Classification of Arrhythmia Using Deep Learning

The arrhythmia is a problem of the irregular heartbeat or irregular rhythm of heartbeat. It can occur due to stress, tensions and of taking caffeine, nicotine, etc. This may result either in increase of heartbeat or decrease of heartbeat. There are six different types of arrhythmia and are classified as left bundle branch block, right bundle branch block, premature atrial contraction, premature ventricular contraction, normal arrhythmia, ventricular fibrillation. The existing CNN algorithm shows an accuracy of 87% in prediction of the arrhythmia which uses the data collected from the sensors of smart watches. This paper proposes an idea that uses a modified CNN algorithm that will take ECG reports as input and will predict whether the person is suffering from the arrhythmia or not. The novelty is the use of original ECG reports than using the smart watch generated reports. It not only predicts the presence of the arrhythmia but also helps in classification of the arrhythmia. The use of ECG records will provide more efficient results than the watch sensors which are not much accurate and have more fluctuations. Also, the cost of implementation is also reduced as it just takes the ECG reports as the input. The input is given to the CNN algorithm which passes through various layers of CNN. This model has shown an accuracy of 99.98%.