Automatic ECG Classification on Mobile Devices Jose Vigno Moura Sousa, Vilson Rosa de Almeida, Aratã Andrade Saraiva, Pedro Mateus Cunha Pimentel, Luciano Lopes de Sousa

Abstract

In this work, a new method is proposed to classify electrocardiogram signals in mobile devices that can classify different arrhythmias according to the EC57 standard of the Association for the Advancement of Medical Instrumentation. A convolutional neural network was built, trained and validated with the MIT-BIH arrhythmia data set, in which this database has 5 different classes: normal beat, premature supraventricular beat, premature ventricular contraction, ventricular beat fusion, normal and unclassifiable beat. After being trained and validated, the model is submitted to a post-training quantization stage using the TensorFlow Lite conversion method. The results obtained were satisfactory, before and after quantization, the convolutional neural network obtained an accuracy of 98.5%. With the quantization technique it was possible to obtain a reduction in the size of the model, thus enabling the development of the mobile application, this reduction was approximately 90% in relation to the size of the original model.