Automated identification of innocent Still's murmur using a convolutional neural network

Abstract

BackgroundStill's murmur is the most prevalent innocent heart murmur of childhood. Auscultation is the primary clinical tool to identify this murmur as innocent. Whereas pediatric cardiologists routinely perform this task, primary care providers are less successful in distinguishing Still's murmur from the murmurs of true heart disease. This results in a large number of children with a Still's murmur being referred to pediatric cardiologists.ObjectivesTo develop a computer algorithm that can aid primary care providers to identify the innocent Still's murmur at the point of care, to substantially decrease over-referral.MethodsThe study included Still's murmurs, pathological murmurs, other innocent murmurs, and normal (i.e., non-murmur) heart sounds of 1,473 pediatric patients recorded using a commercial electronic stethoscope. The recordings with accompanying clinical diagnoses provided by a pediatric cardiologist were used to train and test the convolutional neural network-based algorithm.ResultsA comparative analysis showed that the algorithm using only the murmur sounds recorded at the lower left sternal border achieved the highest accuracy. The developed algorithm identified Still's murmur with 90.0% sensitivity and 98.3% specificity for the default decision threshold. The area under the receiver operating characteristic curve was 0.943.ConclusionsStill's murmur can be identified with high accuracy with the algorithm we developed. Using this approach, the algorithm could help to reduce the rate of unnecessary pediatric cardiologist referrals and use of echocardiography for a common benign finding.

A Noise-Robust Heart Sound Segmentation Algorithm Based on Shannon Energy

“…Seventeen of the 20 highest-scoring papers in the 2016 PhysioNet Challenge used segmentation (See Table 1). More recent papers also support this finding [10], [11], [16]- [20]. One could argue that segmentation is not the only reason why these methods performed better, but, given that the accuracies of the top six entries ranged within 2% despite using different classifiers, segmentation appears to be a dominant factor.…”

“…Heart sound segmentation (HSS)-identification of the primary heart sounds (S1 and S2)-is often a prerequisite in computerized auscultation. It has been reported to be used in many AI-based heart murmur classification pipelines [10], [12], [13]. However, some deep-learning approaches skip this step.…”

StethAid: A Digital Auscultation Platform for Pediatrics

“…Still’s murmur is the most common innocent murmur in children. It is often difficult to recognize by PCPs [ 12 , 14 , 25 ]. We are conducting a multicenter data acquisition to create the necessary training and validation sets to develop an automated Still’s murmur identification algorithm.…”

“…We have collected 470 recordings. Each recording is 15 s long and is recorded at the lower left sternal border (LLSB) chest location [ 25 ]. The dataset consists of 265 Still’s murmur recordings and 205 pathological heart murmur recordings.…”

Heart Murmur Classification Using a Capsule Neural Network

“…Huang et al [9] aimed to develop an intelligent diagnostic method for detecting heart murmurs in patients with ventricular and atrial septal defects. Shekhar et al [10] developed a computer algorithm to assist primary care providers in identifying Still's murmur in children, thereby decreasing overreferral to pediatric cardiologists. Additionally, there have been studies on the use of machine learning techniques for heart disease prediction, such as the ANN-based approach [2] for the detection and identification of congenital heart disease in pediatric patients and the cardiovascular disease prediction model based on the improved deep belief network [11].…”