Objective: To compare longitudinal structural connectomes derived from whole brain dMRI and PD clinical rating scales as a disease progression biomarker, on three cohorts: de-novo PD, RBD and matched controls

Background: REM behavior disorder (RBD) patients have an estimated lifetime rate of conversion to an alpha-synuclein neurodegenerative disorder of 75-90% and by the time Parkinson’s disease (PD), the most common alpha-synucleinopathy, is diagnosed more than 60% of nigrostriatal neurons have already been lost. There are no known biomarkers for disease diagnosis or progression. Diffusion MRI (dMRI) , combined with machine learning (ML) holds potential as a structural connectivity biomarker to track longitudinal changes in white matter connectivity more objectively.

Methods: The dataset was created by selecting all RBD subjects with at least two dMRI scans as of May 2017 from the publicly available Parkinson’s Progression Markers Initiative (PPMI) dataset. Images were acquired using 3T TIM Trio Siemens scanners with 64 diffusion-weighting gradient directions. The dataset included 16 RBD, 21 PD and 30 controls matched for age, sex, and time between scans. Selected clinical rating scales include the most common scales for motor and cognitive progression (UPDRS-III, SDM and MoCA). Anatomical segmentation was done with a multi-contrast PD atlas, which includes 16 areas of the nigrostriatal pathway. A machine learning (ML) model was trained to identify differences in progression of PD and control group, and used the RBD cohort as a fully external validation. Statistical significance was computed with logistic regression corrected for confounding effects of age and sex.

Results: The best performing ML model was able to distinguish progression between the PD and Control group (AUC=0.89, p<0.001). Most notable was that it found the same PD-specific progression pattern in the RBD group (AUC=0.76, p<0.01). This model was compared with cross-sectional and longitudinal data from clinical rating scales for both motor (UPDRS-III) and cognitive assessment (SDM and MoCA) which showed that the longitudinal brain connectomes had the best performance in identifying differences in progression.

Conclusions: These results demonstrate feasibility of a computational dMRI-based structural connectome biomarker to quantify the progression of neurodegenerative patterns occurring in PD and found in the prodromal RBD group. This approach is a first step towards developing an objective clinical tool to track progression and predict with high specificity and sensitivity the individual RBD patients who will convert to PD.

« Back to 2018 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/tracking-rbd-and-pd-progression-with-longitudinal-structural-brain-connectomes/