UCLA

Individuals with undiagnosed heart disease and those unable to seek advanced care for existing heart-disease related disabilities face a risk of premature death. Guidelines released by the American Heart Association aim to increase healthy life expectancy in such individuals over the next decade. Timely diagnosis and evaluation of heart disease in these individuals with existing tools is a challenge, especially in low-resource primary care settings. Presented here is a computer-aided heart-sound-based system that carries the potential to overcome this challenge and provide these individuals with the opportunity to pursue an accelerated path of care to recovery. This system analyzes phonocardiogram signals collected using acoustic sensors to identify variations in cardiac physiology that are otherwise measured by resource-intensive imaging tools. Signal acquisition is noninvasive, and the entire system operates in a fully automated manner without requirements of training or expert supervision. The acquired signal waveforms are processed using filtering, noise subtraction, heartbeat segmentation, and heartbeat quality assurance algorithms to extract physiologically motivated features for heart disease diagnosis and evaluation. The system and its algorithms were developed and validated using real-world heart sound data from hospital inpatients. When tested on 96 inpatients at the Ronald Reagan University of California Los Angeles Medical Center, the system was able to identify aortic stenosis with a sensitivity and specificity of 92% and 95%. When tested on 34 inpatients being evaluated for heart failure at the Oregon Health & Science University Hospital, the system was able to generate echocardiography-like parameters for left ventricular diastolic function and left atrial pressure evaluation with accuracies of 87.5% and 75%. These results demonstrate the potential of this phonocardiogram-based system in providing clinically relevant heart disease diagnosis and evaluation at the point of primary care and in fulfilling the immediate critical need for improving outcomes in at-risk individuals.