The goal of the proposed research is to identify the best predictive biomarkers of dementia in Parkinson?s disease (PDD) through a multimodal and multivariate statistical model utilizing both neuroimaging derived measures (diffusion-weighted MRI (dMRI), resting-state functional MRI (rsfMRI), and T1-weighted MRI measures) and non- imaging measures such as demographics (age, sex, years of education), clinical (disease duration and severity), genetics (LRRK2), and CSF-measures (Total Tau, ?-Amyloid, ?-synuclein). It is critical to identify biomarkers that can predict dementia in Parkinson?s disease (PD) as approximately 50-80% of PD patients develop PDD within twelve years of diagnosis. Identifying pathophysiology-based biomarkers that could identify PD patients at high risk for PDD reliably is critical for better prognostication, correct identification of PDD in its prodromal stage to recruit in new disease-modifying clinical trials, and better understanding the pathophysiological processes underlining PDD. The proposed project has two important components. The first component of the project is to understand the pathophysiological mechanism underlying PDD through sophisticated voxelwise dMRI-derived measures estimated using a multi-shell high angular and spatial resolution dMRI data acquisition, and understanding network-level white matter (WM)-derived structural connectivity and rsfMRI-derived functional connectivity in PDD. The second component of the project is to identify the biomarkers that predict PDD through multivariate statistical modelling by combining these sophisticated pathologically relevant neuroimaging measures with non-imaging measures (such as clinical, demographics, genetics, and CSF-measures). We will recruit demographically matched healthy controls (HC) along with demographically, disease duration, and disease severity matched PD patients with mild cognitive impairment (PD-MCI), PD-non-MCI (PD-nMCI), and PDD for this project. We will acquire multi-shell dMRI data at three b-values, namely 500s/mm2, 1000s/mm2, and 2500s/mm2 with a high angular and spatial resolution and estimate various unbiased free-water (fiso) corrected Gaussian dMRI-derived measures along with non-Gaussian dMRI-derived measures such as diffusion kurtosis measures, and neurite orientation dispersion and density imaging measures. We will further compare these measures between the groups to identify significant dMRI-derived measures separating the groups, and understanding the neuroanatomical correlates of these measures with various neuropsychological scores. Furthermore, we will estimate dMRI-derived structural connectivity and rsfMRI-derived functional connectivity to understand network-level discrepancies predicting PDD. These pathologically relevant neuroimaging measures will be further combined with various non-imaging measures through a novel machine learning algorithm to identify the comprehensive and best predictors of PDD. The tools developed in our proposal also has great potential for significantly advancing the understanding of other neurodegenerative disorders such as Alzheimer?s disease (AD) thereby helping to understand AD- and PD-specific neuroanatomical changes predicting dementia.
The proposed research is highly relevant to public health because identification of the Parkinson?s disease (PD) patients at a high-risk for developing dementia (PDD) is critical for better prognostication, correct identification of PDD in its prodromal stage to recruit in new disease-modifying clinical trials, and to better understand the pathophysiological processes underlining PDD. The proposed project identifies the best predictive biomarkers for PDD through examining the comprehensive Gaussian and non-Gaussian diffusion-MRI (dMRI)-derived white matter (WM) changes predicting PDD, examining the network-level discrepancies predicting PDD derived from dMRI-derived whole-brain structural connectivity and resting-state functional MRI (rsfMRI)-derived whole-brain functional connectivity, and developing a multivariate and multimodal statistical model by combining these sophisticated pathologically relevant neuroimaging measures with non-imaging measures such as clinical (disease duration and severity), demographics (age, sex, years of education), genetics (LRRK2), and CSF- measures (?-Amyloid, ?-Synuclein, tau) to identify the best predictive biomarkers of PDD through machine learning approach. The machine learning algorithm developed in our proposal also has great potential for significantly advancing the understanding of other neurodegenerative disorders such as Alzheimer?s disease (AD) thereby helping to understand AD- and PD-specific neuroanatomical changes predicting dementia.
