Enhanced ECG Signal Analysis Using Hybrid CNN-LSTM and Augmentation Techniques

The nature and timely nature of the detection of the arrhythmias are critical in preventing severe cardiac repercussions like stroke or sudden cardiac death. First, this research proposes a deep learning model that integrates convolutional neural network (CNN) as well as recurrent neural network with long short-term memory (LSTM) for the classification of arrhythmias from ECG data. CNN component is responsible for extracting geographical features whereas LSTM identifies temporal dependency, creating a good base for understanding consecutive ECG signals. MIT-BIH Arrhythmia Database was used for training and validation, showing near perfect performance of the model with an accuracy of 98.2%, precision of 97.8%, recall of 98.5%, and F1-score of 98.1%. Preprocessing measures such as normalization, R-peak detection for signal segmentation and augmentation ensured data accuracy and generality. The comparative study with standalone models revealed that the proposed hybrid architecture outperformed CNN (accuracy: 92. Traditional classifiers, like SVM, SVR, attained 85.3% accuracy, while the universal LSTM model achieved an impressive 94.2%, RNN 80.5% and GRU 7%. This was further supported by ROC curves where AUC of the model was high on five different classes of arrhythmia. For extensibility and compliance with clinical decision-making processes, forecaster was incorporated into the framework with SHAP which ensure that forecasts are explicable and believable to many health care professionals.