Improving Accuracy Through Preprocessing and Data Augmentation Techniques with a Deep Learning-Based Approach for Arrhythmia Detection

Abstract

Arrhythmia detection plays a critical role in the early diagnosis and management of cardiovascular diseases. In this study, we propose a deep learning-based model for arrhythmia classification using advanced preprocessing and data augmentation techniques. The proposed model is evaluated on the MIT-BIH Arrhythmia Dataset and the PTB Diagnostic ECG Dataset and achieves 98% and 95% accuracy rates, respectively. These results demonstrate the strong ability of the model to classify complex heartbeat patterns, achieving higher accuracy, precision, sensitivity, and F1 score compared to existing methods. The model uses a convolutional neural network (CNN) architecture trained on pre-processed ECG signals with data segmented into individual heartbeats. Data augmentation techniques are applied to reduce data imbalances and improve the generalization ability of the model. Experimental results highlight that the model provides a significant increase in accuracy rates over traditional methods. The results of this study highlight the potential of deep learning architectures in biomedical signal analysis, especially for realtime arrhythmia detection. This approach offers promising potential for clinical applications by enabling higher diagnostic accuracy and timely intervention in cardiovascular healthcare.