Ultra-strong diffusion-weighted MRI reveals cerebellar grey matter abnormalities in movement disorders

Abstract

Structural brain MRI has proven invaluable in understanding movement disorder pathophysiology. However, most work has focused on grey/white matter volumetric (macrostructural) and white matter microstructural effects, limiting understanding of frequently implicated grey matter microstructural differences. Using ultra-strong spherical tensor encoding diffusion-weighted MRI, a persistent MRI signal was seen in healthy cerebellar grey matter even at high diffusion-weightings (b ​≥ 10,000 s/mm²). Quantifying the proportion of this signal (denoted f_s), previously ascertained to originate from inside small spherical spaces, provides a potential proxy for cell body density. In this work, this approach was applied for the first time to a clinical cohort, including patients with diagnosed movement disorders in which the cerebellum has been implicated in symptom pathophysiology. Five control participants (control group 1, median age 24.5 years (20-39 years), imaged at two timepoints, demonstrated consistency in measurement of all three measures - MD (Mean Diffusivity) f_s, and D_s (dot diffusivity)- with intraclass correlation coefficients (ICC) of 0.98, 0.86 and 0.76, respectively. Comparison with an older control group (control group 2 (n = 5), median age 51 years (43-58 years)) found no significant differences, neither with morphometric nor microstructural (MD (p = 0.36), f_s (p = 0.17) and D_s (p = 0.22)) measures. The movement disorder cohort (Parkinson's Disease, n = 5, dystonia, n = 5. Spinocerebellar Ataxia 6, n = 5) when compared to the age-matched control cohort (Control Group 2) identified significantly lower MD (p < 0.0001 and p < 0.0001) and higher f_s values (p < 0.0001 and p < 0.0001) in SCA6 and dystonia cohorts respectively. Lobar division of the cerebellum found these same differences in the superior and inferior posterior lobes, while no differences were seen in either the anterior lobes or with D_s measurements. In contrast to more conventional measures from diffusion tensor imaging, this framework provides enhanced specificity to differences in restricted spherical spaces in grey matter (including small cells) by eliminating signals from cerebrospinal fluid and axons. In the context of human and animal histopathology studies, these findings potentially implicate the cerebellar Purkinje and granule cells as contributors to the observed signal differences, with both cell types having been implicated in several neurological disorders through both postmortem and animal model studies. This novel microstructural imaging approach shows promise for improving movement disorder diagnosis, prognosis, and treatment.

Keywords: Cerebellum; Diffusion; MRI; Microstructure; Movement Disorders.

Fig. 1

A. Linear b-tensor diffusion encoding (LTE, top) with diffusion sensitisation along a single axis, and spherical b-tensor diffusion encoding (STE, bottom) with diffusion sensitisation in all directions. Timings for the first waveform, temporal gap (180° pulse), and second waveform were [28.6, 6.9, 28.6] ms for the linear encoding and [35.5, 6.9, 25.6] ms for STE, respectively. The figure show averages across DWIs per b-value, with the intensity min–max normalised per b-value according to the LTE image. B. Graphical overview of the estimation of dMRI features from the STE signal as a function of b-value (black line), here simulated as a tri-exponential decay with $f=\left[0.2,0.72,0.08\right]$ and $D=\left[3,1,0.1\right]\mu m^2/ms$ where the first and last compartment mimic free water and a spherical restricted compartment, respectively. The blue line is estimated by fitting a mono-exponential function to b ​$\ge$ 10,000 s/mm² and has slope $D_s$ and y-intercept $S_{0,s}$. The red line is estimated by fitting a bi-exponential function $S=S_0\left(f_t\exp \left(-b·MD\right)+\left(1-f_t\right)\exp \left(-b·3\right)\right)$ to b ​$\le$ 1500 s/mm² and has slope $\mathit{MD}$ and y-intercept $f_t{S}_0$. At low b-values, the deviation from mono-exponential behaviour is assumed to be arising from the free water compartment with a diffusivity of $3\mu m^2/ms$. C. Cerebellar segmentation in lobules (top) and lobes (bottom). D. Schematic cross-sectional representation of cerebellar grey matter microstructure including the granule cell layer (yellow), Purkinje cell layer (grey) and the molecular layer (blue). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

A. Intraclass correlation coefficient (ICC) for different dMRI measures (rows) in Control Group 1. B. Comparison of estimated dMRI measures between Group 1 and 2 across lobes. C. Comparison of estimated dMRI measures between Group 1 and 2 for different lobes.

Fig. 3

A. Boxplots of the median values of mean diffusivity ($\mathit{MD})$, signal fraction ($f_s)$ and diffusivity ($D_s$) from each cerebellar region (I-X) (circles) across the whole cerebellum, excluding the flocculonodular lobe. B. Boxplots of the median values for each voxel within each region (I-X) (circles) analysed by cerebellar lobe; anterior, superior posterior and inferior posterior. *** denotes statistically significant comparisons, beyond that of the Bonferroni correction for multiple metric comparisons (p < 0.003).