Background
Multiple sclerosis (MS) is a chronic disorder of the CNS, characterized by focal white matter (WM) plaques along with diffuse normal appearing WM (NAWM) damage and cortical demyelination [
1]. Diffusion tensor imaging (DTI) is one of the most widely used methods in detecting microstructural abnormalities based on water diffusion measures with the assumption that the diffusion displacement of water molecule in an unrestricted environment has a Gaussian approximation [
2].In reality, water molecules often show non-Gaussian diffusion due to the presence of barriers of cell membranes, axon sheaths, and water compartments in biological tissues [
3]. So it is thought that DTI may not be capable to provide accurate values at dense intersections of fiber tracts [
4]. In contrast, as a clinically feasible extension of DTI, diffusion kurtosis imaging (DKI) has been proposed to characterize the deviation of water diffusion in neural tissues from Gaussian diffusion [
5,
6]. Both diffusion parameters including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (Da), radial diffusivity (Dr) and kurtosis parameters including mean kurtosis (MK), axial kurtosis (Ka), radial kurtosis (Kr) could be obtained from DKI data. DKI can be regarded as a more sensitive indicator of diffusional heterogeneity and can be used to investigate abnormalities in tissues with isotropic structure [
6,
7].
The sensitivity of DKI has been evaluated in age-related diffusion patterns in the prefrontal brain [
8], reactive astrogliosis in traumatic brain injury [
9], and cuprizone-induced demyelination in mice [
10], which showed better demonstration of microstructural changes than with DTI. However, there were few studies to validate the merits of DKI in evaluating patients with MS [
11,
12]. Tract-based spatial statistics (TBSS) provides a powerful and objective method to perform multi-subject comparisons [
13].
In this study, the microstructural alterations reflected by both DKI and DTI parameters in relapsing-remitting multiple sclerosis (RRMS) were investigated using TBSS. Our aim is to assess the performances of 11 commonly used parameters derived from DKI (MK, Ka, Kr, FA, MD, Da and Dr) and DTI (FA, MD, Da and Dr) in detecting microstructural abnormalities in RRMS.
Discussion
Although DTI has been widely used in investigating structural changes in the NAWM in MS [
16,
17], it may not provide accurate parameters at dense intersections of fiber tracts [
4]. In contrast, DKI can be used to quantify non-Gaussian diffusion, thus providing accurate parameters at dense intersection of fiber tracts [
6]. To our knowledge, there were only a limited number of studies using DKI in MS patients [
11,
12,
18,
19], Raz E et al. measured FA, MD, and MK values of the entire cross-sectional cord area, normal-appearing gray matter (NAGM) and WM in MS patients by DKI using region-of-interest (ROI) analysis, they thought that DKI could provide additional and complementary information to DTI on spinal cord pathology [
11]. In another research, DKI was used to evaluate diffusional changes in NAWM regions remote from MS plaques using ROI analysis, the results indicated that DKI might be an additional sensitive indicator for detecting tissue damage in MS patients [
12]. They concluded that DKI was sensitive for detecting tissue damage in MS patients and could provide information that was complementary to that of conventional DTI-derived metrics. However, most of these above-mentioned studies adopted ROI-based analysis, which had poor reproducibility of ROI positioning and only a limited number of specific regions can be examined. In contrast, the TBSS method used in this study was relatively a novel hypothesis-free and user-independent voxel-wise analysis.
In this study, TBSS analysis of both DKI and DTI derived parameters showed widespread WM damage in RRMS patients compared with healthy controls, which was consistant with previous studies using DKI [
19] or DTI [
20,
21]. Similarly to a research study using DKI in schizophrenia patients, we also observed that DKI-derived kurtosis and diffusion parameters had differernt sensitivity to detect abnormality in WM areas with different fiber architecture [
15]. Moreover, we found that the MK decrease in the WM of RRMS patients was predominantly caused by the Kr decrease, and the FA decrease was mainly driven by the increase of Dr.
FA measures anisotropic water diffusion and is proven to be most applicable for assessing WM regions with coherent fiber arrangement. However, it is not suitable for detecting diffusion changes of complex WM architecture, such as crossing fiber regions [
22,
23]. As the most characteristic parameter of DKI, MK measures the deviation of the diffusion displacement profile from a Gaussian distribution and enables to probe WM regions with complex fiber arrangement [
24]. Therefore, the combination of diffusion and kurtosis parameters may provide improved sensitivity and specificity in detecting alterations in various WM structures. This theoretical prediction has been validated by a previous study in schizophrenia patients [
15], and confirmed by our findings that altered diffusion parameters (especially reduced DKI_FA) were observed mainly in WM regions with coherent fiber arrangement (such as the corpus callosum and anterior limb of internal capsule). The percentage of abnormal DKI_FA voxels (74.1%) relative to the whole skeleton voxels was higher than that of DTI_FA (68.6%) in this study, which suggest that DKI_FA might have higher sensitivity than DTI_FA in detecting abnormality in WM regions, while reduced kurtosis parameters were mainly located in WM regions with complex fiber arrangement (such as the juxtacortical WM and corona radiate). The percentage of abnormal MK voxels relative to the whole skeleton voxels was 78.2%, which suggest that MK might have higher sensitivity than DTI in detecting abnormality in WM regions. Therefore, appropriate DKI derived parameters should be selected to probe altered diffusion pattern in specific WM regions in RRMS patients.
As we know, there is strong directional dependence of water distribution within myelinated WM tracts. However, once inflammation and demyelination occur, diffusivity will increase and directionality will decrease. The increase of diffusivity manifested as increase of Dr (diffusion perpendicular to the long axis) and Da (diffusion along the long axis). However, decreased Da was reported in some animal experiments [
25,
26]. In our opinion, these studies may not take into account the full complexity of pathological processes occurred in RRMS. The increase of Da found in our RRMS patients was consistent with a TBSS study using DTI in RRMS patients [
20]. The cause may be explained by severe decreases in axonal packing density which would lead to a whole increase in extracellular water, resulting in larger Dr increases and subsequent Da increases. Other reported reasons include fiber re-organization, increased axonal diameter and membrane permeability [
27,
28]. In our study, the percentage of abnormal DKI_Dr voxels (79.8%) relative to the whole skeleton voxels is significantly higher than that of DKI_Da (28.3%), which demonstrated that the increase of MD (mean diffusivity) and decreased FA mainly caused by the increased DKI_Dr. Similarly, this pattern of changes was also found in DTI_Da and DTI_Dr. Interestingly, when assessing the contribution of DKI_Ka and DKI_Kr in those regions showing significantly decreased DKI_MK, we found these were driven predominantly by decreases in DKI_Kr (76.5% vs DKI_Ka 53.5%). All these above-mentioned findings suggested demyelination might be regarded as a key factor among various pathological changes in RRMS. So Dr and DKI_Kr might be regarded as useful surrogate markers for reflecting pathological changes and improving clinical–radiological correlations in MS. Furthermore, Dr and DKI_Kr measured by TBSS might have great potential to be a MRI biomarker in monitoring remyelination in MS patients.
Acknowledgements
We would like to gratefully thank Jia Jia Zhu of Department of Radiology Affiliated Hospital, Anhui Medical University for his support and assistance.
We also would like to gratefully thank Pro. Chi-Shing ZEE, Keck hospital of USC for his kindly revised the manuscript finally.