A Deep Learning Approach for the Classification of Arrhythmias in ECG Signal

ECG or electrocardiogram assesses an individual's cardiac rhythm as a signal. This signal is the significant result of repolarisation and depolarisation of heart's four chambers, through which voltage can be interpreted over time. The presumption of this paper is the implementation of supervised deep learning to identify the visualisation and illustration of labelled rhythmic aberrations. The proposed technique uses a series of one dimensional convolutional paired with set of multilayer perceptron to classify and detect the most common arrhythmias. The model was trained with 75% of the data available which was first sampled, and then, tested on the natural distribution of data. The accuracy of model proved to be very efficient after the implementation of the model trained by using supervised deep learning and some techniques of signal processing. Thus, providing an accuracy of 98.3% with a tolerance value of 0.0040, iterating over multiple number of iterations. Future scope of this proposition includes different processing techniques with slight adjustments within the model and its architecture.