CHF Detection with LSTM Neural Network.

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Heart rate variability (HRV) is a recognized predictor of heart failure risk.
• Traditional HRV analysis relies on manual feature extraction, introducing potential errors and reducing robustness.
• Congestive heart failure (CHF) detection requires reliable and automated methods.

Purpose of the Study:

• To introduce a deep learning model for robust heart failure risk prediction.
• To evaluate the performance of a long short-term memory (LSTM) network for HRV analysis.
• To assess the feasibility of automated CHF detection using intelligent systems.

Main Methods:

• A deep learning approach utilizing a long short-term memory (LSTM) network was developed.
• The LSTM model was applied to heart rate variability data without pre-processing.
• Three different RR interval lengths (N=50, N=100, N=500) were used for analysis.

Main Results:

• The LSTM method achieved accuracies of 82.47% (N=50), 85.13% (N=100), and 84.91% (N=500).
• The model demonstrated high performance across varying RR interval lengths.
• No pre-processing was required, simplifying the analysis pipeline.

Conclusions:

• Deep learning, specifically LSTM, offers a robust and accurate method for heart failure prediction from HRV.
• This automated approach can potentially be integrated into intelligent hardware or mobile applications for early CHF detection.
• The findings support the use of AI in improving cardiovascular risk assessment.