WEARABLE SENSORS AND HEALTH MONITORING SYSTEMS Received 16 June 2020; revised 14 September 2020; accepted 5 October 2020. Date of publication 8 October 2020; date of current version 20 October 2020. Digital Object Identifier 10.1109/JTEHM.2020.3029690 Quantification of Resting-State Ballistocardiogram Difference Between Clinical and Non-Clinical Populations for Ambient Monitoring of Heart Failure ISAAC SUNGJAE CHANG 1 , (Member, IEEE), SUSANNA MAK 2 , NARGES ARMANFARD 3 , JENNIFER BOGER 4,5 , (Member, IEEE), SHERRY L. GRACE 6,7 , AMAYA ARCELUS 7 , CAROLINE CHESSEX 7 , AND ALEX MIHAILIDIS 7 1 Institute of Biomaterials and Biomedical Engineering, University of Toronto, ON M5S 3G9, Canada 2 Division of Cardiology, Department of Medicine, Mount Sinai Hospital, Toronto, ON M5G 1X5, Canada 3 Department of Electrical and Computer Engineering, McGill University, Montreal, QC H3A 0G4, Canada 4 Department of Systems Design Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada 5 Research Institute for Aging, Waterloo, ON N2J 0E2, Canada 6 Faculty of Health, York University, Toronto, ON M3J IP3, Canada 7 Toronto Rehabilitation Institute, University Health Network, Toronto, ON M5T 2S8, Canada CORRESPONDING AUTHOR: I. S. CHANG (isaac.chang@mail.utoronto.ca) This work was supported in part by Canadian Institutes of Health Research (CIHR), AGE-WELLL NCE Inc., and Natural Sciences and Engineering Research Council (NSERC). This article has supplementary downloadable material available at http://ieeexplore.ieee.org, provided by the authors. ABSTRACT A ballistocardiogram (BCG) is a versatile bio-signal that enables ambient remote monitoring of heart failure (HF) patients in a home setting, achieved through embedded sensors in the surrounding environ- ment. Numerous methods of analysis are available for extracting physiological information using the BCG; however, most have been developed based on non-clinical subjects. While the difference between clinical and non-clinical populations are expected, quantification of the difference may serve as a useful tool. In this work, the differences in resting-state BCGs of the two cohorts in a sitting posture were quantified. An instrumented chair was used to collect the BCG from 29 healthy adults and 26 NYHA HF class I and II patients while seated without any stress test for five minutes. Five 20-second epochs per subject were used to calculate the waveform fluctuation metric at rest (WFMR). The WFMR was obtained in two steps. The ensemble average of the segmented BCG heartbeats within an epoch were calculated first. Mean square errors (MSE) between different ensemble average pairs were then retrieved. The MSEs were averaged to produce the WFMR. The comparison showed that the clinical cohort had higher fluctuation than the non-clinical population and had at least 82.2% separation, suggesting that greater errors may result when existing algorithms were used. The WFMR acts as a bridge that may enable important features, including the addition of error margins in parameter estimation and ways to devise a calibration strategy when resting-state BCG is unstable. INDEX TERMS Ballistocardiogram, resting-state, heart failure, ambient monitoring. I. INTRODUCTION In Canada, more than 660,000 people aged over 40 years had heart failure (HF) in 2013 [1], and 50,000 new cases of HF emerged yearly according to a 2016 report, costing more than $2.8 billion per year [2]. In the United States, HF cost over $11 billion and more than 1 million people were hospitalized due to HF in 2014 [3]. The high cost of HF is primarily attributable to the elevated rate of hospital VOLUME 8, 2020 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ 2700811