Deep Learning Based Heart Murmur Detection using Frequency-time Domain Features of Heartbeat Sounds

Jungguk Lee1, Taein Kang1, Narin Kim1, Soyul Han1, Hyejin Won1, Wuming Gong2, Il-Youp Kwak1
1Chung-Ang University, 2University of Minnesota, Lillehei heart institute


The goal of the George B. Moody PhysioNet Challenge 2022 was to use heart sound recordings gathered from various auscultation locations to identify murmurs and clinical outcomes. Our team, CAU_UMN, propose a deep learning-based model that automatically identifies heart murmurs from a phonocardiogram (PCG). We converted the heartbeat sound into 2D features in the frequency-time domain through feature extraction techniques such as log-mel spectrogram, Short Time Fourier Transform (STFT), and Constant Q Transform (CQT). The frequency-temporal 2D features were modeled using voice classification models such as Convolutional neural networks (CNN) and Light CNN (LCNN). The model using log-melspectrogram and LCNN was ranked 31 out of 303 submitted methods with a murmurs score of 0.734 and 24 out of 303 submitted methods with a outcomes score of 9493 in the official phase of the George B. Moody PhysioNet Challenge. We believe that our deep learning based heart murmur detection system will be a promising system for automatic heart murmur detection from PCG.