TY - GEN
T1 - Hybrid Transfer Learning with ECOC Ensemble Configuration for COVID-19 CXR Detection
AU - Tan, Darryl Wen Shen
AU - Wong, W. K.
AU - Juwono, Filbert H.
AU - Tiong, Teckchai
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - Coronavirus Disease 2019 (COVID-19) is a viral pneumonia that causes symptoms in the lungs of those infected. The presence of the symptoms must be diagnosed as soon as possible. If no test kits are available, the next best alternative is a computer-aided diagnostic of a patient's chest X-ray scan for a quick and accurate diagnosis. This paper proposes a hybrid transfer learning method with Error-Correction Output Codes (ECOC) by combining networks including GoogLeNet, ResNet-18, and ShuffleNet for feature extraction. X-ray input data are collected from open-source repositories. In this implementations, Support Vector Machine (SVM) as the base classifier. The proposed network attempts to categorize the input data into one of three categories: COVID-19, healthy, and non-COVID-19 pneumonia. The mean accuracy of our method is 96.21%, compared fine tuning existing pre-trained model which yielded 89.1% for GoogLeNet, 88.95% for ResNet-18, and 89.31% for ShuffleNet.
AB - Coronavirus Disease 2019 (COVID-19) is a viral pneumonia that causes symptoms in the lungs of those infected. The presence of the symptoms must be diagnosed as soon as possible. If no test kits are available, the next best alternative is a computer-aided diagnostic of a patient's chest X-ray scan for a quick and accurate diagnosis. This paper proposes a hybrid transfer learning method with Error-Correction Output Codes (ECOC) by combining networks including GoogLeNet, ResNet-18, and ShuffleNet for feature extraction. X-ray input data are collected from open-source repositories. In this implementations, Support Vector Machine (SVM) as the base classifier. The proposed network attempts to categorize the input data into one of three categories: COVID-19, healthy, and non-COVID-19 pneumonia. The mean accuracy of our method is 96.21%, compared fine tuning existing pre-trained model which yielded 89.1% for GoogLeNet, 88.95% for ResNet-18, and 89.31% for ShuffleNet.
KW - COVID-19
KW - ECOC
KW - GoogLeNet
KW - ResNet-18
KW - ShuffleNet
KW - hybrid transfer learning
UR - http://www.scopus.com/inward/record.url?scp=85147032104&partnerID=8YFLogxK
U2 - 10.1109/GECOST55694.2022.10010683
DO - 10.1109/GECOST55694.2022.10010683
M3 - Conference Proceeding
AN - SCOPUS:85147032104
T3 - 2022 International Conference on Green Energy, Computing and Sustainable Technology, GECOST 2022
SP - 155
EP - 158
BT - 2022 International Conference on Green Energy, Computing and Sustainable Technology, GECOST 2022
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2022 International Conference on Green Energy, Computing and Sustainable Technology, GECOST 2022
Y2 - 26 October 2022 through 28 October 2022
ER -