Robin Chen’s research while affiliated with University of Florida and other places

Importance Magnetic resonance imaging (MRI) paired with appropriate disease-specific machine learning holds promise for the clinical differentiation of Parkinson disease (PD), multiple system atrophy (MSA) parkinsonian variant, and progressive supranuclear palsy (PSP). A prospective study is needed to test whether the approach meets primary end points to be considered in a diagnostic workup. Objective To assess the discriminative performance of Automated Imaging Differentiation for Parkinsonism (AIDP) using 3-T diffusion MRI and support vector machine (SVM) learning. Design, Setting, and Participants This was a prospective, multicenter cohort study conducted from July 2021 to January 2024 across 21 Parkinson Study Group sites (US/Canada). Included were patients with PD, MSA, and PSP with established criteria and unanimous agreement in the clinical diagnosis among 3 independent, blinded neurologists who specialize in movement disorders. Patients were assigned to a training set or an independent testing set. Exposure MRI. Main Outcomes and Measures Area under the receiver operating characteristic curve (AUROC) in the testing set for primary model end points of PD vs atypical parkinsonism, MSA vs PSP, PD vs MSA, and PD vs PSP. AIDP was also paired with antemortem MRI to test against postmortem neuropathology in a subset of autopsy cases. Results A total of 316 patients were screened and 249 patients (mean [SD] age, 67.8 [7.7] years; 155 male [62.2%]) met inclusion criteria. Of these patients, 99 had PD, 53 had MSA, and 97 had PSP. A retrospective cohort of 396 patients (mean [SD] age, 65.8 [8.9] years; 234 male [59.1%]) was also included. Of these patients, 211 had PD, 98 had MSA, and 87 had PSP. Patients were assigned to the training set (78%; 104 prospective, 396 retrospective) or independent testing set, which included 145 (22%; 60 PD, 27 MSA, 58 PSP) prospective patients (mean age, 67.4 [SD 7.7] years; 95 male [65.5%]). The model was robust in differentiating PD vs atypical parkinsonism (AUROC, 0.96; 95% CI, 0.93-0.99; positive predictive value [PPV], 0.91; negative predictive value [NPV], 0.83), MSA vs PSP (AUROC, 0.98; 95% CI, 0.96-1.00; PPV, 0.98; NPV, 0.81), PD vs MSA (AUROC, 0.98; 95% CI, 0.96-1.00; PPV, 0.97; NPV, 0.97), and PD vs PSP (AUROC, 0.98; 95% CI, 0.96-1.00; PPV, 0.92; NPV, 0.98). AIDP predictions were confirmed neuropathologically in 46 of 49 brains (93.9%). Conclusions and Relevance This prospective multicenter cohort study of AIDP met its primary end points. Results suggest using AIDP in the diagnostic workup for common parkinsonian syndromes.

It is established that one of the best predictors of a future diagnosis of Parkinson’s disease is a current diagnosis of rapid eye movement behaviour disorder (RBD). In such patients, this provides a unique opportunity to study brain physiology and behavioural motor features of RBD that may precede early-stage Parkinson’s disease. Based on prior work in early-stage Parkinson’s disease, we aim to determine if the function of corticostriatal and cerebellar regions are impaired in RBD using task-based functional MRI and if structural changes can be detected within the caudate, putamen and substantia nigra in RBD using free-water imaging. To assess motor function, we measured performance on the Purdue Pegboard Test, which is affected in patients with RBD and Parkinson’s disease. A cohort of 24 RBD, 39 early-stage Parkinson’s disease and 25 controls were investigated. All participants were imaged at 3 Telsa. Individuals performed a unimanual grip force task during functional imaging. Participants also completed scales to assess cognition, sleep and motor symptoms. We found decreased functional activity in both RBD and Parkinson’s disease within the motor cortex, caudate, putamen and thalamus compared with controls. There was elevated free-water-corrected fractional anisotropy in the putamen in RBD and Parkinson’s disease and elevated free-water in the putamen and posterior substantia nigra in Parkinson’s disease compared with controls. Participants with RBD and Parkinson’s disease performed significantly worse on all tasks of the Purdue Pegboard Test compared with controls. The both hands task of the Purdue Pegboard Test was most sensitive in distinguishing between groups. A subgroup analysis of early-stage RBD (<2 years diagnosis) confirmed similar findings as those in the larger RBD group. These findings provide new evidence that the putamen is affected in early-stage RBD using both functional and free-water imaging. We also found evidence that the striatum, thalamus and motor cortex have reduced functional activity in early-stage RBD and Parkinson’s disease. While the substantia nigra shows elevated free-water in Parkinson’s disease, we did not observe this effect in early-stage RBD. These findings point to the corticostriatal and thalamocortical circuits being impaired in RBD patients.

Background Diffusion‐weighted magnetic resonance imaging (dMRI) examines tissue microstructure integrity in vivo. Prior dementia with Lewy bodies (DLB) diffusion tensor imaging studies yielded mixed results. Objective We employed free‐water (FW) imaging to assess DLB progression and correlate with clinical decline in DLB. Methods Baseline and follow‐up MRIs were obtained at 12 and/or 24 months for 27 individuals with DLB or mild cognitive impairment with Lewy bodies (MCI‐LB). FW was analyzed using the Mayo Clinic Adult Lifespan Template. Primary outcomes were FW differences between baseline and 12 or 24 months. To compare FW change longitudinally, we included 20 cognitively unimpaired individuals from the Alzheimer's Disease Neuroimaging Initiative. Results We followed 23 participants to 12 months and 16 participants to 24 months. Both groups had worsening in Montreal Cognitive Assessment (MoCA) and Movement Disorder Society‐Unified Parkinson's Disease Rating Scale (MDS‐UPDRS) scores. We found significant FW increases at both time points compared to baseline in the insula, amygdala, posterior cingulum, parahippocampal, entorhinal, supramarginal, fusiform, retrosplenial, and Rolandic operculum regions. At 24 months, we found more widespread microstructural changes in regions implicated in visuospatial processing, motor, and cholinergic functions. Between‐group analyses (DLB vs. controls) confirmed significant FW changes over 24 months in most of these regions. FW changes were associated with longitudinal worsening of MDS‐UPDRS and MoCA scores. Conclusions FW increased in gray and white matter regions in DLB, likely due to neurodegenerative pathology associated with disease progression. FW change was associated with clinical decline. The findings support dMRI as a promising tool to track disease progression in DLB. © 2024 International Parkinson and Movement Disorder Society.

Background Diffusion‐weighted MRI (dMRI) examines tissue microstructure integrity and white matter pathways in vivo. Prior dementia with Lewy bodies (DLB) studies focused on white matter or voxel‐based techniques. Less is known about longitudinal white and grey matter microstructural changes in DLB. We aimed to characterize longitudinal diffusion microstructural changes in individuals with DLB. Method dMRI scans were collected on individuals with DLB from the Mayo Clinic Longitudinal Imaging Biomarkers of DLB Program. Demographics, clinical evaluations and MRI were collected at baseline, and 12‐ or 24‐months. Free water (FW) and FW‐corrected fractional anisotropy (FA T ) were analyzed in grey and white matter using 122 bilateral template regions and tracts of interest from the Mayo Clinic Adult Lifespan Template. Primary outcomes were differences in FW and FA T between baseline and 12‐ or 24‐months. Result We identified 23 individuals with DLB or mild cognitive impairment with Lewy bodies (MCI‐LB) from baseline to 12‐months [mean age 69.3 years (SD 9.5); 95.8% male], and 16 individuals from baseline to 24‐months [mean age 67.5 years (SD 9.3); 100% male]. Participants at both time‐points showed worsening in Montreal Cognitive Assessment and Movement Disorders Society Unified Parkinson’s Disease Rating Scale motor scores (p<0.05). We found significant differences in FW, but not FA T , from baseline to 12‐months. Significant increases in FW at both time‐points compared to baseline were observed in the insula, putamen, parahippocampal, and rolandic operculum regions (p fdr <0.05). We found additional more widespread microstructural changes from baseline to 24‐months, with increased FW in the amygdala, entorhinal, mid and posterior cingulum, inferior frontal, hippocampal, pallidum, precuneus and retrosplenial regions (p fdr <0.05). We found decreased FA T from baseline to 24‐months in the cerebellar and superior occipital regions (p fdr <0.01). Conclusion Longer follow‐up at 24‐months identified broader networks of free water changes in individuals with DLB. Results support dMRI as a promising, non‐invasive tool to track disease progression in DLB. More investigation is needed to evaluate the relationship of microstructural changes with clinical progression in DLB.

With the advent of group equivariant convolutions in deep networks literature, spherical CNNs with $\mathsf{SO}(3)$-equivariant layers have been developed to cope with data that are samples of signals on the sphere $S^2$. One can implicitly obtain $\mathsf{SO}(3)$-equivariant convolutions on $S^2$ with significant efficiency gains by explicitly requiring gauge equivariance w.r.t. $\mathsf{SO}(2)$. In this paper, we build on this fact by introducing a higher order generalization of the gauge equivariant convolution, whose implementation is dubbed a gauge equivariant Volterra network (GEVNet). This allows us to model spatially extended nonlinear interactions within a given receptive field while still maintaining equivariance to global isometries. We prove theoretical results regarding the equivariance and construction of higher order gauge equivariant convolutions. Then, we empirically demonstrate the parameter efficiency of our model, first on computer vision benchmark data (e.g. spherical MNIST), and then in combination with a convolutional kernel network (CKN) on neuroimaging data. In the neuroimaging data experiments, the resulting two-part architecture (CKN + GEVNet) is used to automatically discriminate between patients with Lewy Body Disease (DLB), Alzheimer's Disease (AD) and Parkinson's Disease (PD) from diffusion magnetic resonance images (dMRI). The GEVNet extracts micro-architectural features within each voxel, while the CKN extracts macro-architectural features across voxels. This compound architecture is uniquely poised to exploit the intra- and inter-voxel information contained in the dMRI data, leading to improved performance over the classification results obtained from either of the individual components.