Cardiac Pathologies Detection and Classification in 12-lead ECG

Cardiac Pathologies Detection and Classification in 12-lead ECG

Stať ve sborníku v databázi WoS či Scopus

Background: Automatic detection and classification of cardiac abnormalities in ECG is one of the basic and often solved problems. The aim of this paper is to present a proposed algorithm for ECG classification into 19 classes. This algorithm was created within PhysioNet/CinC Challenge 2020, name of our team was HITTING. Methods: Our algorithm detects each pathology separately according to the extracted features and created rules. Signals from the 6 databases were used. Detector of QRS complexes, T-waves and P-waves including detection of their boundaries was designed. Then, the most common morphology of the QRS was found in each record. All these QRS were averaged. Features were extracted from the averaged QRS and from intervals between detected points. Appropriate features and rules were set using classification trees. Results: Our approach achieved a challenge validation score of 0.435, and full test score of 0.354, placing us 11 out of 41 in the official ranking. Conclusion: The advantage of our algorithm is easy interpretation. It is obvious according to which features algorithm decided and what thresholds were set.

Background: Automatic detection and classification of cardiac abnormalities in ECG is one of the basic and often solved problems. The aim of this paper is to present a proposed algorithm for ECG classification into 19 classes. This algorithm was created within PhysioNet/CinC Challenge 2020, name of our team was HITTING. Methods: Our algorithm detects each pathology separately according to the extracted features and created rules. Signals from the 6 databases were used. Detector of QRS complexes, T-waves and P-waves including detection of their boundaries was designed. Then, the most common morphology of the QRS was found in each record. All these QRS were averaged. Features were extracted from the averaged QRS and from intervals between detected points. Appropriate features and rules were set using classification trees. Results: Our approach achieved a challenge validation score of 0.435, and full test score of 0.354, placing us 11 out of 41 in the official ranking. Conclusion: The advantage of our algorithm is easy interpretation. It is obvious according to which features algorithm decided and what thresholds were set.

ECG classification, cardiac pathologies classification

ECG classification, cardiac pathologies classification

SMÍŠEK, R.; NĚMCOVÁ, A.; MARŠÁNOVÁ, L.; SMITAL, L.; VÍTEK, M.; KOZUMPLÍK, J.

2021

30.12.2020

IEEE

Rimini, Italy

Computing in Cardiology 2020

2325-887X

Computing in Cardiology

47

1

Spojené státy americké

1

4

4

http://www.cinc.org/archives/2020/pdf/CinC2020-171.pdf

http://hdl.handle.net/11012/196704

CinC2020-171