A Hybrid CNN-LSTM Model for Heart Failure Detection Using Raw ECG Signals

Heart failure (HF) represents a significant burden on healthcare systems that affects more than 64.3 million people worldwide. Its complications contribute to the annual demise of approximately 7 million people. This complex syndrome encompasses three distinct classes based on left ventricle ejection fraction (LVEF), i.e., preserved (HFpEF), mid-range (HFmEF), and reduced (HFrEF). Early and accurate diagnosis of these subclasses is crucial for optimal management and prognosis. This study focuses on utilizing raw electrocardiogram (ECG) signals, without the need for feature engineering, for the precise detection of HF classes. The latter is achieved by leveraging a deep learning architecture consisting of Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) layers. The proposed model demonstrated promising results with an overall accuracy of 86%, with accuracy for detecting HFpEF, HFmEF, and HFrEF classes at 89%, 80%, and 84%, respectively. By integrating deep learning techniques with raw ECG data from remote sensing and wearable devices, healthcare providers can achieve early diagnosis and personalized management of HF, ultimately improving patient outcomes and enabling real-time monitoring.