Heart Disease Classification System Hybrid Method using CNN-LSTM

This paper proposes an ECG-based classification system for detecting heart diseases using deep learning models. The system utilizes both Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) architectures to classify ECG images into four distinct categories: myocardial infarction (MI), history of MI, abnormal heart rhythm, and normal ECG. The dataset comprises 11,148 ECG images sourced from Kaggle, which are preprocessed through resizing, normalization, and augmentation techniques. The CNN model consists of six convolutional layers followed by max-pooling layers, dropout regularization, and dense layers, achieving a test accuracy of 98.88% and an F1-score of 0.99 across all classes. Additionally, an LSTM model is incorporated to capture sequential dependencies in ECG data, further improving classification performance. A Random Forest classifier is also applied as a secondary model for comparison. The system is deployed as a web application using Streamlit, allowing real-time ECG image classification. These results highlight the robustness and clinical applicability of the system, making it a valuable tool for heart disease detection in healthcare settings.