Open AccessRESEARCH OF THE CETLIN METHOD OF AUTOMATIC ARRHYTHMIA DETECTION BY ECG SIGNALS FROM MIT-BIH
Author(s) -
Aleksei K. Khalaidzhi
Publication year - 2021
Publication title -
avtomatizaciâ processov upravleniâ
Language(s) - English
Resource type - Journals
ISSN - 1991-2927
DOI - 10.35752/1991-2927-2021-1-63-98-109
Subject(s) - false positive paradox , computer science , quality (philosophy) , segmentation , set (abstract data type) , pattern recognition (psychology) , artificial intelligence , signal (programming language) , duration (music) , data mining , art , philosophy , literature , epistemology , programming language
This article presents and solves the problem of evaluating the quality of the Cetlin method, which classifies the sequence of RR-intervals by the recordings of ECG signals from MIT-BIH, which have labels on R-peaks. To solve this problem, author proposes new quality metrics and describes developed algorithms for calculating them in real time with taking into account segmentation errors. Author analyzes the influence of the accuracy of the segmentation procedure for extracting the positions of R-peaks from ECG signal on the proposed quality metrics. Paper compares the quality of the Cetlin method and other existing algorithms for arrhythmia detection that analyze the duration of RR-intervals in accordance with a set of rules in real time. Article reveals advantages and limitations of the method. Paper shows that the method successfully detects SVEB and VEB arrhythmias. but has inertia, that leads to false positives, and is immune to morphological abnormalities that do not change the duration of RR-intervals. Author analyzes the influence of parameters of the Cetlin method on its quality according to the proposed metrics. Paper describes the procedure for searching the best parameters configuration. In conclusion, author reveals that there is no the only configuration, that achieves the best quality for each signal from MIT-BIH.