In this pilot study we sought to evaluate the feasibility and effectiveness of using the Mean Apparent Propagator (MAP) to characterize differences in diffusion MRI (dMRI) that reflect changes in tissue microstructure in hemodialysis (HD) patients compared to healthy controls. Prior literature has shown that theoretically and empirically MAP parameters can be sensitive to brain microstructural changes such as inflammation and degeneration, in silico and in a transgenic Alzheimer's mouse model [[1], [2], [3], [4], [5], [6], [7]]. HD patients experience >2-fold higher rates of cognitive impairment compared to age matched peers [8]. Structural deficits underlying cognitive impairment include decreased cortical thickness and volume, loss of white matter integrity indicated by diffusion tensor imaging (DTI) and increased white matter hyperintensity (WMH) volume as measured by T2 FLAIR [[9], [10], [11], [12], [13]]. Cross-sectional studies have unambiguously been successful in detecting diffuse dMRI deficits across the WM in HD patients [9,11,12]. However, studies in rats and humans have shown that HD patients have increased cerebral water content [14,15]. Further, Schaier et al. 2019 found that a single dialysis session can cause reductions in fractional anisotropy (FA) and increases in mean, radial and axial diffusivity (MD, RD and AD) [16]. As a result, studies using DTI may be complicated by these cerebral water content changes in HD patients, which would depend on individual solute transport, solute reduction during HD and time since last HD session. These diffusion dependencies could obscure detection of the tissue microstructural changes. Alternatively, this physiological effect could be another element of neurodegeneration in HD patients which deserves further study.

MAP might be useful in overcoming limitations of DTI in HD patients. DTI metrics such as FA, MD, AD, and RD are primarily used to determine tissue integrity in white matter and have been related to axonal packing, diameter, and demyelination [[17], [18], [19]]. However, interpretation of DTI parameters makes several assumptions about the organization of tissue and is not specifically sensitive to the complex interplay of osmotic shifts, inflammation, and neurodegeneration present in HD patients [20]. To adequately model multiple fiber populations and non-Gaussian diffusion, q-space needs to be sampled densely with a large number of directions over multiple shells [21]. Unlike the diffusion tensor model that represents diffusion in a voxel as a simple ellipsoid, the mean apparent propagator (MAP) model estimates the actual diffusion propagator in each voxel. In this technique, the 3D MR signal attenuation in q-space (E(q¯)) is expressed as the sum of a set of basis functions, which is then used to calculate the full diffusion propagator [2]. This representation using basis functions makes it easier to calculate a set of scalar metrics that are related to the extracellular volume fraction, apparent intracellular volume, apparent axonal cross-sectional area, and apparent axonal length [1,22,23]. These metrics allow for analysis on intra- and extra-cellular water contributions to diffusion that could allow for detection of progression of disease such as inflammation and necrosis [6]. In recent studies, MAP metrics have been found to be more sensitive than DTI metrics in Parkinson's disease to microstructural changes in subcortical gray matter, better at predicting motor clinical outcomes post-stroke and microstructural changes in development [[24], [25], [26]].

In order to assess the utility of MAP in detecting diffusion changes in HD patients, we collected diffusion data in a small cohort of HD patients and age and sex matched controls. We hypothesized that MAP would be more sensitive to differences between controls and HD patients and provide indicators of microstructural deficits in HD patients. Secondarily, we postulated that MAP metrics would be less sensitive to head motion during the scan than DTI metrics as was found by Pines et al. 2020 [26]. Additionally, we hypothesized that the MAP propagator anisotropy would be a better measure of anisotropy than DTI fractional anisotropy in regions of complex fiber architecture.