Cardiac Conduction Simulation: A Hybrid Deep Learning Approach to Classify Realistic ECG Signals

Cardiac conditions such as myocardial infarction (MI) and arrhythmias pose significant diagnostic challenges, making accurate classification of ECG signals critical for effective treatment and prevention. This project focuses on a comprehensive comparative analysis of four advanced deep learning algorithms—CNN, MobileNet, DenseNet, and an ensemble model combining MobileNet with LSTM—to classify ECG images into four categories: myocardial infarction, history of MI, abnormal heartbeat, and normal heart conditions. The dataset comprises labeled ECG images, enabling the models to learn both spatial and temporal features critical for accurate classification. Each algorithm’s performance is evaluated using metrics like accuracy, precision, recall, and F1-score. Results demonstrate that the ensemble model outperforms individual architectures, leveraging MobileNet's spatial feature extraction and LSTM's sequential pattern recognition to achieve superior accuracy. This approach showcases the potential for robust, automated diagnostic tools in clinical applications. Future work aims to incorporate multi-lead ECG signals and additional metadata to further enhance the system's reliability and scalability for accurate classification in cardiac healthcare systems.

Keywords:
ECG classification, cardiac conditions, CNN, MobileNet, DenseNet, LSTM, hybrid model, deep learning, myocardial infarction, comparative analysis, diagnostic tools.