Detección de fibrilación auricular en electrocardiogramas de corta duración (Detection of atrial fibrillation in short duration electrocardiograms)

Abstract

In this chapter, we present the general guidelines in the application of two machine learning algorithms to detect a common cardiac arrhythmia, atrial fibrillation (AF), when employing short-duration recordings taken through a portable single-lead electrocardiographic (ECG) electronic device. Due to the importance of the early diagnosis of cardiovascular pathologies such as AF, our goal is to improve the classification performance of the mobile device that, in practice, leaves a relevant set of ECG recordings unclassified. We analyze the performance of supervised classification techniques such as recursive partitioning and random forests
in combination with ECG signal feature extraction methods. Our methodology applies to an international ECG training set and a national test set of ECG recordings generated in 2019 for the elder adult population of Uruguay, under a collaboration
between the CHSCV and the Ibirapitá Plan. The available diagnoses of the ECG signals performed by expert clinical cardiologists allow the interpretation of the obtained results.