Title:  Continuous Cardiorespiratory Monitoring Using Ballistocardiography From Load Cells Embedded in a Hospital Bed

Committee:

Dr. Omer Inan, ECE, Chair, Advisor

Dr. Ying Zhang, ECE

Dr. Woon-Hong Yeo, ME

Dr. Nima Ghalichechian, ECE

Dr. Rishikesan Kamaleswaran, Emory

Abstract: The objective of this research is to explore signal processing and machine learning techniques to allow continuous monitoring of cardiorespiratory parameters using the ballistocardiogram (BCG) signals recorded with sensors embedded in a hospital bed. First, the heart rate (HR) estimation algorithms were presented. The first is signal processing-based HR estimation with array processing for multi-channel combination. The second uses a deep learning (DL) model that tranforms BCG signals into an interpretable triangular waveform, from which hearbeat locations can be estimated. Second part of the work focuses on the estimation of respiratory rate (RR) and respiratory volume (RV) using the respiration waveforms derived from the low-frequency components of the load cell signals. Lastly this work presents two models for blood pressure (BP) estimation -- 1) Conventional pulse transit time (PTT)-based model, and 2) DL-based model, both using multi-channel BCG and the photoplethysmogram (PPG) signals to extract features. Overall, this work established methods that would enable non-invasive and continuous monitoring of standard vital signs utilizing the sensors already embedded in commonly-deployed commercially available hospital beds. Such technologies could potentially improve the continuous assessment of the patients' physiologic state without adding an extra burden on the caregivers.