Heart Murmur and Abnormal PCG Detection via Wavelet Scattering Transform & a 1D-CNN

TL;DR

The paper addresses automated heart murmur and abnormal PCG detection using two publicPhysioNet datasets and a trio of neural models. By combining denoising, segmentation, noise relabeling, normalization, and time–frequency feature extraction with wavelet scattering, the authors show that a 1D-CNN consistently outperforms LSTM-RNN and CRNN, especially after cleaning the CirCor Digiscope 2022 data. The approach achieves state-of-the-art-like performance on murmur detection (notably in E3) and surpasses prior methods on abnormal PCG detection (E4), highlighting the value of fixed-filter wavelet features for small datasets. The work also provides practical preprocessing and evaluation strategies, including a voting-based aggregation and careful handling of class imbalance, with potential clinical impact for early valve disease diagnosis and future directions in murmur grading and data generation.

Abstract

Heart murmurs provide valuable information about mechanical activity of the heart, which aids in diagnosis of various heart valve diseases. This work does automatic and accurate heart murmur detection from phonocardiogram (PCG) recordings. Two public PCG datasets (CirCor Digiscope 2022 dataset and PCG 2016 dataset) from Physionet online database are utilized to train and test three custom neural networks (NN): a 1D convolutional neural network (CNN), a long short-term memory (LSTM) recurrent neural network (RNN), and a convolutional RNN (C-RNN). We first do pre-processing which includes the following key steps: denoising, segmentation, re-labeling of noise-only segments, data normalization, and time-frequency analysis of the PCG segments using wavelet scattering transform. We then conduct four experiments, first three (E1-E3) using PCG 2022 dataset, and fourth (E4) using PCG 2016 dataset. It turns out that our custom 1D-CNN outperforms other two NNs (LSTM-RNN and C-RNN). Further, our 1D-CNN model outperforms the related work in terms of accuracy, weighted accuracy, F1-score and AUROC, for experiment E3 (that utilizes the cleaned and re-labeled PCG 2022 dataset). As for experiment E1 (that utilizes the original PCG 2022 dataset), our model performs quite close to the related work in terms of weighted accuracy and F1-score.

Figures (8)

• Figure 1: Cardiac auscultation locations: (PCG 2022 dataset) 2022challengepaper.
• Figure 2: Graphical user interface for guided re-labeling of dataset segments.
• Figure 3: PCG time-series and the corresponding WST for the two situations (murmur present, murmur absent) across the four heart valves, i.e., AV, PV, TV, MV (for CirCor Digiscope 2022 dataset). Note that WST is quite effective in differentiating between the two classes (murmur present, murmur absent).
• Figure 4: 1D-CNN Model architecture: KS (kernel size), SS(stride size), Pad (padding).
• Figure 5: CRNN Model architecture: KS (kernel size), SS (stride size), Pad (padding).