Implementación de un sistema de anotación y evaluación de los detectores de QRS: PAN&TOMPKINS, basado en la transformada de ondaletas sobre electrocardiogramas de esfuerzo

Abstract

The aim of this project is to develop a semiautomatic annotation RR time-series system using an eight leads stress database named DICARDIA. In a first stage the user realizes a visual inspection validating or denying the channel proposed by the system based in statistics measure. In the second stage the system has two options based in the quality of RR-time series, if the series contains few artifacts the annotations is realized using interval of annotations, in the case of noisy series, the system allows annotations beat by beat. Finally, an annotation file is generated under standard formats that are used to evaluate QRS detectors. The performance measures were the Sensibility (Se) and positive predictive value (P+). The results realized over 255432 beats provide a Se of 97.347630% and P+ of 96.84413% for the Pan & Tompkins (PT) and Se of 95.540104% and P+ of 93.0337086% for based wavelet transform (WT). The procedure developed in this work can be seen as a valuable starting point in semiautomatic annotation of large electrocardiographic databases, as well to evaluate and to improve stress ECG delineations and provides of delineations to evaluate the performance measure of QRS detectors. This work has been published in two indexed journals. The semiautomatic annotation system will be presented and published in TIC.EC 2016 memories and the evaluation of QRS detection algorithm will be published in AndesCon 2016 memories.