Personalized Input-Output Hidden Markov Models for Disease Progression Modeling

Abstract

Disease progression models are important computational tools in healthcare and are used for tasks such as improving disease understanding, informing drug discovery, and aiding in patient management. Although many algorithms for time series modeling exist, healthcare applications face particular challenges such as small datasets, medication effects, disease heterogeneity, and a desire for personalized predictions. In this work, we present a disease progression model that addresses these needs by proposing a probabilistic time-series model that captures individualized disease states, personalized medication effects, disease-state medication effects, or any combination thereof. The model builds on the framework of an input-output hidden Markov model where the parameters are learned using a structured variational approximation. To demonstrate the utility of the algorithm, we apply it to both synthetic and real-world datasets. In the synthetic case, we demonstrate the benefits afforded by the proposed model as compared to standard techniques. In the real-world cases, we use two Parkinson's disease datasets to show improved predictive performance when ground truth is available and clinically relevant insights that are not revealed via classic Markov models when ground truth is not available.

Human-centered explainability for life sciences, healthcare, and medical informatics

“…To provide a real-world example, let us take disease progression modeling as a use case to describe what the potential roles each healthcare persona can play (see Figure 2 ). The goal of disease progression modeling (DPM) 63 is to model the natural continuous progression of chronic diseases identifying multiple irreversible stages, each having diverse disease symptoms. Clinicians are responsible for interacting with the end users, the patients, and as such they typically use their medical expertise to define the overall goal of the DPM, mimicking the clinical progression of the disease through the generation of clinical hypothesis.…”

Discovery of Parkinson's disease states and disease progression modelling: a longitudinal data study using machine learning

“…Progression modelling was done using a machine learning approach: a personalised inputoutput hidden Markov model (figure 1). 20 The primary goal of the analysis was to learn a small number of clinically useful states while accounting for medication effects. A state is a discrete label of a patient and has two primary characteristics: a transition model, which describes the probability of changing states, and an observation model, which describes the distribution of clinical measures associated with a state.…”

Fully Personalised Degenerative Disease Modelling - A Duchenne Muscular Dystrophy Case Study

“…In this case, time is again explicitly modelled, with separate models again for each patient, but certain parameters in the models (the trend between each time step, and the variances) are made the same. Severson et al 10 used hidden Markov models for disease progression, with the disease able to progress between different latent states over time. In this model, these latent states are different for each patient, but the parameters controlling the random processes are the same between patients, with the exception of an additive inter-patient random-effect term and a personalised parameter for the effect of treatments.…”