Classifying ECG exams of different formats and sources Using Convolutional Networks

ECG is an important exam to detect cardiac conditions. New algorithms to aid the classification of these exams are arising, but although the exam is widely available, the format of the generated data is not standardized. This paper presents an algorithm developed to detect abnormalities in the exam...

Full description

Saved in:
Bibliographic Details
Published in2020 IEEE International Conference on E-health Networking, Application & Services (HEALTHCOM) pp. 1 - 4
Main Authors dos Santos De Oliveira, Jessica, Marques, Clement Bernardo, Wanderley, Maria Fernanda, Wagner, Priscilla Koch, Filho, Walter Martins
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.03.2021
Subjects
artificial intelligence
Classification algorithms
data extraction
ECG
Electrocardiography
image processing
Myocardium
Rhythm
signal analysis
Online AccessGet full text

Cover

More Information
Summary:ECG is an important exam to detect cardiac conditions. New algorithms to aid the classification of these exams are arising, but although the exam is widely available, the format of the generated data is not standardized. This paper presents an algorithm developed to detect abnormalities in the exam and classify them between rhythm and beat abnormalities. Three different sources were used with at least six different formats in total. The results show the generalization power of the model, achieving an AUC of 0.96 to detect rhythm conditions, 0.91 to detect abnormalities in general, and a minimum AUC of 0.84 to detect specific beat conditions.
DOI:10.1109/HEALTHCOM49281.2021.9399032