|
Abstract:
|
Parkinson's disease is heterogeneous in symptom presentation and progression. Increased understanding of both aspects can enable better patient management and improve clinical trial design. In past work, we hypothesized that it is critical to model the longitudinal patterns of PD progression given its observed heterogeneity as opposed to characterizing subtypes using cross-sectional data. To that end, we developed a personalized input-output hidden Markov model, trained using up to seven years of data from the Parkinson’s Progression Markers Initiative, which enabled the discovery of PD disease states, each of which has associated clinical presentation and progression patterns. We validated the clinical implications of that model by assessing the differential prevalence of key clinical outcomes not used to train the model in two patient cohorts. In this work, we demonstrate how the disease progression model can be used to discover candidate biomarkers for improved quantitative assessment of patient disease state.
|
