Variance Approximation and Probabilistic Decomposition Noise Removal Framework for Arrhythmia Detection and Classification on Internet of Medical Things Environment

In the Internet of Medical Things (IoMT), medical devices and applications are linked via networking to create a healthcare data framework. In the IoMT environment, a healthcare tracking framework is required to detect health abnormalities on a proper basis and prescribe treatments based on the data obtained from wearable sensors. In the IoMT platform, an Electrocardiogram (ECG) helps to analyze the electrical function of the heart in a simplified and comfortable way. ECG monitoring is used to diagnose heart diseases, such as arrhythmias. ECG signal analysis automatically detect and classify life-threatening arrhythmias associated with various cardiac situations. Arrhythmia diagnosis and classification can be misinterpreted when ECG signals are corrupted with different noises. To overcome such interpretation Variance Approximation and Probabilistic Decomposition Noise Removal method (VA-PDNR) has been developed to denoise the ECG signal for arrhythmia detection and classification. VA has been used to reduce high-frequency noise in ECG signals by controlling the operational variables to minimize noise variation. Furthermore, PDNR breaks up the power line interference into different mode parameters and remove noise by utilizing Linear Quadratic Estimation. The dimension-reduction analysis aims to remove redundancies and verify characteristic textural characteristics. All such hybrid features are blended and fed to the Artificial Neural Network (ANN) for Arrhythmia classification. Improves the signal-to-noise ratio (SNR) to 34.55 dB, the accuracy to 98.99 per cent, and the mean square error rate. VA-PDNR method is estimated using ECG signals collected from MIT-BIH Arrhythmia dataset and improves Signal-to-Noise Ratio (SNR) to 34.55 dB, accuracy to 98.99%, and lesser mean square error (MSE) rate.