Classificação de subclasses de fibrilação atrial utilizando estatística de alta ordem e aprendizado de máquina

Abstract

The electrocardiogram (ECG) is a simple and routine procedure of great importance for the diagnosis of cardiac pathologies. This exam gives us a graphic representation of the electrical activity of the heart, which results in its interpretation, as waves, segments and possible intervals for measuring and identifying the changes it presents in the cardiac organ. This dissertation aims to develop a classification model based on the beats of four groups of desired: with paroxysmal atrial fibrillation, intracardiac atrial fibrillation, atrial fibrillation and normal sinus rhythm. The methodology of extraction of characteristics based and adapted to classify with Atrial Fibrillation, its subtypes and healthy, with and without the use of the Independent Component Analysis (ICA) technique. As evaluated, they were evaluated based on the characteristics of the statistics of the four databases, evaluating as metrics the K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Artificial Neural Network (ANN) algorithms, obtaining accuracy of 93.4% to 99.85%.