Deep Learning Methods for Heart Sounds Classification: A Systematic Review.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Automated heart sound classification is crucial for diagnosing cardiovascular diseases (CVDs).
• Deep learning (DL) approaches are increasingly used for heart sound analysis, driven by big data and AI advancements.
• Current DL methods face limitations including data scarcity, inefficient training, and suboptimal model performance.

Purpose of the Study:

• To systematically review and analyze deep learning methods for heart sound classification.
• To focus on convolutional neural network (CNN) and recurrent neural network (RNN) models developed in the last five years.
• To identify challenges and future trends in applying DL to heart sound classification.

Main Methods:

• Systematic literature review of deep learning techniques for heart sound classification.
• Analysis focused on CNN and RNN architectures from the past five years.
• Discussion of limitations and future research directions.

Main Results:

• Deep learning, especially CNNs and RNNs, offers significant potential for accurate heart sound classification.
• Identified limitations include data availability, training efficiency, and model generalizability.
• The study highlights the need for further research to overcome these challenges.

Conclusions:

• Deep learning methods, particularly CNNs and RNNs, are advancing heart sound classification for cardiovascular disease diagnosis.
• Addressing data limitations and improving training efficiency are key for developing more effective models.
• This review serves as a reference for future research in AI-driven cardiac auscultation.