Rule-Based Method and Deep Learning Networks for Automatic Classification of ECG

Abstract

The objective of the study is to explore the potentiality of combining a classical rule-based method with a Deep Learning method for automatic classification of ECG for participation in PhysicNet/Computing in Cardiology Challenge 2020. Six databases are considered for training set. They consist 43101 12 -leads ECG recording, lasting from 6 to 60 seconds considering 24 diagnostic classes. The rule-based method is using morphological and time-frequency ECG descriptors, characterizing each diagnostic labels. These rules have been extracted from the knowledge-base of a physician, with no direct learning procedure in the first phase, while a refinement have been tested in the second phase. The Deep Learning method consider both raw ECG signals and median beat signals. These data are processed by continuous wavelet transform analysis obtaining a time-frequency domain representtation, with the generation of specific images. These images are used for training Convolutional Neural Networks for ECG diagnostic classification. Official result of the classification accuracy of the ECGs Test set of our team named ‘Gio_Ivo’ produced a challenge validation score of 0.325 for the rule based method, and a 0.426 for the Deep learning methodology with GoogleNet, which was chosen for the final score, obtaining a full test score of 0.298, placing us 12th out of 41 in the official ranking.