Heart failure detection using deep learning and Gradient Boosting classifier

Authors : Ahmad Ahmad

Pages : 1-8

View : 99 | Download : 70

Publication Date : 2025-01-31

Article Type : Research Paper

Abstract :Heart failure (HF) is marked by a diminished capacity of the heart to effectively pump blood. Traditionally, the electrocardiogram (ECG) has served as a non-invasive diagnostic tool, gauging the heart\\\'s electrical activity and rhythm. Recent advancements have leveraged machine learning (ML) and deep learning (DL) techniques to automate the identification and classification of HF types from ECG data. This study introduces a novel deep learning architecture, blending the efficacy of a convolutional neural network (CNN) for feature extraction with an eXtreme Gradient Boosting (XGBoost) layer for final classification. The first CNN model operates on ECG segments in the time domain, while the second CNN processes the Continuous Wavelet Transform (CWT) of the same segments. This composite model offers superior automatic HF detection, particularly with 2-second ECG fragments, by capturing intricate features from both time and frequency domains. Training and testing utilize datasets from the MIT-BIH, BIDMC, and PTB Diagnostic ECG databases. Through 10-fold cross-validation, the proposed approach attains remarkable accuracy, sensitivity, and F1-score, all surpassing 99.9\\\\%. This modality represents a significant stride in DL applications for ECG diagnosis, holding promise for enhanced clinical utility.
Keywords : Convolutional neural network, deep learning, electrocardiogram, eXtreme Gradient Boosting, heart failure.