A0702
Title: Advanced HMMs for physiological time series
Authors: Sida Chen - MRC BSU University of Cambridge (United Kingdom)
Barbel Finkenstadt - University of Warwick (United Kingdom)
Francis Levi - Warwick Medical School - University of Warwick (United Kingdom)
Kristian Romano - University of Warwick (United Kingdom) [presenting]
Abstract: Wearable devices allow for non-invasive telemonitoring of patients. Physiological data, such as physical activity and body temperature, can be collected in the daily living environment of the patient without the need for hospitalization. The devices can be used to monitor the Circadian Timing System (CTS), which regulates many critical cellular processes, such as the cell cycle and metabolism, and to detect adverse events which pose a risk to health. The appraisal of the CTS in real time could lead to chronopharmacological strategies and personalized medicine. To model the physiological data arising from wearable sensors we developed non-homogeneous hidden Markov and semi-Markov models with nonparametric emission distributions. To accommodate for the oscillating nature of the CTS, the transition probabilities are driven by a circadian oscillator. We will illustrate the novel models, estimation algorithms and some (preliminary) results for simulations.
