Automated ECG Arrhythmia classification using Resnet and AutoML Learning Model

The prevalence of cardiovascular diseases (CVDs) means that they account for a large percentage of all deaths worldwide. Based on the development of AI, various automatic classifications of cardiac arrhythmias have recently been successfully applied to numerous models. But during training, most models separate the intrinsic properties of each lead in a 12-lead ECG on their own, leaving them short on inter-lead features to automate the classification of normal rhythms and 26 cardiac diseases. In this paper, we present a systematic approach that combines Auto-CardiacML's (Auto-ML) classification with ResNet-50's feature extraction techniques (Auto-Resnet). The model uses a 12-lead ECG that was digitally reconstructed from the original. Auto-ML is used to classify cardiac arrhythmias, and the ResNet model is used to get both the inner and interlead features of an ECG at the same time. It has been determined, based on experimental results from the CPSC 2018 test set, that our model has an average accuracy of 0.82 for distinguishing normal rhythm from cardiac arrhythmias. In the future, the results will be used to make a structure that uses both PCG and ECG signals.