Explainable attention-based deep learning for classification and interpretation of heart murmurs using
Bollapalli Althaph1, Nagendra Panini Challa2
1School of Computer Science and Engineering (SCOPE), VIT-AP University, Amaravati, Vijayawada, 522237, Andhra Pradesh, India.
Insights
This study introduces an explainable deep learning model for accurate heart murmur detection using phonocardiogram (PCG) signals. The framework enhances diagnostic reliability and provides visual explanations for clinical trust.
Area of Science:
- Cardiology
- Artificial Intelligence
- Biomedical Signal Processing
Background:
- Cardiovascular diseases (CVDs) pose a global health challenge, with traditional heart murmur diagnosis relying on subjective auscultation.
- Existing automated methods often lack reproducibility and clinical interpretability, hindering widespread adoption.
- There is a critical need for accurate, reproducible, and transparent diagnostic tools for heart murmurs.
Purpose of the Study:
- To develop and validate an Explainable Attention-Based Deep Learning framework for heart murmur classification and interpretation.
- To improve diagnostic accuracy and clinical trust in automated phonocardiogram (PCG) analysis.
- To provide visual explanations for model predictions, highlighting critical murmur segments.
Main Methods:
- Utilized a Transformer architecture for time-frequency feature extraction from PCG signals (spectrograms, MFCCs).
- Integrated Gradient-weighted Class Activation Mapping (Grad-CAM) for generating visual explanations of model predictions.
- Validated the framework across multiple large-scale datasets (HeartWave, CirCor DigiScope, PhysioNet, Shenzhen) using robust A-Test methods.
Main Results:
- Achieved high diagnostic performance: 96.7% accuracy, 95.5% macro-F1 score, and AUC > 0.97.
- Outperformed ten baseline models by 3-5% in accuracy and 2-4% in macro F1 score.
- Demonstrated strong alignment between model-generated explanations and expert annotations, confirming interpretability.
Conclusions:
- The proposed Explainable Attention-Based Deep Learning framework offers a significant advancement in automated heart murmur diagnosis.
- The model provides both high accuracy and crucial clinical interpretability, fostering greater trust in AI-driven cardiac diagnostics.
- Future work aims to enhance scalability and integrate multimodal data for comprehensive clinical decision support.
Abstract:
Cardiovascular diseases (CVDs) remain a leading global health challenge, necessitating diagnostic solutions that combine high accuracy with clinical interpretability and reproducibility. Traditional auscultation methods rely extensively on clinician expertise, resulting in variability and potential diagnostic delays, especially for subtle murmurs indicative of existing cardiac abnormalities. While automated methods have improved diagnostic accuracy, they frequently lack reproducibility and transparency, contributing to clinical mistrust. To address these challenges, we propose an Explainable Attention-Based Deep Learning framework specifically designed for classification and interpretation of heart murmurs using phonocardiogram (PCG) signals. Our approach employs a Transformer architecture tailored for robust time frequency feature extraction-such as spectrograms and Mel Frequency Cepstral Coefficients (MFCCs) applied to PCGs. Visual explanations generated through Gradient weighted Class Activation Mapping (Grad-CAM) explicitly highlight critical systolic and diastolic murmur segments driving the model's diagnostic predictions. We rigorously validated our framework across multiple datasets, including the HeartWave dataset (over 1,300 recordings), and further corroborated our results using CirCor DigiScope, PhysioNet, and Shenzhen datasets. Our revised validation strategy, adopting robust A-Test methods, demonstrated enhanced reliability with an accuracy of 96.7%, macro-F1 score of 95.5%, and an AUC above 0.97. Compared to ten baseline models-including CNN-RNN hybrids, ResNet variants, and Time Growing Neural Networks (TGNNs) our framework showed a 3-5% improvement in accuracy and a 2-4% increase in macro F1 score, particularly excelling in identifying rare conditions such as valvular defects and congenital anomalies. Ablation studies underscored the crucial role of attention mechanisms for both accuracy enhancement and interpretability, showing strong alignment between model-generated explanations and expert annotations. Future work will further explore model scalability, robustness in diverse clinical environments, and integration with multimodal data, including electrocardiograms, aiming for comprehensive and clinically trusted diagnostic support.
Related Concept Videos
01:15Heart Sounds
Auscultation is the process of listening to these internal body sounds using a stethoscope. The heart produces four types of sounds, but only two—S1 and S2—can usually be heard with a stethoscope.
S1, also known as the "lub" sound, is caused by the closure of atrioventricular (A-V)...
01:25Cardiovascular System Abnormal Findings II: Auscultation
Abnormal Heart Sounds
Gallops:
01:30Equipments Used To Measure Blood Pressure
This invasive approach involves cannulating a peripheral artery. During each cardiac contraction, pressure generates mechanical motion within the catheter, transmitted through rigid, fluid-filled tubing to a transducer. This transducer converts mechanical motion into electrical signals displayed as waveforms on a monitor. An automatic flushing system prevents blood backflow. Due to the potential risk of unexpected arterial blood loss, this method is primarily used in intensive...