Volume 17 - Issue 5

Mini Review Biomedical Science and Research Biomedical Science and Research CC by Creative Commons, CC-BY

Biomarkers of Neurodegenerative Disease using Diffusion Magnetic Resonance Imaging

*Corresponding author: Yuya Saito, Department of Radiology, Juntendo University Graduate School of Medicine, Tokyo Japan

Received: December 05, 2022; Published: December 21, 2022

DOI: 10.34297/AJBSR.2022.17.002392


With the significant global growth in the number of aging societies, neurological illnesses have become more prevalent. Urgently required are biomarkers that may be utilized to identify pathological alterations prior to the onset of severe neuronal loss and hence permit early intervention with disease-modifying treatment approaches. Diffusion magnetic resonance imaging (MRI) is a promising technique that can be used to infer microstructural characteristics of the brain, including microstructural integrity and complexity, as well as axonal density, order, and myelination, by utilizing water molecules that are diffused within the tissue, with displacement at the micron scale. For assessing the pathophysiology of neurodegenerative disorders, diffusion tensor imaging is the most used diffusion MRI method. New methods, such as neurite orientation dispersion and density imaging, diffusion kurtosis imaging, and free-water imaging, have been developed to circumvent the limitations of diffusion tensor imaging. This article presents an overview of these technologies and their potential as biomarkers for the early diagnosis and development of significant neurodegenerative illnesses.

Keywords: Biomarker, Diffusion Kurtosis Imaging, Diffusion Tensor Imaging, Free-water Imaging, Neurite Orientation Dispersion and Density Imaging, Alzheimer’s Disease, Parkinson’s Disease


The frequency of neurodegenerative illnesses is rising in tandem with the fast aging of the world’s civilizations. Alzheimer’s disease (AlzD) and Parkinson’s disease (PD), two of the most prevalent neurodegenerative illnesses, are estimated to affect 35 million [1] and 6 million [2] persons worldwide, respectively. Importantly, it is anticipated that the prevalence rates of neurodegenerative disorders would climb even more quickly as the world population ages, given that aging is a key risk factor for these illnesses, signifying a rising public health concern. Neurodegenerative disorders presently have no curative treatments; consequently, the discovery of disease-modifying medications that may halt the progression of underlying pathological alterations is eagerly awaited. Urgently required are biomarkers that may be utilized to identify pathological alterations prior to the onset of severe neuronal loss and therefore permit early intervention with disease-modifying treatment methods. Among the several possible biomarkers that have been suggested, magnetic resonance imaging (MRI) is an outstanding candidate biomarker, since it provides a potent method for noninvasive in vivo brain examination. Specifically, diffusion MRI is promising because it can infer microstructural characteristics of the brain, such as microstructural integrity and complexity, as well as axonal density, order, and myelination, by utilizing water molecules that diffuse within the tissue with micron-scale displacement [3]. Neurite orientation dispersion and density imaging (NODDI), diffusion kurtosis imaging (DKI), and free-water imaging (FWI) have been developed as alternatives to DTI. Diffusion tensor imaging (DTI) is the most commonly used diffusion MRI technique to assess pathophysiology in neurodegenerative diseases (see Table 1 for summary). This article presents an overview of these technologies and their potential as biomarkers for early neurodegenerative disease diagnosis and progression prevention.

Diffusion MRI Techniques


The biologically structured structure of the brain, including axons, myelin, cerebrospinal fluid (CSF), and neuronal soma and dendrites, influences water diffusion in the brain. Isotropic diffusion, in which water diffuses equally in all directions (e.g., in cerebrospinal fluid (CSF) and gray matter (GM)), may be distinguished from anisotropic diffusion, in which water diffusion is unidirectional (e.g., in white matter (WM)). The uses and efficacy of DTI in brain illnesses have been discussed earlier [3,4]. DTI characterizes brain structures on the basis of water diffusion using four metrics: fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) [5].

DTI is applied extensively in neurodegenerative illnesses such as Alzheimer’s disease [6-8], Parkinson’s disease [9-12], multiple sclerosis [13-15], stroke, and traumatic brain injury [16-18]. However, a number of deficiencies restrict its clinical applicability. First, DTI can not account for the non-Gaussian diffusion characteristics of water molecules in some biological tissue components, such as the cell membrane and myelin sheath, which results in biological limitations [19,20]. Therefore, DTI is incapable of detecting microstructural changes in GM, which is mostly constituted of neuronal cell bodies and displays greater isotropic water transport [19,20]. Second, DTI assumes that each voxel contains a single tissue compartment, which creates a partial volume effect due to the presence of extracellular free water, such as CSF [5] and significantly affects the accuracy of DTI measurements at the GM/WM boundary [21,22] and that of the GM voxels contaminated by CSF [23,24]. Third, DTI parameters lack diseasespecific and pathological information [3]. For instance, it is unclear whether a drop in FA means a reduction in axon density or axonbundle cross-section, and the interpretation of DTI parameters [25- 28] is contentious. In conclusion, the DTI model oversimplifies the brain’s anatomy. Despite the fact that WM voxels include crossing fibers and account for up to 90% of all adult brain voxels [29,30], DTI reflects only a single major direction; thus, FA diminishes in such voxels even in normal brain tissue [31


A mathematical extension of DTI, DKI was suggested. Kurtosis is dimensionless [19,20] and assesses the degree of non-Gaussian distribution in water diffusion inside a voxel. Consequently, DKI identifies the limitation of water transport caused by the intricacy of brain tissue components such as the cell membrane and myelin sheath [32]. The greater the diffusion kurtosis, the greater the deviation of water molecule diffusion from the Gaussian distribution, indicating a more constrained diffusion environment. In contrast, a lower diffusion kurtosis indicates less limited diffusion, such as in neuronal degeneration [33].

DKI needs at least three b-values and 15 diffusion gradient directions for a more complicated model [34]. DTI requires at least two b-values and six diffusion gradient directions. DKI measures the condition of brain tissue via the use of three kurtosis metrics: mean kurtosis (MK), axial kurtosis (AK), and radial kurtosis (RK) (RK). DKI has been used to assess neurodegeneration in WM with complex architectures, including voxels with crossing fibers [12,35-37]. Moreover, DKI parameters reflect the restriction of water diffusion in anisotropic as well as isotropic environments, in contrast to DTI parameters, which assume non-restriction of water diffusion; thus, the utility of DKI has also been demonstrated for the evaluation of microstructural changes in GM, which is primarily composed of neuronal cell bodies and exhibits isotropic water diffusion [20,28,38,39].

Despite its advantage over DTI for the assessment of diseased brain alterations, DKI has numerous drawbacks. First, the acquisition time of DKI (about 10 minutes) is greater than that of DTI (roughly 5 minutes) [19,40], which reduces its clinical value since the more sophisticated DKI model needs more parameters than DTI. Because neither model involves biophysical assumptions [19,20], neither DKI nor DTI can explain disease-specific and pathological alterations such as the density, dispersion, and crosssection of axons or dendrites in neurons.


As explained in Section 2.1, DTI can reliably estimate tissuespecific indices only in voxels containing a single kind of brain tissue, but cannot quantify tissue-specific indices in voxels polluted by extracellular free water, such as CSF [5,21-24]. FWI was first suggested for the distinct interpretation of microstructures inside brain tissue and extracellular fluid within the same voxel [41]. FWI is a two-compartment (or bi-tensor) model consisting of anisotropic brain tissue and isotropic free water. After eliminating extracellular free water contamination, FWI calculates the free water volume fraction (FW) map with the free water compartment and the traditional DTI map with the tissue compartment [41]. Thus, FWI may increase the accuracy of single-tensor DTI indices and analyzes particularly the microstructures of brain tissue after the removal of free water. Moreover, the free-water map is regarded as a potential biomarker for discriminating between neuronal degeneration and the increase of free water in the extracellular space, which is linked to neuronal disorders such as neuroinflammation [41- 43]. FWI can quantitatively measure the degree of edema and atrophy, as well as neuroinflammation, and hence may contribute to a better understanding of the pathophysiology underlying neurodegenerative illnesses such as Parkinson’s disease [42], schizophrenia [44,45], and depression [46].

FWI may be estimated from clinically common single-shell diffusion data using the same method as DTI [41], and its accuracy is equivalent to that obtained from multi-shell diffusion data [47]. Nonetheless, the estimate of FWI is highly dependent on the regularization restrictions of spatial smoothing, which may result in decreased sensitivity for modest diseases [41,47]. Other approaches that do not need spatial regularization may be able to reconstruct more accurate FWI indices [47,48] when multishell diffusion acquisition is employed to estimate FWI.


NODDI was designed to provide a more precise description of brain tissue microstructures than signal representations like DTI and DKI. NODDI simulates three compartments of brain tissue. The intracellular compartment corresponds to the space bordered by the neurite membrane, the extracellular compartment to the space surrounding by neurites, and the isotropic water pool to the space filled by CSF [49]. Before NODDI, other methods that assume numerous compartments, such as the composite hindered and limited water diffusion (CHARMED) model [50,51], were created. NODDI’s innovation, however, is in its capacity to reveal the properties of angular variances of neurites inside each voxel. NODDI not only quantifies the isotropic volume fraction (ISOVF, volume fraction of extracellular isotropic free water), but also the orientation dispersion index (ODI, index of intracellular neurite dispersion) and the intracellular volume fraction (ICVF, neurite density) [49,52]. After the elimination of extracellular free water from the voxel, ICVF and ODI may thereby characterize the biological microstructures of axons and dendrites. In addition, an increased ISOVF in WM may potentially account for neurodegeneration accompanied by an increase in extracellular isotropic fluid, such as in neuroinflammation [49].

Consequently, NODDI may be utilized to represent microstructures more precisely than signal representation approaches [49,53,54], notwithstanding NODDI’s limitations. First, whereas ODI is very accurate at predicting single-shell diffusion data, ICVF and ISOVF need multi-shell diffusion data with a minimum of two shells [49,55], comparable to DKI. Therefore, a lengthy acquisition period is required for estimating neurite density and extracellular fluid. Second, while the anisotropic orientation dispersion of neurites induced by bending and fanning fibers is seen across the whole brain [56], NODDI cannot assess the complicated anisotropic neurite dispersion since it models only the isotropic neurite dispersion.

Conclusion and Future Directions

Neurodegenerative illnesses have been elucidated by the use of advanced diffusion MRI methods, such as FWI, DKI, and NODDI, which give novel information on brain microstructures. Due to the absence of clinical proof of their efficacy, these sophisticated approaches have not yet been used in clinical settings, such as the regular use of DWI to evaluate myocardial infarction. In addition, unlike the measurement of hippocampal volume, a biomarker for Alzheimer’s disease that is used to reduce the sample size and cost of clinical trials for the detection of neurodegenerative changes [57], the evidence regarding the utility of advanced diffusion MRI-based biomarkers in neurodegenerative diseases is insufficient. Clinical trials are impeded by the high costs involved with conducting them. Consequently, the cost-effectiveness of newer diffusion MRI methods hinders the gathering of clinical data.

To accomplish the practical use of sophisticated diffusion MRI methods for the diagnosis of neurodegenerative disorders, many obstacles must be overcome. First, the link between pathological alterations in neurodegenerative illnesses and improved diffusion MRI measures is yet unknown. DKI, FWI, and NODDI only model and forecast brain microstructures via diffusion MRI, and it is uncertain to what degree these models can reflect and explain particular neurodegenerative illnesses. In order to understand the association between pathological results and the advanced diffusion MRI metrics of FWI, DKI, and NODDI, more investigations on neurodegenerative disorders in postmortem human tissues or animal models are required. Second, the repeatability and dependability of the findings of research using sophisticated diffusion MRI methods are extremely poor due to the small sample sizes and thus low statistical power. Therefore, it is necessary to show the efficacy of FWI, DKI, and NODDI as biomarkers for neurodegenerative illnesses based on robust evidence from multisite studies with greater sample numbers to increase their statistical power. Although a number of large-scale multi-site investigations are now ongoing, MRI scanners and acquisition settings are very variable and site-dependent [58]. These variations across research locations may reduce the repeatability and reliability of sophisticated MRI diffusion investigations. Andica et al. [59] assessed the scan-rescan and inter-vendor repeatability of DTI and NODDI using two 3-T MRI scanners from two manufacturers. The scan-rescan coefficient of variation of NODDI measurements with both scanners was close to that of DTI metrics (0.2% to 3.8%). However, the inter-vendor CoV for NODDI measurements was greater than the scan-rescan CoV (2.3-14%). In addition, the inter-sequence variability of DTI measurements for three distinct sequences revealed that the CoVs for FA and MD were 5.45-7.34% and 1.72-5.55%, respectively [60]. In addition, Kamagata et al. [61] assessed the inter-site reproducibility of DTI measurements using identical 3-T MRI scanners and acquisition conditions at two distinct locations. According to the authors, the CoV of DTI varied from 0.6% to 5.6%. Consequently, changes in diffusion MRI metrics induced by site differences, such as MRI scanners and acquisition settings, may impair their statistical power, resulting in poor repeatability and reliability in multi-site studies using sophisticated diffusion MRI methods [58]. Specifically, changes in diffusion MRI metrics in neurocognitive and psychiatric disorders are subtle (approximately 5-6%) compared to healthy controls and on the same order as that of site difference; thus, it is challenging for a multi-site study to detect pathological changes in patients with neurocognitive and psychiatric disorders [62-64]. Therefore, it is vital to decrease inter-site variability in diffusion MRI metrics by harmonizing multi-site diffusion MRI data and standardizing MRI procedures, including MRI scanners and acquisition settings. Several harmonization strategies for diffusion MRI, such as the combined association test ComBat [65], linear regression based on rotation-invariant spherical harmonics [66], and the deep learning approach [67], have been suggested to decrease the variance across MRI scanners and protocols. ComBat utilizes the regression of variables for the data harmonization of diffusion MRI measurements using an empirical Bayesian inference. The linear rotation invariant spherical harmonics are used for diffusion MRI signal harmonization and mapping of diffusion MRI data from a target location to a reference site. The deep learning harmonization technique optimizes neural network parameters using diffusion MRI signals received from target and reference locations during the learning phase, and then uses the trained neural network to harmonize diffusion MRI data. These harmonization strategies not only eliminate undesirable fluctuations in DTI measures caused by site differences, but they also retain biological diversity caused by age and gender.

Another issue is the unknown link between neurodegenerative disease-induced pathological alterations and advanced diffusion MRI measurements. DKI, FWI, and NODDI can only model and forecast brain microstructures using diffusion MRI, and it is uncertain to what degree these models represent and explain particular neurodegenerative illnesses. In order to understand the association between pathological results and the advanced diffusion MRI metrics of DKI, FWI, and NODDI, more research of neurodegenerative illnesses in postmortem human tissues or animal models are necessary. Resolving these constraints should result in the practical use of enhanced diffusion MRI methods as diagnostic biomarkers for neurodegenerative disorders.

Recently, frameworks that integrate diffusion tensor MRI and relaxometry have been developed to increase their specificity for quantifying myelin and axonal characteristics regardless of the complexity of fiber organization inside the voxel, even in the presence of crossing fibers [68,69]. In addition, accumulating evidence implicates excessive iron accumulation in the pathogenesis of neurodegenerative disorders [70]. Quantitative susceptibility mapping is a potential imaging tool for the complete study of iron distribution in the brain. By assessing various biological tissue qualities, the use of the diffusion tensor-relaxometry framework or the combination of modern diffusion MRI methods and quantitative susceptibility mapping may offer a more complete picture of neurodegenerative disorders.


Sign up for Newsletter

Sign up for our newsletter to receive the latest updates. We respect your privacy and will never share your email address with anyone else.