Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Intelligent Cardiac Auscultation for Murmur Detection via Parallel-Attentive Models with Uncertainty Estimation

Zixing Zhang, Tao Pang, Jing Han, Björn W. Schuller

TL;DR

This work tackles automatic murmur detection from phonocardiograms by presenting a parallel-attentive transformer that blends global self-attention with local convolutional processing. By incorporating Monte Carlo Dropout for uncertainty estimation and temperature scaling for calibration, the approach yields reliable predictions on the CirCor Digiscope dataset, achieving a leading weighted accuracy of $79.8\%$ and macro-F1 of $65.1\%$, surpassing prior state-of-the-art methods. The combination of local-global feature learning and probabilistic calibration enhances trust and decision support in a clinical setting, addressing both performance and reliability. The findings suggest that uncertainty-aware, calibrated models can facilitate practical deployment of automated auscultation tools in cardiovascular screening and diagnosis.

Abstract

Heart murmurs are a common manifestation of cardiovascular diseases and can provide crucial clues to early cardiac abnormalities. While most current research methods primarily focus on the accuracy of models, they often overlook other important aspects such as the interpretability of machine learning algorithms and the uncertainty of predictions. This paper introduces a heart murmur detection method based on a parallel-attentive model, which consists of two branches: One is based on a self-attention module and the other one is based on a convolutional network. Unlike traditional approaches, this structure is better equipped to handle long-term dependencies in sequential data, and thus effectively captures the local and global features of heart murmurs. Additionally, we acknowledge the significance of understanding the uncertainty of model predictions in the medical field for clinical decision-making. Therefore, we have incorporated an effective uncertainty estimation method based on Monte Carlo Dropout into our model. Furthermore, we have employed temperature scaling to calibrate the predictions of our probabilistic model, enhancing its reliability. In experiments conducted on the CirCor Digiscope dataset for heart murmur detection, our proposed method achieves a weighted accuracy of 79.8% and an F1 of 65.1%, representing state-of-the-art results.

Intelligent Cardiac Auscultation for Murmur Detection via Parallel-Attentive Models with Uncertainty Estimation

TL;DR

This work tackles automatic murmur detection from phonocardiograms by presenting a parallel-attentive transformer that blends global self-attention with local convolutional processing. By incorporating Monte Carlo Dropout for uncertainty estimation and temperature scaling for calibration, the approach yields reliable predictions on the CirCor Digiscope dataset, achieving a leading weighted accuracy of and macro-F1 of , surpassing prior state-of-the-art methods. The combination of local-global feature learning and probabilistic calibration enhances trust and decision support in a clinical setting, addressing both performance and reliability. The findings suggest that uncertainty-aware, calibrated models can facilitate practical deployment of automated auscultation tools in cardiovascular screening and diagnosis.

Abstract

Heart murmurs are a common manifestation of cardiovascular diseases and can provide crucial clues to early cardiac abnormalities. While most current research methods primarily focus on the accuracy of models, they often overlook other important aspects such as the interpretability of machine learning algorithms and the uncertainty of predictions. This paper introduces a heart murmur detection method based on a parallel-attentive model, which consists of two branches: One is based on a self-attention module and the other one is based on a convolutional network. Unlike traditional approaches, this structure is better equipped to handle long-term dependencies in sequential data, and thus effectively captures the local and global features of heart murmurs. Additionally, we acknowledge the significance of understanding the uncertainty of model predictions in the medical field for clinical decision-making. Therefore, we have incorporated an effective uncertainty estimation method based on Monte Carlo Dropout into our model. Furthermore, we have employed temperature scaling to calibrate the predictions of our probabilistic model, enhancing its reliability. In experiments conducted on the CirCor Digiscope dataset for heart murmur detection, our proposed method achieves a weighted accuracy of 79.8% and an F1 of 65.1%, representing state-of-the-art results.
Paper Structure (13 sections, 7 equations, 4 figures, 4 tables)

This paper contains 13 sections, 7 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Parallel-Attentive Modelling
  5. Uncertainty Estimation with Monte Carlo Dropout
  6. Calibration with Temperature Scaling
  7. Experiments
  8. Selected Dataset
  9. Implementation Details
  10. Results and Discussion
  11. Accuracy
  12. Uncertainty and Calibration
  13. Conclusion

Figures (4)

  • Figure 1: The flowchart of the proposed murmur detection system.
  • Figure 2: Illustration of the Parallel-Attentive Modelling.
  • Figure 3: Confidence Histogram (above) and Reliability Diagram (below) on the segments (left) and patients (right) before the confidence calibration.
  • Figure 4: Confidence Histogram (above) and Reliability Diagram (below) on the segments (left) and patients (right) after the confidence calibration.