Mixture of Input-Output Hidden Markov Models for Heterogeneous Disease Progression Modeling

A particular challenge for disease progression modeling is the heterogeneity of a disease and its manifestations in the patients. Existing approaches often assume the presence of a single disease progression characteristics which is unlikely for neurodegenerative disorders such as Alzheimer's disease and Parkinson's disease. In this paper, we develop a hierarchical approach based on mixtures of hidden Markov models that can identify similar groups of patients through time-series clustering and separately represent the progression of each group, unlike hidden Markov models which assume that a single dynamics is shared among all patients. The proposed model is an extension of an input-output hidden Markov model that takes into account the clinical assessments of patients' health status and the prescribed medications. We illustrate the benefits of our approach using a synthetically generated dataset and a real-world longitudinal dataset for Parkinson's disease, obtained from the Parkinson's Progression Markers Initiative observational study. While the synthetic data experiments demonstrate the ability of mixture of personalized hidden Markov models to simultaneously learn personalized state effects and multiple disease progression dynamics when the true disease progression dynamics is known, real-data experiments show that a mixture of input-output hidden Markov models is favoured over an input-output hidden Markov model for disease progression modeling.