Detection of Dysrhythmias from Electrocardiogram Signals Using Noise Removal and Deep Learning Techniques

Heart related diseases have created a major loss in human lives. Healthy heart ensures good standard of living from infant to aged people for a long period. Predicting heart related diseases prior to serious cause would mostly reduce death and save human life. One such heart related issue is dysrhythmias which is an unstable rhythm occurs in the normal heart beat of the humans due to improper electrical dispositions that synchronizes the heartbeat. This endangered disease would definitely lead to the end of life irrespective of the age group. This unstable rhythm would be diagnosed mostly using Electrocardiogram (ECG) which is an electrical waveform that depicts the heartbeat. This research aimed to discover a novel method that makes use of the ECG waveforms as a training and testing dataset with the most incipient Deep Learning (DL) algorithms’ to predict dysrhythmia earlier. The DL models like Convolutional Neural Network (CNN) and Bi directional Long Short Term Memory (BLSTM) are combined to experiment the input data by segmenting the signals. A noise removal technique was used to remove the noise in the waveforms of the ECG data and built a two dimensional image which is the input to the proposed method. The proposed method produced remarkable results as 99.36 percent in accuracy, 0.99 as F1 score. The objective of the investigation is to save human lives from serious and severe dysrhythmias by predicting it as early as possible by using the proposed methodology.