Multiclasi ficación de arritmias cardiacas usando una red neuronal y la tarjeta MyRio-1900

Abstract

Cardiovascular diseases (CVD), and particularly cardia arrhythmias, have become one of the main
causes of death in the world, regardless of the level of development of the countries. The detection of cardiac
arrhythmias on the electrocardiogram (ECG) is a laborious task for physicians, due to the large amount of
information that must be analyzed, which could lead to inadvertent errors in diagnosis. For this reason, this work
presents an automatic system for the classification/detection of cardiac arrhythmias. To extract the vector of
characteristics of the heartbeats, a set of linear and non-linear techniques has been used to generate thirty-three
characteristics, which are used from the input of an artificial neural network (ANN) for the classification of seven
types of heartbeats. The experimental results, developed on the ECG signals from the MIT-BIH database, ordered
according to the AAMI standard, demonstrate a Cohen's Kappa index value of 0;9953, with an error of 0;04 %,
show an accuracy of 99;48 %, even under noisy conditions. Later, this system has been implemented in hardware
using the MyRio-1900 card. which is composed of a Xilinx FPGA Z-7010.