AI-Enhanced Detection of Heart Murmurs: Advancing Non-Invasive Cardiovascular Diagnostics

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Cardiovascular diseases (CVDs) are a leading global cause of mortality, with over 17 million deaths annually.
• Early detection of heart murmurs, indicative of valve abnormalities, is crucial for patient outcomes.
• Current diagnostic methods face limitations in accessibility, cost, and invasiveness, particularly in resource-limited settings.

Purpose of the Study:

• To develop and evaluate a novel Convolutional Recurrent Neural Network (CRNN) model for non-invasive heart murmur classification.
• To assess the model's performance in accurately identifying heart murmurs from audio recordings.
• To explore the potential of AI in enhancing cardiac diagnostics and accessibility.

Main Methods:

• Utilized a CRNN model integrating convolutional and recurrent layers to process heart sound recordings.
• Applied pre-processing techniques including z-score normalization, band-pass filtering, and data augmentation (Gaussian noise, time shift, pitch shift).
• Trained and validated the model on heart sound datasets to capture spatial and temporal audio features.

Main Results:

• The CRNN model achieved high diagnostic performance with 90.5% accuracy.
• The model demonstrated strong precision (89%) and recall (87%) in heart murmur classification.
• The study highlights the model's robustness and effectiveness in analyzing cardiac auscultation data.

Conclusions:

• The developed CRNN model shows significant potential for revolutionizing cardiac diagnostics through non-invasive heart murmur detection.
• This AI-driven approach offers a scalable and accessible solution for early identification of cardiovascular conditions.
• The technology holds promise for broader applications in healthcare, especially in regions with limited access to traditional diagnostic resources.