Transfer Learning in ECG Classification from Human to Horse Using a Novel Parallel Neural Network Architecture

Glenn Van Steenkiste,Gunther Van Loon,Guillaume Crevecoeur

doi:10.1038/s41598-019-57025-2

Glenn Van Steenkiste, Gunther Van Loon + Show 1 more

Open Access

PDF Available

https://doi.org/10.1038/s41598-019-57025-2

Copy DOI

Export

Save

Cite

Journal: Scientific Reports	Publication Date: Jan 13, 2020
Citations: 42	License type: open-access

Affiliation: Ghent University, Flanders Make (Belgium)

Abstract
Highlights/Summary
Full-Text PDF
Similar Papers

Abstract

Listen

Automatic or semi-automatic analysis of the equine electrocardiogram (eECG) is currently not possible because human or small animal ECG analysis software is unreliable due to a different ECG morphology in horses resulting from a different cardiac innervation. Both filtering, beat detection to classification for eECGs are currently poorly or not described in the literature. There are also no public databases available for eECGs as is the case for human ECGs. In this paper we propose the use of wavelet transforms for both filtering and QRS detection in eECGs. In addition, we propose a novel robust deep neural network using a parallel convolutional neural network architecture for ECG beat classification. The network was trained and tested using both the MIT-BIH arrhythmia and an own made eECG dataset with 26.440 beats on 4 classes: normal, premature ventricular contraction, premature atrial contraction and noise. The network was optimized using a genetic algorithm and an accuracy of 97.7% and 92.6% was achieved for the MIT-BIH and eECG database respectively. Afterwards, transfer learning from the MIT-BIH dataset to the eECG database was applied after which the average accuracy, recall, positive predictive value and F1 score of the network increased with an accuracy of 97.1%.

Highlights

Automated ECG interpretation is challenging since an ECG signal can vary between and within patients under different physical circumstances[12]
The aim of this study is to develop algorithms for an end-to-end system for beat-to-beat equine ECGs (eECGs) analysis for horses based upon conventional ECG signal processing techniques along with state of the art deep learning techniques for feature extraction and classification
The advantage of using deep learning above the conventional techniques is that the essential steps, namely feature extraction, feature selection and classification can be developed without explicit definition

Summary

Introduction

Automated ECG interpretation is challenging since an ECG signal can vary between and within patients under different physical circumstances[12]. The aim of this study is to develop algorithms for an end-to-end system for beat-to-beat eECG analysis for horses based upon conventional ECG signal processing techniques along with state of the art deep learning techniques for feature extraction and classification. In most automated ECG interpretation studies, the authors concentrate on conventional machine learning approaches: pre-processing, feature extraction, feature reduction and feature classification[16]. Another study by Isin and Ozdalili[21] converts R-T segments to 256 × 256 × 3 images and uses the convolutional layers of a pre-trained alexNet[22] as neural network architecture to extract the features, that on their turn are fed into a hidden layer after a principal component analysis is applied to reduce the number of features. Transfer learning for ECGs has already successfully been used in different studies for human ECG classification, but has not yet been applied between different species[17,21,27]

Objectives

Methods

Results

Conclusion