Platform for Detecting, Managing, and Manipulating Characteristic Points of the ECG waves through Continuous Wavelet Transform Implementation.

Abstract

This work presents open-source software that incorporates detection and delineation algorithms of characteristic points of QRS complexes and P and T waves in ECG recordings. The tool facilitates the identification of significant points in the ECG waves, allowing manual correction of the results based on user criteria, exporting the detected points, and a simultaneous visualization of the recordings and the obtained points. The main objective is to improve the management of long- and short-term recordings by reducing detection errors caused by noise, interference, and artifacts, while also providing the capability for manual results correction. To achieve these objectives, the software uses an SQL Server database, which efficiently manages the data, and detection and delineation algorithms based on the continuous wavelet transform with splines, along with alternatives to optimize processing time. The QRS complex detection algorithm was validated in a previous work with the manually annotated ECG databases: MIT-BIH Arrhythmia, European ST-T, and QT. The QRS detector obtained a Se=99.91% and a P+=99.62% on the first channel of the MIT-BIH, ST-T and QT databases over the 986,930 QRS complexes analyzed. To evaluate the delineation algorithms of the characteristic points of QRS, P and T waves, the QT and PTB databases were used. The mean and standard deviations of the differences between the automatic and manual annotations by CSE experts were calculated. The mean errors range obtained was smaller than one sample (4 ms) to around two samples (8 ms); and the mean standard deviations range was around of two samples (8 ms) to six samples (24 ms).