High-Spatial-Resolution Diffusion MRI in Parkinson Disease: Lateral Asymmetry of the Substantia Nigra

Abstract

Background

Motor symptoms in Parkinson disease (PD) have exhibited lateral asymmetry, suggesting asymmetric neuronal loss in the substantia nigra (SN). Diffusion MRI may be able to help confirm tissue microstructural alterations in the substantia nigra to probe for the presence of asymmetry.

Purpose

To investigate lateral asymmetry in the SN of patients with PD by using diffusion MRI with both Gaussian and non-Gaussian models.

Materials and Methods

In this cross-sectional study conducted from March 2015 to March 2017, 27 participants with PD and 27 age-matched healthy control (HC) participants, all right handed, underwent MRI at 3.0 T. High-spatial-resolution diffusion images were acquired with a reduced field of view by using seven b values up to 3000 sec/mm². A continuous-time random-walk (CTRW) non-Gaussian diffusion model was used to produce anomalous diffusion coefficient (D_m) and temporal (α) and spatial (β) diffusion heterogeneity indexes followed by a Gaussian diffusion model to yield an apparent diffusion coefficient (ADC). Individual or linear combinations of diffusion parameters in the SN were unilaterally and bilaterally compared between the PD and HC groups.

Results

In the bilateral comparison between the PD and HC groups, differences were observed in β (0.67 ± 0.06 [standard deviation] vs 0.64 ± 0.04, respectively; P = .016), ADC (0.48 μm²/msec ± 0.08 vs 0.53 μm²/msec ± 0.06, respectively; P = .03), and the combination of CTRW parameters (P = .02). In the unilateral comparison, differences were observed in all diffusion parameters on the left SN (P < .03), but not on the right (P > .20). In a receiver operating characteristic (ROC) analysis to delineate left SN abnormality in PD, the combination of D_m, α, and β produced the best sensitivity (sensitivity, 0.78); the combination of D_m and β produced the best specificity (specificity, 0.85); and the combination of α and β produced the largest area under the ROC curve (area under the ROC curve, 0.73).

Conclusion

These results suggest that quantitative diffusion MRI is sensitive to brain tissue changes in participants with Parkinson disease and provide evidence of substantia nigra lateral asymmetry in this disease.

© RSNA, 2019

Online supplemental material is available for this article.

Summary

By using high-spatial-resolution diffusion MRI that was sensitive to tissue microstructures, significant changes of diffusion parameters were observed only on the left side of substantia nigra of right-handed patients with Parkinson disease.

Key Points

• ■ Diffusion characteristics of the substantia nigra were different in patients with Parkinson disease compared with healthy control participants.

• ■ Differences in all diffusion parameters in Parkinson disease were localized to the left substantia nigra (P < .03) compared with healthy control participants in right-handed participants, suggesting lateral asymmetry of the substantia nigra in Parkinson disease.

Introduction

Parkinson disease (PD) is a neurodegenerative disorder characterized by progressive degeneration of dopaminergic neurons in the substantia nigra (SN), leading to abnormalities of movement and other functions. Motor symptoms in PD onset and progression have exhibited lateral asymmetry (1), which suggests that the neuronal loss in the SN can be asymmetric, as reported in a neuropathologic study (2). In a majority of patients, symptoms begin on the side of the dominant hand (1). SPECT has demonstrated left hemispheric predominance of nigrostriatal dysfunction in patients with PD who are right handed, providing further evidence that dopaminergic denervation in the SN can be asymmetric (3). Furthermore, signal lateral asymmetry in the SN was indicated in an anatomic MRI study at 7.0 T (4) and considered in a susceptibility-weighted imaging study at 3.0 T (5). Neuroimaging evidence of underlying tissue microstructural alterations in the SN, particularly with the presence of asymmetry, ultimately may help detect those at risk for PD or exhibiting prodromal or mild motor symptoms, and identify who would be candidates for potential future disease-modifying therapies.

Diffusion MRI emerged as an important tool to probe tissue microstructures. Apparent diffusion coefficient (ADC) from diffusion-weighted imaging and fractional anisotropy and mean diffusivity from diffusion-tensor imaging have been used extensively to study brain tissue alterations, including those in PD (6–9). ADC, fractional anisotropy, and mean diffusivity are derived from a monoexponential diffusion model, which assumes that diffusion displacement of water molecules follows a Gaussian distribution. However, this assumption breaks down as tissue structural complexity becomes greater (10–14), particularly when measured at high b values (eg, b > 1500 sec/mm² in brain tissues), leading to non-Gaussian diffusion displacement. A number of non-Gaussian diffusion models have been developed to probe various aspects of tissue microstructures (10–16). Among these models, the continuous-time random-walk (CTRW) model (14) is of particular interest because it offers two new parameters related to temporal diffusion heterogeneity (α) and spatial diffusion heterogeneity (β) that can be linked to intravoxel tissue structural heterogeneities, thereby providing an avenue to studying microstructural alterations in the human brain.

Gaussian or non-Gaussian diffusion imaging is typically performed by using a single-shot echo-planar imaging pulse sequence to shorten acquisition times and reduce motion sensitivity. However, conventional single-shot echo-planar imaging offers limited spatial resolution and is subject to image distortion, which causes challenges for studying small structures located in the distortion-prone areas such as the brainstem that contains the SN. The spatial resolution can be substantially improved by restricting the field of view, which was demonstrated by the use of two-dimensional radiofrequency excitation pulses (17–20). Additionally, the reduced field-of-view technique also lessens geometric distortion from magnetic susceptibility variations.

In this study, we hypothesized that if the structural abnormalities in the SN in patients with PD were laterally asymmetric, it would be represented by an asymmetry in diffusion. Therefore, our aim was to investigate the possible lateral asymmetry by using a high-spatial-resolution reduced field-of-view diffusion imaging sequence with a Gaussian (ie, the monoexponential) and a non-Gaussian diffusion model (ie, the CTRW model).

Materials and Methods

Participants

This cross-sectional study was approved by the institutional review board of the University of Illinois at Chicago and Rush University in accordance with Health Insurance Portability and Accountability Act guidelines. All participants provided written informed consent for the study, which was conducted from March 2015 to March 2017.

Thirty-four participants with PD were recruited by convenience from the Rush University Movement Disorder Clinic as part of an ongoing study of clinical and neuroimaging markers of cognitive and behavioral symptoms related to PD (21). All participants with PD were examined by a movement disorders neurologist (J.G.G., with more than 15 years of experience) and met United Kingdom Parkinson’s Disease Society Brain Bank criteria (22). The enrolled participants were aged 60–80 years with more than 4 years of disease duration, on stable medication regimens, and right handed. Exclusionary criteria were as follows: atypical or secondary forms of parkinsonism, severe or unstable depression, anticholinergic medications, other medical or neurologic reasons for cognitive impairment, or contraindications to MRI. Twenty-seven right-handed healthy control (HC) participants were also recruited from the community and age matched to the participants with PD. The HC participants met the following inclusion criteria: no cognitive complaints, healthy cognition with Mini-Mental State Examination scores greater than 28, normal neurologic examinations, absence of known neurologic disorders (eg, brain tumors, epilepsy, and multiple sclerosis), and no contraindications to MRI.

Reduced Field-of-View Pulse Sequence for Diffusion MRI

A custom reduced field-of-view diffusion imaging pulse sequence (19,20) was used in the study to achieve high spatial resolution in the brainstem (Fig 1). Central to the pulse sequence was a two-dimensional radiofrequency excitation pulse that was designed by using a tilted excitation k-space trajectory to enable multisection imaging (17–19). To achieve this, a square k-space trajectory was first designed in the plane defined by the phase-encoding and section-selection directions. The square k-space trajectory was then rotated by 60° in plane, followed by scaling through changing the relative amplitude of the phase-encoding and section-selection gradients to achieve an optimal excitation k-space coverage (20). The two-dimensional radiofrequency pulse, which used a fly-back scheme to avoid issues associated with Nyquist ghosts, was incorporated into a diffusion-weighted single-shot echo-planar imaging pulse sequence with a subsequent radiofrequency refocusing pulse to achieve fat suppression (20).

Figure 1:

Diagram shows reduced field-of-view pulse sequence (top) used in this study and its two-dimensional (2D) radiofrequency (RF) pulse (middle). For the RF pulse, the period of the accompanying gradient was 1.4 msec and 11 subpulses were used. The k-space trajectory was first designed as a square in the plane defined by the phase-encoding (k_y) and section-selection (k_z) directions, and then rotated by 60° in plane, followed by scaling the relative amplitude of the phase-encoding and section-selection gradients to achieve an optimal excitation k-space coverage (bottom). Note that the 2D RF pulse used a fly-back design to avoid the issues associated with Nyquist ghosts. EPI = echo-planar imaging, G_diff = diffusion gradient, G_pe = phase-encoding gradient, G_ss = section-selection gradient.

Image Acquisition

The pulse sequence in Figure 1a was implemented by using a 3.0-T imager (MR750; GE Healthcare, Waukesha, Wis). All participants were imaged with a 32-channel head coil (Nova Medical, Wilmington, Mass). The imaging protocol included noncontrast agent–enhanced sagittal T1-weighted imaging and axial T2-weighted fast spin-echo imaging with full brain coverage, followed by diffusion imaging with seven b values (0, 50, 200, 500, 1000, 2000, and 3000 sec/mm², with four, two, two, four, four, four and four signal averages, respectively) over a small field of view of 10 × 6 cm² in the axial plane covering the brainstem. At each nonzero b value, the diffusion gradient was applied successively along the three orthogonal directions to obtain a trace-weighted image to mitigate the effects of diffusion anisotropy. The other key acquisition parameters for the diffusion examination were as follows: repetition time msec/echo time msec, 3080/86; diffusion gradient separation, 47.0 msec; diffusion gradient pulse width, 32.2 msec; section thickness, 3 mm; intersection spacing, 0 mm; number of sections, 26; 160 × 96 matrix; voxel size, 0.6 × 0.6 × 3 mm³; and total examination time, approximately 3.5 minutes.

Diffusion Image Analysis

The reduced field-of-view diffusion-weighted images were reconstructed off-line by using customized software (Matlab; Mathworks, Natick, Mass). Complex signal averaging was used in the image reconstruction to increase the signal-to-noise ratio (20). After reconstruction of the individual images acquired with different gradient orientations, trace-weighted diffusion images were obtained and analyzed by using a conventional Gaussian monoexponential model and a non-Gaussian CTRW model, respectively.

In the Gaussian diffusion model, ADC was obtained from a monoexponential fitting with the following equation:

by using two b values of 50 and 1000 sec/mm², where S is the signal intensity acquired at b, S₀ is the signal intensity without diffusion weighting, exp is an exponential function, and b is the b value. ADC calculation with higher b values (eg, 3000 sec/mm²) was not attempted because of considerable deviation from Gaussian diffusion.

In the CTRW model, the signal intensity in diffusion-weighted imaging normalized to S₀ is described by a Mittag-Leffler function E_α:

where b is the b value, D_m is an anomalous diffusion coefficient, and α and β are parameters related to temporal and spatial diffusion heterogeneities, respectively, as further explained in reference (14). Equation (2) was used to fit to the multiple-b-value diffusion images, pixel by pixel, by using an iterative Levenberg-Marquardt method in the Matlab Optimization Toolbox (Mathworks).

For each participant, regions of interest were drawn bilaterally on the hypointense region of the SN by using the images with a b value of 0 (Fig 2; regions of interest were drawn by Z.Z. with the supervision of two neuroradiologists who had more than 8 years of experience, on the basis of their mutual agreement). The structure of SN was identified in at least three contiguous sections, and the middle section was chosen to draw the regions of interest to minimize partial-volume effects. A region of interest of a similar size was also drawn on the midbrain tegmentum as an internal reference (bottom row in Fig 2). After obtaining the pixel-by-pixel parameter maps by using Equation (1) or (2), the mean values of D_m, α, β, and ADC within each region of interest were calculated for each participant and used for bilateral and unilateral comparisons.

Figure 2:

Images (b = 0 sec/mm²) with the regions of interest (ROI) used in the image analysis. Bilateral ROIs were placed on the right and left substantia nigra (SN) separately (top), and an additional ROI in the tegmentum (bottom) was used as an internal reference in the comparisons.

Statistical Analysis

Statistical analyses were performed by using software (SPSS version 24.0; IBM, Armonk, NY) for both bilateral and unilateral SN comparisons between the PD and HC groups. In the bilateral comparison, the right and left regions of interest in the SN were combined for each participant, and an average value from the combined region of interest for each diffusion parameter was obtained. The mean value from all participants in each group was computed and used to detect the difference between the PD and HC groups by using a two-tailed Student t test; P values less than .05 indicated statistical significance. A multivariate analysis of variance was also performed to test for statistically significant difference between the PD and HC groups by using the combination of all CTRW parameters. For validation, the same statistical analyses were performed on the internal reference tegmentum to confirm the absence of other confounding factors that might lead to false-positive findings.

To investigate the unilateral abnormalities in the SN, the right and left regions of interest in Figure 2 were treated separately. A preliminary two-way mixed analysis of variance was first performed to examine whether there were significant group interactions between the SN side (ie, right SN and left SN) and the participant type (ie, participant with PD and HC participant). After confirming significant group interactions between the two factors, a post hoc analysis of the simple main effects was performed to investigate the unilateral difference between the PD and HC groups by using a two-tailed Student t test.

On the side of the SN that exhibited a statistically significant difference between the PD and HC groups, a receiver operating characteristic (ROC) analysis was performed by using the individual and the combination of CTRW parameters and ADC to identify unilateral SN abnormality. The combination of the three CTRW parameters (D_m, α, and β) was accomplished by using a multivariable logistic regression analysis according to the following function:

where P₀ is the probability of detecting the abnormality; exp is an exponential function; a₀ is a constant; and a₁, a₂, and a₃ are the regression coefficients of the corresponding CTRW parameters D_m, α, and β, respectively. The probability P₀ was then used as a test variable in the ROC analysis. Additional ROC analyses were performed analogously to investigate other combinations, such as D_m and α, α and β, and D_m and β. Youden index was used to determine the best cut-off sensitivity and specificity of each ROC curve, together with the diagnostic accuracy and the area under the curve to identify unilateral SN abnormality in participants with PD.

Results

Image data from seven participants with PD were excluded from the analysis because of excessive motion during the MRI. Data from the remaining 27 participants with PD were analyzed and compared with those from the 27 HC participants. The characteristics of all participants are summarized in Table 1. The mean age of participants with PD was 71.2 years ± 6.3 (standard deviation); the mean age of HC participants was 69.7 years ± 5.6. The average duration of disease for patients with PD was 11.4 years.

Table 1:

Participant Characteristics

Note.—Mean data are ± standard deviation; data in parentheses are range. HC = healthy control, MDS-UPDRS = Movement Disorder Society Unified Parkinson’s Disease Rating Scale, NA = not applicable, PD = Parkinson disease, PDD = Parkinson disease dementia, PD-MCI = Parkinson disease mild cognitive impairment, PD-NCI = Parkinson disease no cognitive impairment.

*Side of onset was determined by patient report and record review.

Representative Diffusion-weighted Images and Diffusion Parameter Maps

Representative diffusion-weighted images of the brainstem from a participant with PD and an HC participant are shown on the top row of Figure 3, where structures in the brainstem were resolved with high spatial resolution (voxel size, 0.6 × 0.6 × 3 mm³) and in the absence of visible image distortion. The other images on Figure 3 show a set of maps of D_m, α, β, and ADC obtained in the same participants. ADC within the brainstem was higher than D_m, which was likely because of the non-Gaussian behavior that was not accounted for in the ADC calculation.

Figure 3:

A set of representative axial diffusion-weighted (DW) images and different diffusion parameter maps from a patient with Parkinson disease (PD; 71-year-old man with PD duration of 10 years) and a healthy control (HC) participant (71-year-old man). The DW MR images (DW imaging; b = 1000 sec/mm²; top row) illustrate high spatial resolution and absence of image distortion achieved by the reduced field-of-view pulse sequence in Figure 1. α = temporal diffusion heterogeneity index, β = spatial diffusion heterogeneity index, ADC = apparent diffusion coefficient, D_m = anomalous diffusion coefficient.

Bilateral SN Comparison between the PD and HC Groups

In the bilateral comparison between the PD and HC groups, lower ADC (P = .03) and higher β (P = .016) were observed in the PD group compared with the HC group (Table 2). Although D_m (P = .12) and α (P = .13) did not exhibit a difference between the two groups, the combination of the three CTRW parameters showed a difference (P = .02) as revealed by the multivariate analysis of variance analysis. In all comparisons, no difference was observed in the midbrain tegmentum, which was the internal reference (P > .05).

Table 2:

Bilateral Substantia Nigra Comparison

Note.—Unless otherwise indicated, data are mean ± standard deviation. Comparison was performed between the Parkinson disease and healthy control groups by using a two-tailed Student t test for each individual Gaussian and non-Gaussian diffusion parameter. α = temporal diffusion heterogeneity index, β = spatial diffusion heterogeneity index, ADC = apparent diffusion coefficient, CTRW = continuous-time random walk, D_m = anomalous diffusion coefficient, HC = health control, MANOVA = multivariate analysis of variance, PD = Parkinson disease.

*P value indicates statistical significance.

Unilateral SN Comparison between the PD and HC Groups

The preliminary two-way mixed analysis of variance showed significant group interactions between the SN side and the participant type (typical P < .02), which provided the justification to use a post hoc analysis of simple main effects to investigate the differences between the PD and HC groups on each side. To establish a reference in unilateral comparison, the right and left sides of SN were first compared within the HC group, which showed no lateral difference in any of the diffusion parameters (P > .67; Fig 4). Unilateral differences between the PD and HC groups were observed in the left side (P < .03), but not in the right (P > .20) for each of the three CTRW parameters and for the ADC (Fig 5). This observation was further substantiated by significant correlations between two diffusion parameters and the motor score only on the left side of SN (ADC, r = −0.384, P = .047; and β, r = 0.390, P = .04). The results in Figure 5 also indicated that the bilateral differences between the PD and HC groups, summarized in Table 2, arose primarily from the left SN in the PD group. It is worth noting that α and D_m showed differences in the unilateral comparison in the left side, but not in the bilateral comparison (Table 2), which is likely because of averaging with the noncontributing right side. For the comparisons that did not show significance, a power analysis was performed to determine the minimum detectable differences (Appendix E1 [online]).

Figure 4:

Box plots of unilateral comparison between the left and right side of substantia nigra within the healthy control (HC) group for individual Gaussian or non-Gaussian diffusion parameters. No significant difference was observed in any of the four parameters, which is indicated by the P values. The dotted vertical lines indicate the range of the data. The mean value and standard deviation of the diffusion parameters for the right and left substantia nigra in the HC group were as follows: temporal diffusion heterogeneity index (α), 0.87 ± 0.05 versus 0.88 ± 0.05, respectively; spatial diffusion heterogeneity index (β), 0.64 ± 0.06 versus 0.63 ± 0.05, respectively; anomalous diffusion coefficient (D_m), 0.33 μm²/msec ± 0.07 versus 0.32 μm²/msec ± 0.06, respectively; and apparent diffusion coefficient (ADC), 0.52 μm²/msec ± 0.08 versus 0.53 μm²/msec ± 0.07, respectively.

Figure 5:

Box plots of unilateral comparison between the Parkinson disease (PD) and healthy control (HC) groups with different diffusion parameters. The unilateral difference was observed on the left side of the substantia nigra (SN; top), but not on the right (bottom) between the PD and HC groups for each of the four diffusion parameters: temporal diffusion heterogeneity index (α), spatial diffusion heterogeneity index (β), anomalous diffusion coefficient (D_m), and apparent diffusion coefficient (ADC). The P value in each comparison is indicated, and bold denotes that the P value was indicative of statistical significance (P < .05). The dotted vertical lines indicate the range of the data.

ROC Analyses on the Left SN Changes in PD Patients

At the ROC analyses (Table 3), ADC produced the best accuracy (38 of 54 [70.4%]) and the best specificity (44 of 54 [81.5%]) among individual parameters for identifying the changes on the left side of SN. When individual CTRW parameters were used, α provided the best sensitivity (40 of 54 [74.1%]), diagnostic accuracy (38 of 54 [70.4%]), and the area under the curve (area under the curve, 0.68); D_m produced the best specificity (38 of 54 [70.4%]). Compared with ADC, only α had a higher sensitivity and a comparable diagnostic accuracy.

Table 3:

Receiver Operating Characteristic Curve Analysis

Note.—Unless otherwise indicated, data are numerator/denominator and data in parentheses are percentages. Analysis is based on the changes in the left side of the substantia nigra. α = temporal diffusion heterogeneity index, β = spatial diffusion heterogeneity index, ADC = apparent diffusion coefficient, D_m = anomalous diffusion coefficient.

* Data in parentheses are 95% confidence intervals.

^†Best value of sensitivity, specificity, or accuracy among all receiver operating characteristic curves.

^‡Largest area under the receiver operating characteristic curve in the column.

When the CTRW parameters were combined, the performance was notably improved. Among the four combinations, the combination of D_m, α, and β produced the best sensitivity (42 of 54 [77.8%]), the combination of D_m and β yielded the best specificity (46 of 54 [85.2%]), and the combination of α and β yielded the largest area under the curve (area under the curve, 0.73). Overall, ADC produced one of the best diagnostic accuracies, whereas the combinations of the CTRW parameters provided more balanced sensitivity and specificity. The ROC curves of two representative combinations of CTRW parameters, the combination of D_m, α, and β and the combination of α and β, together with ADC, are shown in Figure 6.

Figure 6:

Selected receiver operating characteristic curves by using the left substantia nigra diffusion parameters for detection of Parkinson disease. Two combinations of continuous-time random-walk parameters, anomalous diffusion coefficient (D_m), temporal diffusion heterogeneity index (α), and spatial diffusion heterogeneity index (β), and α and β, as well as the apparent diffusion coefficient (ADC) were selected to identify the abnormality on the left side of substantia nigra in the Parkinson disease group compared with the healthy control group. The sensitivity, specificity, diagnostic accuracy, and area under the curve are summarized in Table 3.

Discussion

By using high-spatial-resolution reduced field-of-view diffusion MRI with Gaussian and non-Gaussian diffusion models, we observed significant differences only on the left SN between the PD and HC groups. This suggests that the unilateral difference is the major contributor to the bilateral difference observed in our study and reported by others (7,8,23).

Handedness has been proposed as a possible cause to explain this asymmetry (24). In our study, all participants were right handed, which can explain the observed unilateral difference only on the left side of the SN. This explanation is consistent with previous studies that reported a significant association between the handedness and the side of the initial motor symptom (1), and the dominant side of symptom in patients with PD (25). Another study that used SPECT reported left hemispheric predominance of nigrostriatal dysfunction among right-handed patients with PD, further suggesting that hemispheric asymmetry plays an important role in the development of PD (3). Even though handedness may not be the only reason, the results of our study indicate that it can be an important factor for understanding the unilateral structural alteration in the SN in patients with PD.

Dopaminergic neuronal degeneration in the SN can lead to the underlying tissue microstructural alterations. These tissue structural alterations can be effectively probed with diffusion MRI. For example, fractional anisotropy changes revealed by diffusion-tensor imaging have been reported in the SN (6–8); however, the reported trends for fractional anisotropy change are inconsistent (6–8,26,27). Water diffusivity, measured by ADC or mean diffusivity, has also been used to investigate possible changes in the SN of patients with PD. Most of these studies did not find a significant difference between participants with PD and HC participants (27–29), whereas other studies reported higher (6,30) or lower (7) diffusivity in patients with PD. The inadequate spatial resolution (eg, ∼2 × 2 × 2 mm³) that prevents a more sensitive unilateral analysis in these studies can be a possible explanation of the inconsistent findings.

Unlike previously published studies, we used a high-spatial-resolution diffusion MRI technique with reduced field of view that allowed us to investigate the unilateral as well as bilateral differences in the SN in patients with PD. The spatial resolution used in this study (voxel size, 0.6 × 0.6 × 3 mm³) is, to our knowledge, among the highest ever applied to PD by using diffusion MRI and is particularly important for imaging small structures such as the SN by substantially reducing the partial-volume effects. By using high-spatial-resolution diffusion imaging, a reduction in ADC was observed only on the left side of the SN in the PD group compared with the HC group. This is an indication of a more restricted diffusion environment possibly because of increased cellularity or cell volume fraction related to the formation of Lewy bodies, the glial response to the neuronal loss in SN (mainly microglial cells and to a lesser extent reactive astrocytes), and neuroinflammation when neuromelanin-containing neurons die and neuromelanin granules are released into the extracellular space (31).

The non-Gaussian CTRW model and its variations were successfully applied to tissue characterization in cancer imaging studies (11,14,32–34). This model offers three parameters: D_m, α, and β. Unlike ADC, D_m accounts for non-Gaussian diffusion behavior in biologic tissues. The smaller D_m and ADC observed in our study are in agreement with a report on SN in PD (7). α and β have been associated with the temporal and spatial diffusion heterogeneity, respectively (14), both of which may reflect different aspects of intravoxel tissue structural heterogeneity. The smaller α on the left side of SN in the PD group suggests that water molecules are diffusing through a more heterogeneous environment temporally (ie, water molecules take a variable amount of time to make a move), whereas the increased β indicates a spatially more homogeneous environment (ie, the water molecules diffuse with a more uniform step length in each move). The different trend of the diffusion parameters can be a reflection of the compounded microstructural changes related to PD such as neuronal loss and reactive gliosis. An additional value of non-Gaussian CTRW diffusion imaging is that it combines multiple diffusion parameters because of their complementary roles in describing complex diffusion processes. The combinations of two or three CTRW parameters provide multiple possibilities to complement ADC in improving sensitivity and specificity.

Our study had limitations. One limitation is the lack of comparison with a group of left-handed participants because of the small occurrence in the general population (<10%) (35), which makes it difficult to recruit a sufficient number of participants with PD. The limited number of participants also prevented us from performing an independent validation in the ROC analysis. Another limitation was that the reported unilateral alterations in diffusion parameters were not validated by or correlated with another independent method such as histopathologic analysis. Thus, the exact tissue structural underpinning of the observed changes, to our knowledge, remains unknown. Finally, the accuracy of our quantitative analysis was negatively affected by the relatively low signal-to-noise ratio in the diffusion-weighted images because of simultaneous increases in spatial resolution and b value. The signal-to-noise ratio limitation forced us to use a relatively thick section compared with the high in-plane resolution. However, the use of a 32-channel coil together with signal averaging helped to alleviate this problem.

In conclusion, our results provided in vivo neuroimaging evidence of substantia nigra lateral asymmetry in patients with Parkinson disease and suggested a set of diffusion imaging markers for characterizing such asymmetry.