Parkinson disease (PD) and essential tremor (ET) are some of the most common movement disorders. In individuals older than 65, the estimated prevalence of ET and PD is 5% and 1.8%, respectively [1]. ET is predominantly characterized by kinetic or postural tremors involving bilateral upper limbs [2], while PD is characterized by asymmetric resting tremors, rigidity, and bradykinesia [3]. Additionally, PD is often associated with significant non-motor symptoms such as depression, anxiety, cognitive impairment, and sleep dysregulation. Although non-motor symptoms are increasingly reported as concomitant clinical features of ET [4, 5, 6], some studies have shown no such association [7], suggesting that the pathophysiological relevance of non-motor symptoms in ET is still unclear. While ET and PD are considered distinct diagnostic entities, the co-existence of these pathologies within individuals has been reported. Some studies have further suggested that ET may be a risk factor for PD, conferring up to a 5-fold increase in risk [8, 9], although this has not been substantiated in other studies [10, 11, 12]. Regardless, overlapping motor and non-motor symptoms can lead to ambiguity in diagnosis, further hampered by the possibility of concurrent disease.

The pathophysiology of PD is well studied and linked to neurodegeneration, but the pathophysiology of ET remains controversial. Magnetic resonance imaging (MRI) biomarkers may lend insight into the underlying pathophysiology of ET and PD. To that end, a better understanding of the pathological changes between ET and PD may aid in the differential diagnosis as well as in understanding the unique features of these movement disorders. MRI biomarkers based on diffusion properties have previously been employed to assess white matter changes associated with ET [13, 14, 15] and PD [16, 17].

Diffusion MRI (dMRI) measures the diffusion of water molecules in brain tissues, which can occur either without restriction (isotropic diffusion) or limited by obstacles, such as cell membranes (anisotropic diffusion). Diffusion tensor imaging (DTI) allows for microstructure-sensitive metrics, such as fractional anisotropy (FA) and axial and radial diffusivities (AxD and RD, respectively), to be derived. FA represents the directionality of the water diffusion [18] and provides the degree of anisotropy of the diffusion tensor. Biologically, FA is often considered a direct marker of white matter integrity and is related to several factors, including axonal count and density, degree of myelination, and fiber architecture. The diffusivity parameters AxD and RD represent the water diffusivity along and perpendicular to the principal axis, respectively [19]. AxD is generally associated with axonal damage, while RD has been associated with myelin integrity, axonal diameter and density, and fiber coherence [20, 21]. However, it is important to note that the interpretation of these DTI-related metrics may be altered in neuropathology, reflecting the differential evolution of pathophysiological changes (e.g., axonal damage and loss, demyelination) across disease types and disease progression [22, 23, 24].

Previous studies using DTI in ET have generally shown higher mean diffusivity (MD) and lower FA values compared to controls, with changes occurring predominantly in the cerebellum and brainstem, as well as in cerebral white matter tracts [25, 26]. While DTI metrics in PD have shown heterogeneous patterns in early PD [27], clearer trends of increasing MD and decreasing FA have been observed longitudinally and in the later stages of the disease [28]. Direct comparisons between ET and PD have shown white matter changes in the cerebello-thalamic cortical pathways involving the visual network [29].

Standard single-tensor DTI is a common model for studying white matter microstructural integrity; however, this method has several limitations. Notably, it cannot resolve different single-fiber orientations within each voxel. The inability of DTI to resolve intravoxel fiber orientations can impact both DTI metrics (FA, AxD, and RD) and tractography [30]. As a result, it is increasingly recognized that the direct biological interpretation of DTI-based metrics is hampered by the complex white matter fiber geometry in the brain [31, 32]. To overcome this limitation, more advanced dMRI models have been proposed to resolve multiple fiber populations inside a single voxel [33]. Fixel-based analysis (FBA) is an advanced diffusion model that estimates microstructural changes of individual populations of specific fiber bundles on a microscopic level [34], where ‘fixel’ describes a fiber population within a voxel. Physiologically, FBA-derived metrics are thought to represent intra-axonal volume changes associated with specific fiber populations (fiber density, FD) and macroscopic cross-sectional sizes of individual fiber bundles (fiber-bundle cross-section, FC) [35]. These metrics reflect on pathophysiological processes such as fiber loss or atrophy and thus allow for direct assessment of microstructural changes on a sub-voxel level [36].

In this retrospective study, we compared white matter microstructural differences between ET and PD cohorts with moderate to severe disease. Standard DTI-based FA was assessed for comparison with previous studies, while advanced FBA metrics were assessed to probe white matter microstructural differences between these groups. Our goal was to investigate regional differences in white matter microstructure between these two movement disorders.

Participant characteristics are summarized in Table 1. No significant differences in head motion were detected between PD and ET (t = 0.502; p = 0.617) patients; however, 1 ET and 6 PD patients were excluded from the analysis due to substantial motion during the dMRI scan (motion $>$5 mm); therefore, a total of 78 participants (16 ET and 62 PD) were included in the final analysis. From the t-test, PD and ET groups (from the participants included in the final analysis) did not differ significantly in age (t = 0.956; p = 0.342), but they did differ in disease duration (t = 2.249; p = 0.027).

Differences between the 2 groups detected by DTI through the FA index are shown in Fig. 2. Through FWE and effect-size analysis, several significant clusters in white matter were found where the ET group had lower FA values than the PD group. Table 2 reports the complete list of clusters detected by FA analysis, with their volume percentage, t, and g values. FA analysis showed large clusters of difference in the corpus callosum (CC) (volumes: $\approx$29% and $\approx$16% for FWE and effect size, respectively) and in the left tapetum (volumes: $\approx$51% and $\approx$35% for FWE and effect size, respectively).

Fig. 2.

For FA, significant clusters at FWE 0.05 and g $\mathrm{>}$0.80 were found in several white matter areas. Compared with PD, the ET patients showed lower FA values. In the figure, the top sub-panel image represents the FWE analysis, and the bottom sub-panel image represents the effect size. PD, Parkinson disease; ET, essential tremor; FA, fractional anisotropy.

Fig. 3 shows the differences in FBA (FWE and effect size) between the 2 groups. For all metrics, significant clusters of differences between the 2 groups (top sub-panels of (a)-(c)) were found mainly in the CC ($\approx$9%, $\approx$8%, and $\approx$18% of the volume of the CC for FD, log FC, and FDC, respectively) and left tapetum (volumes: $\approx$16.5%, $\approx$4.5%, and $\approx$56% for FD, log FC, and FDC, respectively). These results were also confirmed by the effect size (bottom sub-panels). Additionally, for log FC, the effect-size analysis showed a large difference (with g $\mathrm{>}$0.80) between the 2 groups in the cerebral peduncle (see Table 3 for complete regional results). Compared with the PD patients, we found lower values of all FBA-related metrics in the ET patients.

Fig. 3.

For FBA, significant clusters at FWE 0.05 and g $\mathrm{>}$0.80 were found for FD, log FC, and FDC metrics. (a) FD, (b) log FC, (c) FDC. Compared with PD, the ET patients showed lower values in different parts of the white matter for all metrics. For each metric, the top sub-panel images represent the FWE analysis, and the bottom sub-panel images represent the effect size. FBA, fixel-based analysis; FD, fiber density; FC, fiber-bundle cross-section; FDC, fiber density bundel cross-section.

As a secondary analysis to confirm that the scanner covariate did not drive the group-wise results, Supplementary Fig. 1 and Supplementary Table 1 show the results from the same statistical analysis for FBA for GE and Philips scanners separately. Similar regional trends are observed for Philips (top sub-panels) and GE (bottom sub-panels) for each FBA metric ((a)-(c)), and in general, these trends match those of Fig. 2 and Table 2. Notably, differences between groups in the CC and left tapetum were confirmed for both combined and separate analyses.

In this study, we evaluated white matter tract differences between a well-characterized clinical cohort of advanced ET and moderate-to-severe PD using a novel fixel-based method. FBA overcomes several limitations associated with standard DTI; most notably, DTI cannot differentiate multiple sub-voxel fiber orientations often present in crossing fibers, which may lead to inaccurate DTI metrics. Additionally, FBA allows for the estimation of different orientations of fibers within a voxel, yielding metrics reflecting FD, FC, and a combination of these measures. For comparison, results from standard DTI are also presented. With standard DTI-based FA, widespread non-specific differences were observed between ET and PD groups, with many changes occurring more superior and anterior. On the other hand, results from FBA demonstrated significant differences were primarily concentrated in posterior regions of the corpus callosum, as well as the cerebral peduncle and left tapetum.

Understanding the advanced stages of the disease, particularly in patients seeking neurosurgical interventions, may reveal unique regional vulnerabilities in white matter pathways that may be a factor underlying treatment outcomes with DBS in tremors. A previous study by Juttukonda et al. [29] considered DTI-based changes in a similar cohort of ET and PD patients presenting for DBS. In that study, the authors found lower FA in ET patients compared to PD patients in white matter tracts posteriorly along the cerebello-thalamo-cortical network, as well as in the splenium of the corpus callosum and tapetum. These results are consistent with our results showing lower FA in ET than PD group, including in the tapetum; however, our standard DTI-based FA showed greater involvement of anterior cortical projections, including the genu/body of corpus callosum, anterior corona radiata, and cingulum (cingulate gyrus). Interestingly, the FBA results showed posterior-dominant involvement, most notably in the splenium of the corpus callosum and tapetum, which anatomically aligns with the prior study. Taken together, the previous findings of lower FA and higher RD and the current findings of reduced FD and FDC may indicate demyelination processes in the splenium and tapetum of ET patients. However, caution is warranted in directly interpreting dMRI metrics, as these metrics may reflect complex biological processes that cannot be fully modeled [36]. Nonetheless, the findings of white matter microstructural differences between ET and PD groups are strongly bolstered by consistent findings replicated across studies and populations. Furthermore, these findings support the hypothesis that a vulnerable white matter pathway may exist in ET.

The involvement of the corpus callosum in PD has been previously noted in several studies. Compared to healthy controls, PD patients have shown decreased FD in the corpus callosum with increased FD in the corticospinal tract, possibly reflecting a compensatory process [51]. Additionally, decreased FDC has been observed longitudinally in the splenium in PD patients, as well as in the tapetum [52]. To our knowledge, no studies have previously investigated FBA-based changes in ET; however, standard DTI has shown variable results in comparisons between ET and healthy controls. For example, 2 studies in ET revealed significant differences in diffusivity measures in the cerebellum [53] and red nuclei [15] but no corresponding FA differences. On the other hand, decreased FA has been observed using a white matter skeletonized approach in the corpus callosum and cortical spinal tract, as well as in the frontal, occipital, and temporal white matter [54]. Although the present study did not compare to healthy controls, our results suggested differences in white matter integrity in the corpus callosum.

Interestingly, few similarities were observed between regions identified by FA and FBA-based metrics. Although the same underlying data with inherent sensitivity to white matter microstructure are used for both analyses, the model used for each is fundamentally different. As noted above, DTI models a single component within each voxel, while FBA can resolve multiple fiber populations within each voxel. Recent studies have suggested that the majority of cerebral white matter (up to 90% [39]) contains multiple fiber bundles, which may lead to erroneous inferences using DTI-based FA alone [32]. For example, we recently demonstrated that an observed paradoxical increase in FA in Alzheimer disease (conventionally interpreted as higher white matter integrity) could be rather attributed to sub-voxel neurodegeneration; by correcting for this factor, the resulting FA changes were more consistent with known Alzheimer disease pathology [55]. Interestingly, increased FA remained after correction in several regions, including the cortical spinal tract; using an advanced anisotropy metric (the mode of anisotropy), we showed that these changes were indicative of degeneration in crossing fiber populations that are not accounted for by free-water correction. On the other hand, FBA estimates tissue microstructure for multiple compartments using constrained spherical deconvolution, which has previously been shown to provide reliable and stable measurements over time [56]. As a result, FBA is likely more reliable in assessing white matter changes in voxels with multiple fiber populations. Other studies comparing DTI and FBA metrics in various pathologies have similarly shown disparate results, suggesting that these models may detect different underlying microstructural changes [57, 58].

The FBA framework provides an advanced analysis pipeline for assessing fiber-specific features associated with white matter in the brain. These features are extracted from dMRI data, allowing the FBA framework to overcome many known challenges associated with more conventional DTI analysis. The currently recommended dMRI protocol for FBA analysis is acquisitions with greater than 45 directions and b-values greater than 3000 s/mm${}^2$ [36]. This protocol increases specificity to intra-axonal signal and the ability to robustly resolve individual white matter FODs. Despite these recommendations, it is technically feasible to perform FBA on dMRI with fewer directions and/or lower b-values; based on a recent review, 8% of studies performed FBA with less than 30 directions, and nearly 60% performed FBA with b-values less than 2500 s/mm${}^2$ [36]. The predominant effect of suboptimal acquisitions is the bias in the measured FD values due to signal contributions from the extra-cellular space outside the axons [36], which may limit the biological interpretation of FBA-based metrics. However, previously published FBA-based studies with a b-value of 1000 s/mm${}^2$ have shown conclusive results [59, 60]. In the current study, we performed a secondary analysis for each scanner independently, allowing comparisons between FBA from b = 1000 (low b-value, GE) and b = 2000 (high b-value, Philips) acquisitions (Supplementary Fig. 1). These results were similar and thus corroborated the combined results. Nonetheless, our findings regarding FBA in the PD and ET groups should be interpreted cautiously.

There are limited studies assessing unique pathological processes in ET compared to other movement disorders such as PD. Interestingly, the ET group in our analysis showed lower FA compared to the PD group, suggesting disruption of white matter pathways. Pathological understanding of ET is limited, with changes in cerebellar Purkinje neurons primarily reported by some groups [40, 61, 62]. Other groups have refuted these findings and have demonstrated a lack of cerebellar involvement [63, 64, 65]. Although our findings did not demonstrate cerebellar group differences, this may be attributed to sample size or methodological differences. It should be noted that the disease stage of both groups is more advanced, reflecting the typical population referred for DBS. Although the exact mechanism is unknown, differential white matter changes between ET and PD have been observed previously [29] and are confirmed herein. Future work should investigate the association of these white matter changes with neurofunctional scores in PD and ET, thus allowing a better understanding of the role of white matter abnormalities on both motor and non-motor symptoms.

In addition to white matter, microstructural changes observed using dMRI and white matter hyperintensities (WMH) are commonly observed in aging populations [66]. They may be associated with a range of neurodegenerative diseases, including both PD [67] and ET [68]. dMRI studies have shown that WMH are associated with reduced FA, reflecting decreased white matter integrity [68]. Additionally, WMHs may be associated with cortical microstructural changes [69], suggesting an interplay between white matter (WM) and gray matter neurodegenerative processes. Differences in WMH burden could contribute to the group-wise differences observed herein, and future studies should investigate the impact of WMH on dMRI biomarkers across groups and whether these changes extend beyond WM tract integrity.

There are some limitations in the present study. The sample sizes were not matched between the two cohorts, and as noted above, the ET cohort was relatively small (n = 17). For this reason, we also analyzed the effect size. Additionally, ET diagnosis was performed based on the 1998 criteria [37], and we did not stratify the ET cohort into different subtypes [70, 71], as this information was not available as part of this retrospective study. Future prospective studies should investigate unique white matter changes across these subtypes. Although our main goal was to analyze white matter differences between PD and ET, the present study lacked a healthy control cohort for comparison. However, several previous studies have analyzed white matter differences between healthy control and PD cohorts [72] and between healthy control and ET cohorts [73], while the focus herein was to investigate regional differences between these 2 movement disorders. Another limitation of our retrospective analysis is the unmatched protocols between the 2 clinical scanners used; to overcome this limitation, we performed a secondary analysis for each scanner independently that corroborated the combined results. Additionally, the dMRI acquisition differed from current recommendations, which may limit the biological interpretation of our findings. Another minor limitation is that full coverage of the cerebellar grey matter was not obtained in the clinical protocol; as all white matter tracts were included in the group template, this did not impact our analysis. Future prospective studies with larger cohorts and at earlier stages are needed to validate and expand the current findings.

In conclusion, the present study showed significant differences in white matter integrity between ET and PD cohorts in the later stages of the disease. Differences in posterior white matter projections were observed using advanced FBA-based metrics, which permits the modeling of multiple intra-voxel fiber orientations. Standard DTI-based FA was also included for comparison and showed group differences in several brain regions. In both cases, white matter integrity was more impacted in ET than PD in regions previously implicated in these diseases. Advanced diffusion-based metrics may provide a better understanding of the similarities and differences in white matter microstructure between these disparate motor-associated diseases with different underlying phenotypes.

The datasets generated and/or analyzed during the current study are not publicly available due to clinical purpose but are available from the corresponding author on reasonable request.

Conceptualization: MB, SA and AMS; Methodology: SA and AMS; Software: MB; Validation: MB, SA and AMS; Formal Analysis: MB; Data Curation: MB and JJK; Visualization: SA and AMS, LA; Writing, Original Draft Preparation: MB, SA and AMS; Writing, Review & Editing: MB, SA, LA and AMS; Supervision: SA and AMS. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

A waiver of informed consent was obtained for this retrospective study. This study was conducted according to the guidelines of the Declaration of Helsinki and was approved by the Institutional Review Board of St. Joseph’s Hospital and Medical Center (protocol PHX-22-500-031-73-04, approved 08/13/2021).

The authors would like to thank Dr. Holly Shill for helpful discussions.

This work was supported by the Barrow Neurological Foundation, Sam & Peggy Grossman Family Foundation, and Samuel P. Mandell Foundation.

The authors declare no conflict of interest.