Classification of Abnormalities in ECG using Machine Learning

Continuous health tracking using implantable and portable medical devices with cellular networks is envisioned as a transformative healthcare solution. Rapid advances and in electronics and sensors provides low power remote monitoring medical devices and machine learning methods to classify the abnormalities in the monitoring parameter. However, primary issues faced by the remote monitoring devices are storage memory battery and optimal processor load. This paper presents an internet of things based health parameter monitoring system for Ambulatory electrocardiogram (ECG) recorders, which are increasingly in use by people suffering from cardiac abnormalities. However, the ECG signal acquired by the ambulatory recorder is influenced by motion artifacts induced by any Body Movement Activity (BMA). In the proposed methodology, the Heart Rate Variability method combined with Modified Support Vector Machine (MSVM) based Machine Learning algorithm are used to analyze the abnormality of the acquired ECG signal. The ECG signals are simulated from the standard database and the proposed methodology is applied to the ECG signals. The results shows the proposed methodology able to extract the data from ECG to identify the abnormality using HRV parameters and continuous remote monitoring.