A Bayesian Circadian Hidden Markov Model to Infer Rest-Activity Rhythms Using 24-hour Actigraphy Data

by   Jiachen Lu, et al.

24-hour actigraphy data collected by wearable devices offer valuable insights into physical activity types, intensity levels, and rest-activity rhythms (RAR). RARs, or patterns of rest and activity exhibited over a 24-hour period, are regulated by the body's circadian system, synchronizing physiological processes with external cues like the light-dark cycle. Disruptions to these rhythms, such as irregular sleep patterns, daytime drowsiness or shift work, have been linked to adverse health outcomes including metabolic disorders, cardiovascular disease, depression, and even cancer, making RARs a critical area of health research. In this study, we propose a Bayesian Circadian Hidden Markov Model (BCHMM) that explicitly incorporates 24-hour circadian oscillators mirroring human biological rhythms. The model assumes that observed activity counts are conditional on hidden activity states through Gaussian emission densities, with transition probabilities modeled by state-specific sinusoidal functions. Our comprehensive simulation study reveals that BCHMM outperforms frequentist approaches in identifying the underlying hidden states, particularly when the activity states are difficult to separate. BCHMM also excels with smaller Kullback-Leibler divergence on estimated densities. With the Bayesian framework, we address the label-switching problem inherent to hidden Markov models via a positive constraint on mean parameters. From the proposed BCHMM, we can infer the 24-hour rest-activity profile via time-varying state probabilities, to characterize the person-level RAR. We demonstrate the utility of the proposed BCHMM using 2011-2014 National Health and Nutrition Examination Survey (NHANES) data, where worsened RAR, indicated by lower probabilities in low-activity state during the day and higher probabilities in high-activity state at night, is associated with an increased risk of diabetes.


Bayesian Approximations to Hidden Semi-Markov Models

We propose a Bayesian hidden Markov model for analyzing time series and ...

Hierarchical Continuous Time Hidden Markov Model, with Application in Zero-Inflated Accelerometer Data

Wearable devices including accelerometers are increasingly being used to...

A Hidden Markov Model Based Unsupervised Algorithm for Sleep/Wake Identification Using Actigraphy

Study Objective: Actigraphy is widely used in sleep studies but lacks a ...

Mixture of hidden Markov models for accelerometer data

This work is motivated by the analysis of accelerometer data. The analys...

Activity Detection from Wearable Electromyogram Sensors using Hidden Markov Model

Surface electromyography (sEMG) has gained significant importance during...

A three-state coupled Markov switching model for COVID-19 outbreaks across Quebec based on hospital admissions

Recurrent COVID-19 outbreaks have placed immense strain on the hospital ...

Please sign up or login with your details

Forgot password? Click here to reset