ECG heartbeat classification using CNN-LSTM with attention mechanism

This study presents a deep learning-based framework for automatic classification of ECG. ECG signals include specific patterns known as heartbeat signals. The proposed approach utilizes CNN, CNN–LSTM, and CNN–LSTM with recurrent connections to categorize heartbeat signals. The proposed networks yield high classification accuracy. For analysis of ECG data, attention models were used. The signals were from the MIT-BIH-Noised-Simulation dataset. Class imbalance is addressed by generating additional synthetic samples. For classification, integration of machine learning methods for feature selection and classification, and implementation using software environment MATLAB. The use of convolutional layers and LSTM networks for learning temporal patterns. Introducing enhancements to relevant features via an attention mechanism. Better performance. The quality of deep learning models is often measured using a variety of evaluation metrics such as accuracy, precision, recall, F1, and the confusion matrix. In this paper, we constructed a simulation matrix. The results showed that the CNN–LSTM with Attention model got better performance than other traditional networks. These models can be deployed rapidly into real-world healthcare settings and into automated clinical diagnosis systems.