search for




 

Transfer learning-based ensemble deep learning for image classification of COVID-19 patients
Journal of the Korean Data & Information Science Society 2021;32:1219-35
Published online November 30, 2021;  https://doi.org/10.7465/jkdi.2021.32.6.1219
© 2021 Korean Data and Information Science Society.

Ji Hye Heo1 · Su Bin I2 · Won Hyuk Yang3 · Dong Hoon Lim4

14Department of Information Statistics, Gyeongsang National University
2Department of Mathematics, Gyeongsang National University
3Department of Technology Fusion Engineering, Konkuk University
Correspondence to: 1 Graduate student, Department of Information Statistics, Gyeongsang National University, Jinju 52828, Korea.
2 Undergraduate student, Department of Mathematics, Gyeongsang National University, Jinju 52828, Korea.
3 Undergraduate graduation, Department of Technology Fusion Engineering, Konkuk University, Seoul 06226, Korea.
4 Professor, Department of Information Statistics, Department of Bio and Medical Big Data and RINS, Gyeongsang National University, Jinju 52828, Korea. dhlim@gnu.ac.kr
Received October 13, 2021; Revised November 7, 2021; Accepted November 11, 2021.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
The damage caused by the COVID-19 pandemic has a serious impact not only on public health but also on politics, economy, society, and culture as a whole. To date, the RT-PCR test, a COVID-19 standard diagnostic test, may vary depending on the type of sample, sample collection method, and storage, and is also affected by the time of the test after infection with COVID-19. This paper attempts to classify COVID-19 patients with X-ray/CT images using transfer learning-based ensemble deep learning. The transfer learning used here is the AlexNet, ResNet, Inception V3, and DenseNet models based on the convolutional neural network (CNN). The stacking ensemble model proposed in this study takes place over three stages. In the first step, predicted results are obtained using several transfer learning models, in the second step, they are combined through a concatenate layer, and in the third step, a deep neural network (DNN) model is applied and finally classified. For the performance evaluation of the ensemble model proposed in this paper, three actual COVID-19 X-ray/CT image datasets were considered, and various performance evaluation indicators were compared and analyzed with the transfer learning model and the existing ensemble model. As a result of the performance experiment, the overall proposed ensemble model was superior to the transfer learning model and the existing ensemble model.
Keywords : COVID-19, deep learning, stacking ensemble, transfer learning, X-ray/CT image.