HDEC – Enhanced Classification Heart Disease Detection by ECG Signals

Early prediction and detection of heart disease is crucial for rapid intervation and improved patient outcomes. This study proposes HDEC (Heart Disease Detection using ECG), a hybrid framework combining Kalman filtering, Convolutional Neural Networks (CNN), and XGBoost to enhance ECG-based classification accuracy. ECG, a non-invasive tool, captures the heart's electrical activity, offering insights into rhythm, rate, and potential abnormalities. Raw ECG signals often contain noise, which can obscure critical patterns. In this framework, a Kalman filter is applied to estimate the true cardiac signals while suppressing noise, producing cleaner data for analysis. Feature extraction is then performed in both time and frequency domains to identify key characteristics relevant to heart disease diagnosis. To indicate cardiac abnormities, signals are denoised and then converted into time-frequency representations further fed into a CNN to learn complex i. e. irregular patterns. Features extracted from CNN are classified using XGBoost algorithm, as gradient Boost algorithm is best algorithm for structured data. The combined approach effectively categorizes signals as normal or diseased, achieving improved classification accuracy compared to standalone methods. A robust framework is demonstrated by the incorporation of Kalman filtering, CNN feature learning, and XGBoost classification for machine driven ECG-based heart disease detection, results in real-time diagnostic applications in clinical settings.