Non-Invasive Cardiovascular Health Monitoring for Patients with Heart Failure using Seismocardiography
Loading...
Author(s)
Shandhi, Md. Mobashir Hasan Hasan
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Heart failure (HF) is the leading cause of hospitalization and hospital readmission for patients aged over 65 and older in the United States, with roughly one in five individuals hospitalized with heart failure being readmitted within 30 days of discharge. HF affects 6.2 million Americans with health care costs of almost $31 billion per year. Management of HF is a complicated process that requires frequent clinic visits and outpatient management systems for hemodynamic monitoring and patient-reported symptoms. Hemodynamically guided HF management via tracking pulmonary congestion and taking proactive care have shown efficacy in reducing HF-related readmission significantly. However, the cost-prohibitive nature of such pulmonary congestion monitoring systems precludes their usage in the large patient population affected by HF. For that reason, an inexpensive alternative is necessary to bring hemodynamic monitoring systems to the large patient population affected by HF, not only in the United States but also around the world.
Advancement of novel biomedical sensor technologies and advanced signal processing and machine learning algorithms have merit in tracking health parameters unobtrusively. A promising sensing modality is seismocardiography (SCG), defined as the measurement of local chest wall vibrations associated with the cardiac cycle. SCG has shown efficacy in tracking changes in cardiac contractility via the cardiac timing intervals it yields, such as the pre-ejection period (PEP). However, different sensing modalities of SCG acquisition exist using accelerometer and gyroscope based sensors, and inter-subject variability of these acquired signals has made it challenging to develop a robust hemodynamic monitoring system using SCG. Accordingly, most researches in the field of SCG focus on advancing the understanding and processing of the signal in healthy individuals. The translation of the SCG-based hemodynamic monitoring approaches into the actual patient population, for example, in patients with HF, is necessary to validate such a system for both inpatient and outpatient HF management.
This work addresses some of these key aspects. First, the two sensing techniques for acquiring SCG, accelerometer and gyroscope sensors, are compared in their ability to track cardiac contractility changes via PEP estimation. Second, general time, frequency, and amplitude features are extracted from the SCG signals and used in a population level machine learning regression algorithm to estimate key cardiovascular features for healthy subjects and patients with HF, by overcoming the inter-subject variability of the signals. Third, the SCG sensing system, along with the signal processing and machine learning algorithm, is verified and validated with two gold-standard clinical procedures: cardiopulmonary exercise test (CPX) and right heart catheterization (RHC). Gas exchange variables from the CPX and changes in pulmonary congestion from the RHC procedures were estimated using features from simultaneously recorded SCG signals to demonstrate the efficacy of such a sensing system and algorithm to track relevant hemodynamic parameters in patients with HF.
The algorithms and methods presented in this work can enable remote cardiovascular health monitoring for patients with HF to enable personalized titration of care, and improving medication adherence in a hemodynamically-guided HF management system. The inexpensive wearable sensing technology has the potential to be a viable and ubiquitous alternative to the already-proven hemodynamic congestion monitoring systems, which can improve the quality of life and outcome in patients with HF by reducing hospitalization and reducing the overall health care costs.
Sponsor
Date
2020-12-03
Extent
Resource Type
Text
Resource Subtype
Dissertation