Gated recurrent unit-based heart sound analysis for heart failure screening.

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Heart failure (HF) is a cardiovascular disease impacting cardiac structure and function.
• Early HF screening is crucial for timely and effective treatment.
• Heart sounds (HS) contain vital information for HF diagnosis.

Purpose of the Study:

• To develop an automated tool for identifying normal heart sounds (HS) and differentiating between HF with preserved ejection fraction (HFpEF) and HF with reduced ejection fraction (HFrEF).
• To leverage deep learning for efficient and accurate HF screening using HS signals.

Main Methods:

• A novel framework utilizing a gated recurrent unit (GRU) model for HF screening.
• Employing a logistic regression-based hidden semi-Markov model for HS frame segmentation.
• Using normalized HS frames as input for the GRU model to automatically extract deep features for HF classification.

Main Results:

• The proposed GRU model achieved a high average accuracy of 98.82% in HF screening.
• GRU model performance surpassed other comparison methods evaluated.
• The model successfully identified normal, HFpEF, and HFrEF subjects.

Conclusions:

• The GRU model effectively learns features directly from HS, reducing reliance on expert knowledge.
• Heart sound analysis using the GRU model proves effective for early heart failure screening.
• This automated approach offers a promising tool for clinical application in HF detection.