Medical Internet of Things for Classification of Pathological ECG Beats Based on Fractional Fourier Transform and Hyperparameter Tuning

Abstract

The Medical Internet of Things (MIoT) has recently played a key role in developing functional health systems. As a result, automatic detection and prediction of future risks such as heart valve diseases and arrhythmias are still being researched and studied. Additionally, early detection of heart problems can improve treatment and reduce patient mortality. On the other hand, traditional approaches did not produce good results for accurate diagnosis. This paper proposes electrocardiogram (ECG) beat classification using Deep Transfer Learning (DTL) and hyperparameter tuning. After a frequency domain transformation with the Fractional Fourier Transform, images of ECG signals were captured (FrFT). The framework uses multi-access edge computing technology, allowing end users to access available resources and our DTL Model in the cloud. The proposed automated model incorporates a Convolutional Neural Network (CNN) structure with hyperparameter tuning. Our model is validated using the MIT-BIH database. Finally, we classified heart disease into five categories. According to the experimental results, the developed framework could classify ECG signals with 99.68 percent accuracy. The proposed method is more accurate and efficient than other well-known and popular algorithms when compared to other current methods.