T-WAVE ALTERNANS DETECTION: A MACHINE LEARNING AND DEEP LEARNING APPROACH IN PYTHON

Abstract

Sudden Cardiac Death (SCD) is a significant global health issue, accounting for a substantial proportion of cardiovascular disease-related mortality. SCD often results from ventricular arrhythmias, which can occur without prior warning, making early detection crucial. One of the promising indicators for predicting SCD is T- wave alternans (TWA), which are subtle variations (order of ¿V ) in the amplitude or shape of the T-wave in an electrocardiogram (ECG). Detecting TWA is essential as it can help identify individuals at high risk of SCD, enabling timely interventions. In this study, the creation of a comprehensive database by collecting and pre- processing ECG signals was developed. The preprocessing steps included the addition of alternans, removal of baseline wander, delineation of the ST-T complex, heartbeat windowing, ST-T alignment, background subtraction, linear filtering, and building the signals dataframe. These steps ensured the quality and consistency of the data used for analysis. Relevant features were extracted from the preprocessed ECG signals, high- lighting the Kscore, Valt and the noble cumsum-based feature, to improve the de- tection accuracy of TWA. Various machine learning algorithms were employed, including Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), and Support Vector Machine (SVM), to evaluate their performance in detecting TWA. Furthermore, we implemented advanced deep learning techniques, specif- ically Long Short-Term Memory (LSTM) networks and a hybrid CNN-LSTM ar- chitecture. Our results demonstrated that LSTM networks outperformed other models, achieving an outstanding accuracy of 94.08%, which is the highest per- formance reported in the literature for TWA detection. This research highlights the potential of LSTM networks in improving the early detection of TWA, thereby contributing to the prevention of SCD. The findings suggest that incorporating deep learning models, particularly LSTM networks, can significantly enhance the accuracy and reliability of TWA detection, offering a robust tool for clinicians in the fight against SCD