bundles (left ATR, right ATR, left IFOF, right ILF, and right SLF) were negatively correlated with memory function in the WMH-MCI group, which indicated that the more serious the damage with WM microstructure, the worse is the memory ability. Many efforts have been made to explore the correlation between WM microstructure and cognitive function, but some controversies remain. One reason is inconsistencies in detailed anatomical definitions like IFOF whose origin and termination of fiber tract have not been determined accurately. 43 It was widely accepted that the IFOF connected the fronto-marginal gyrus and lateral orbito-frontal gyrus with the inferior occipital gyrus, the inferior part of the middle occipital gyrus, and lingual gyrus, 44  Thomas's finding that only the reduction in tract integrity of the right ILF and the right IFOF can lead to the face recognition impairment suggested that the right hemisphere is more prominent in some cognitive domains. 54 We also found that the right ILF showed a significant negative correlation with episodic memory in the posterior and intermediate component (nodes 26-58). The ILF links the anterior temporal lobe with the extrastriate cortex of the occipital lobe, running along the lateral and inferior wall of the lateral ventricle and was found to be correlated with semantic autobiographical memory. 55 Diffusion tensor imaging studies found that damaged microstructures of the ILF were significantly correlated with decreased memory function, 48,55 indicating that memory function depends on the connection integrity of the temporal lobe with other lobes such as the occipital lobe.
What is interesting in Zemmoura's finding is that it is the posterior portion of the ILF instead of the anterior portion that plays a prominent role in reading impairment, 56 57 Besides, many studies have confirmed the relationship between the integrity of ILF/IFOF and multiple cognitive functions including memory, reading, attention, and visual processing. 43,48,58,59 In conclusion, both the level of fiber bundles and the node level suggest that ILF and IFOF are key WM fiber tracts mediating certain cognitive functions and can be used as imaging markers for early recognition of WMH-related cognitive impairment.
In addition to FA and MD, other diffusion metrics like AD and RD were also employed in our study. AD reflects diffusion parallel to axonal fibers which may reflect axonal injury and RD captures perpendicular diffusion which is linked to demyelination. 42  research, in which the clinical progress of the same subject can be traced during his lifespan, the conclusions will be more convincing.
Last but not the least, the mechanisms tackling how the right IFOF and the right ILF are influenced and why they are the first two tracts to get impaired are not well illustrated in this paper. Further studies are warranted to support and substantiate findings from the present study.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.