CNNAH: Modified CNN to Arrhythmia Heatbeat Classification using ECG signals

Cardiovascular Diseases (CVDs) can be fatal, therefore, quick identification ensures proper diagnosis and management. This paper aims to accurately classify different heartbeats such as Normal Beats (NB), Supraventricular Premature Beats (SPB), Premature Ventricular Contractions (PVC), Fusion of Ventricular and Normal beats (FVN) and Unclassifiable beats (UB) using Convolutional Neural Networks (CNNs). This paper proposes various CNN models aimed at enhancing performance by varying convolutional layers and dropout rates. Signal acquisition using electrocardiogram (ECG) sensors connected to Arduino boards forms the basis of training and testing with MIT- BIH Arrhythmia Dataset from Kaggle. In this paper, various CNN architectures are explored in relation to their effect on classification accuracy. This demonstrates the high level of accuracy achieved at the rate of 96.9% which is worth admiring for an augmentation in the number of convolutional layers without concurrent reduction in the dropout rates. The proposed paper suggests that the designed varied-architecture CNNs are more effective in correctly categorizing different heart beats and hence advances the diagnostic process for arrhythmias. The proposed model provides high recognition accuracy, making this research significant.