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

Abstract

Background Clinicians vary markedly in their ability to detect murmurs during cardiac auscultation and identify the underlying pathological features. Deep learning approaches have shown promise in medicine by transforming collected data into clinically significant information. The objective of this research is to assess the performance of a deep learning algorithm to detect murmurs and clinically significant valvular heart disease using recordings from a commercial digital stethoscope platform. Methods and Results Using >34 hours of previously acquired and annotated heart sound recordings, we trained a deep neural network to detect murmurs. To test the algorithm, we enrolled 962 patients in a clinical study and collected recordings at the 4 primary auscultation locations. Ground truth was established using patient echocardiograms and annotations by 3 expert cardiologists. Algorithm performance for detecting murmurs has sensitivity and specificity of 76.3% and 91.4%, respectively. By omitting softer murmurs, those with grade 1 intensity, sensitivity increased to 90.0%. Application of the algorithm at the appropriate anatomic auscultation location detected moderate‐to‐severe or greater aortic stenosis, with sensitivity of 93.2% and specificity of 86.0%, and moderate‐to‐severe or greater mitral regurgitation, with sensitivity of 66.2% and specificity of 94.6%. Conclusions The deep learning algorithm’s ability to detect murmurs and clinically significant aortic stenosis and mitral regurgitation is comparable to expert cardiologists based on the annotated subset of our database. The findings suggest that such algorithms would have utility as front‐line clinical support tools to aid clinicians in screening for cardiac murmurs caused by valvular heart disease. Registration URL: https://clinicaltrials.gov ; Unique Identifier: NCT03458806.

Machine Learning and IoT Applied to Cardiovascular Diseases Identification through Heart Sounds: A Literature Review

“…Chorba et al [81] (49) affirmed that due to the heterogeneity in the interpretative capacity of medical professionals to detect pathological characteristics during cardiac auscultation, they presented a computational approach as a promising alternative to assist in the diagnosis of pathologies in the medical area through auscultation. The authors proposed the use of DL algorithms with CNN architecture, using the last layer to normalize probability distribution through a softmax function as a goal of detecting murmurss and valvulopathy.…”

“…Title Countries Year Maritsch et al [39] Improving heart rate variability measurements from consumer smartwatches with machine learning USA 2019 Humayun et al [25] Towards domain invariant heart sound abnormality detection using learnable filterbanks Bangladesh/USA 2020 Shuvo et al [68] Cardioxnet: A novel lightweight deep learning framework for cardiovascular disease classification using heart sound recordings Bangladesh/Saudi Arabia/Yemen 2021 Tiwari et al [27] Phonocardiogram signal based multi-class cardiac diagnostic decision support system India/Saudi Arabia 2021 Du et al [69] Accurate prediction of coronary heart disease for patients with hypertension from electronic health records with big data and machine-learning methods:model development and performance evaluation Catar/Malásia 2019 Chowdhury et al [26] Real-time smart-digital stethoscope system for heart diseases monitoring USA 2019 Swarup and Makaryus [71] Digital stethoscope: technology update USA 2018 Amiri et al [73] Mobile phonocardiogram diagnosis in newborns using support vector machine USA 2017 Ukil et al [32] With robust edge analytics and de-risking India/Switzerland/Spain 2019 Gómez-Quintana et al [80] A framework for ai-assisted detection of patent ductus arteriosus from neonatal phonocardiogram Ireland/Ukraine 2021 Chorba et al [81] Deep learning algorithm for automated cardiac murmur detection via a digital stethoscope platform USA 2021 Balakrishnand et al [83] With robust edge analytics and de-riskingAn intelligent and secured heart rate monitoring system using iot India 2020 Brunese et al [85] An intelligent and secured heart rate monitoring system using iot Italy The articles were mapped according to the countries where the institution of the first author is located. Figure 3 organizes articles chronologically according to countries.…”

Machine Learning and IoT Applied to Cardiovascular Diseases Identification through Heart Sounds: A Literature Review

“…Chorba et al [83] (49) affirmed that due to the heterogeneity in the interpretative capacity of medical professionals to detect pathological characteristics during cardiac auscultation, they presented a computational approach as a promising alternative to assist in the diagnosis of pathologies in the medical area through auscultation. The authors proposed the use of deep learning algorithms with CNN architecture, using the last layer to normalize probability distribution through a softmax function as a goal of detecting murmurss and valvulopathy.…”

Machine Learning for Healthcare Wearable Devices: The Big Picture

“…Examples include irregular rhythm notification feature [ 3 ] in Apple Watch, which won U.S. Food and Drug Administration (FDA) approval with a long list of warnings and precautions in 2018 ( https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN180044.pdf ), and Eko's heart murmur detection algorithm, which has been recently published [ 4 ], which is not really for a personal wearable device but an electronic stethoscope. Additionally, some of the wearable devices that were used for monitoring, which were surveyed in [ 5 ], are no longer available in the market.…”