A Deep Learning Algorithm for Automated Cardiac Murmur Detection Via a Digital Stethoscope Platform

John S Chorba; Avi M Shapiro; Le Le; John Maidens; John Prince; Steve Pham; Mia M Kanzawa; Daniel N Barbosa; Brent E White; Jason Paek; Sophie G Fuller; Grant W Stalker; Sara A Bravo; Dina Jean; Subramaniam Venkatraman; Patrick M McCarthy; James D Thomas

Our results also show that the cardiologists and the algorithm perform better in detecting aortic stenosis than detecting mitral regurgitation. This may be attributed to aortic stenosis having a more discernible auscultatory signature. Moreover, its characteristic murmur can be heard in both the aortic and pulmonic positions on auscultation, offering additional data points for both the human and computer decision-makers. Physiologically, mitral regurgitation is a load-dependent murmur, such that severity of regurgitation can vary markedly depending on the minute-to-minute hemodynamics of the patient. Moreover, MR can be directional, and therefore certain types of MR may simply not have murmurs at a single auscultation location, like the cardiac apex. The higher rates of poor signal from the mitral position in our study only exacerbate these challenges. Further study of additional auscultation locations, or alternative, non-murmur auscultatory characteristics, would be useful to aid in improving the performance of VHD detection algorithms.