Model-driven Heart Rate Estimation and Heart Murmur Detection based on Phonocardiogram

Jingping Nie,Ran Liu,Behrooz Mahasseni,Erdrin Azemi,Vikramjit Mitra

2024-07-26

Abstract:Acoustic signals are crucial for health monitoring, particularly heart sounds which provide essential data like heart rate and detect cardiac anomalies such as murmurs. This study utilizes a publicly available phonocardiogram (PCG) dataset to estimate heart rate using model-driven methods and extends the best-performing model to a multi-task learning (MTL) framework for simultaneous heart rate estimation and murmur detection. Heart rate estimates are derived using a sliding window technique on heart sound snippets, analyzed with a combination of acoustic features (Mel spectrogram, cepstral coefficients, power spectral density, root mean square energy). Our findings indicate that a 2D convolutional neural network (\textbf{\texttt{2dCNN}}) is most effective for heart rate estimation, achieving a mean absolute error (MAE) of 1.312 bpm. We systematically investigate the impact of different feature combinations and find that utilizing all four features yields the best results. The MTL model (\textbf{\texttt{2dCNN-MTL}}) achieves accuracy over 95% in murmur detection, surpassing existing models, while maintaining an MAE of 1.636 bpm in heart rate estimation, satisfying the requirements stated by Association for the Advancement of Medical Instrumentation (AAMI).

Sound,Machine Learning,Audio and Speech Processing

What problem does this paper attempt to address?

The paper attempts to address the problem of estimating heart rate (HR) and detecting heart murmurs using Phonocardiogram (PCG) data. Specifically, the researchers utilize publicly available PCG datasets to achieve heart rate estimation through a model-driven approach and further extend it to a Multi-Task Learning (MTL) framework to perform both heart rate estimation and heart murmur detection simultaneously.

### Main Issues:
1. **Heart Rate Estimation**: How to accurately estimate heart rate from short segments of heart sound signals.
2. **Heart Murmur Detection**: How to achieve both heart rate estimation and heart murmur detection within the same model.

### Background and Challenges:
- **Complex Environmental Noise**: Environmental noise and noise introduced by body movements in the real world pose significant challenges to heart rate estimation and heart murmur detection.
- **Limitations of Traditional Methods**: Traditional signal processing methods perform poorly in handling these noises as they struggle to adapt to strong distribution changes and large fluctuations in different scenarios.
- **Advantages of Deep Learning**: Deep learning can learn complex patterns, showing potential in handling noisy PCG signals.

### Solution:
- **Feature Extraction**: The researchers used various acoustic features (such as Mel spectrogram, Mel-frequency cepstral coefficients, power spectral density, and root mean square energy) to characterize the audio signals.
- **Model Design**: A model based on 2D Convolutional Neural Network (2D-CNN) was designed for heart rate estimation, and on this basis, a multi-task learning model (2D-CNN-MTL) was constructed to perform both heart rate estimation and heart murmur detection.
- **Performance Evaluation**: The performance of the model was evaluated using metrics such as Mean Absolute Error (MAE) and Accuracy.

### Main Contributions:
- **High-Precision Heart Rate Estimation**: The 2D-CNN model achieved an MAE of 1.312 bpm in heart rate estimation, meeting the requirements of the Association for the Advancement of Medical Instrumentation (AAMI).
- **Efficient Multi-Task Learning**: The 2D-CNN-MTL model achieved an accuracy of over 95% in heart murmur detection while maintaining a low heart rate estimation error (1.636 bpm).

### Application Prospects:
- **Remote Patient Monitoring**: This technology can be used for remote patient monitoring, enhancing self-care capabilities.
- **Resource-Limited Environments**: In resource-limited environments, this technology can be used for better disease management, early diagnosis, and treatment.

In summary, this paper successfully addresses the issues of heart rate estimation and heart murmur detection through a model-driven approach, combining various acoustic features and deep learning techniques, and demonstrates its potential in practical applications.