The Diagnosis of Congestive Heart Failure Based on Generalized Multiscale Entropy-Wavelet Leaders

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Congestive heart failure (CHF) presents significant mortality risks, with electrocardiogram (ECG) signals offering potential diagnostic insights.
• Manual interpretation of ECGs for CHF is challenging due to signal characteristics, leading to potential diagnostic errors.

Purpose of the Study:

• To develop and validate a novel, automated method for diagnosing congestive heart failure (CHF) using electrocardiogram (ECG) signals.
• To enhance diagnostic accuracy and efficiency for cardiologists through objective ECG interpretation.

Main Methods:

• Utilized generalized multiscale entropy (MSE)-wavelet leaders (WL) combined with extreme learning machine (ELM) for CHF diagnosis.
• Pre-processed ECG signals from normal sinus rhythm (NSR) and CHF patients, determining key parameters like segmentation time and scale factor.
• Employed two datasets (unbalanced A and balanced B) for training and testing the proposed CHF detection model.

Main Results:

• The balanced dataset (B) achieved superior performance with 99.72% accuracy, 99.46% precision, 100% sensitivity, 99.44% specificity, and 99.73% F1 score.
• The unbalanced dataset (A) demonstrated strong results with 99.56% accuracy, 99.44% precision, 99.81% sensitivity, 99.17% specificity, and 99.62% F1 score.
• The method requires fewer ECG segments, eliminates the need for R-wave detection, and improves detection probability for unbalanced datasets.

Conclusions:

• The proposed MSE-wavelet leaders and ELM-based method offers a robust and efficient approach for automated CHF diagnosis from ECG signals.
• This technique provides objective and rapid ECG interpretation, serving as valuable diagnostic assistance for cardiologists.
• The method's ability to handle unbalanced datasets and its reduced reliance on specific signal features enhance its clinical applicability.