Can MRI-based multivariate gray matter volumetric distance predict motor progression and classify slow versus fast progressors in Parkinson’s disease?

Abstract

AbstractIntroductionWhile Parkinson’s disease (PD) related neurodegeneration is associated with structural changes in the brain, magnetic resonance imaging (MRI) has not been helpful in diagnosing PD or predicting the progression of motor symptoms. In this study, we aimed to develop a structural MRI-based biomarker to predict the rate of progression of motor symptoms and to classify patients based on the symptom severity (i.e. slow vs. fast progressors) in the early stages of PD.MethodsThe study included 59 patients with PD (n=40 for the primary analysis, 19 for the validation analysis), and 55 healthy controls with structural MRI from the Parkinson’s Progression Markers Initiative (PPMI) database. We developed a patient-specific multivariate gray matter volumetric distance using Mahalanobis distance (MGMV) to investigate the changes in MGMVover time using longitudinal linear mixed-effect model, its potential as a biomarker to predict the rate of progression of motor function (MDS-UPDRS-part III) using multiple linear regression model, and classification of patients based on symptom severity using machine learning (ML).ResultsMGMVat BL significantly predicted changes in motor severity (p<0.05) and a trend level increase in MGMVover time (p = 0.09) were noted. We obtained 85% accuracy in discriminating patients according to their symptom severity, and on an independent test cohort, an accuracy of 90% was achieved.ConclusionsWe identified a promising structural MRI-based biomarker for predicting the rate of progression of motor symptoms and classification of patients based on motor symptom severity.