Title:  Processing and Learning of Cardiomechanical Signals for Heart Failure

Committee: 

Dr. Rehg, Advisor

Dr. Inan, Co-Advisor     

Dr. Davenport, Chair

Dr. Anderson

Abstract: The objective of the proposed research is to use cardio-mechanical signals such as ballistocardiography (BCG) and seismocardiography (SCG) in heart failure (HF) care. HF is a progressive condition in which the heart is not able to pump enough blood to meet the body’s demands.  Between 2011 and 2014, 6.5 million American adults were diagnosed with HF, with 670,000 patients diagnosed annually. Estimates for 2030 suggest that the number of adults with HF could exceed 8 million. One of the driving factors of the cost and mortality of HF is the high rate of rehospitalization of the patients following initial hospitalization. Rehospitalizations can be reduced by remote monitoring of HF patients. To enable remote monitoring, non-invasive devices that can measure cardio-mechanical signals are required.  Two such signals are called BCG and SCG.  The goal of this proposal is to develop effective processing and learning methods to utilize SCG and BCG signals in HF care. This proposal describes two processing and modeling efforts to utilize BCG and SCG signals in classification of clinical status of the HF patients.  In the first effort, processing pipeline for BCG signals is developed to classify clinical decompensation. AUC of 0.78 is achieved in BCG data collected HF patients from home and hospital settings. In the second effort, similarly, processing pipeline is developed for SCG signals to classify hemodynamic decompensation. AUC of 0.84 is achieved from SCG data collected in the hospital. The proposed work on cardio-mechanical (BCG and SCG) signals aims to extract better features in classifying clinical status of HF patients using deep learning.  Within the domain of cardio-mechanical signals, dataset size is usually not large enough compared to that of successful applications of deep learning. To overcome this problem, quasi-periodic structure in the cardio-mechanical signals can be exploited to have specialized deep learning modules and auxiliary tasks. This exploitation can learn a feature set that outperforms hand-designed features.