A Comparison Study to Detect COVID-19 Chest X-Ray Images with SOTA Deep Learning Models

Qingzhong Liu, Zhongxue Chen, Henry C. Liu
Proceedings of the 1st Workshop on Healthcare AI and COVID-19, ICML 2022, PMLR 184:146-153, 2022.

Abstract

IBy using a recently released chest X-ray (CXR) image database for COVID-19 positive cases along with Normal, Lung Opacity (Non-COVID lung infection), and Viral Pneumonia images, this study compares the performance of SOTA deep learning models in detecting COVID-19 CXR images. Pre-trained deep learning models are retrained under several combinations of optimizers, learning rate schedulers, and loss functions. Our study shows that these SOTA deep learning models perform well if the models and parameters are selected meticulously. Overall, EfficientNet is superior to others especially across different optimizers. Regarding the loss function, the integration of cosine embedding similarity and cross entropy is slightly better than cross entropy itself while we adopt the SGD optimizer. In terms of optimizer, SGD constantly performs well while Adam and AdamW are unstable across different models.

Cite this Paper

BibTeX
@InProceedings{pmlr-v184-liu22a, title = {A Comparison Study to Detect COVID-19 Chest X-Ray Images with SOTA Deep Learning Models}, author = {Liu, Qingzhong and Chen, Zhongxue and Liu, Henry C.}, booktitle = {Proceedings of the 1st Workshop on Healthcare AI and COVID-19, ICML 2022}, pages = {146--153}, year = {2022}, editor = {Xu, Peng and Zhu, Tingting and Zhu, Pengkai and Clifton, David A. and Belgrave, Danielle and Zhang, Yuanting}, volume = {184}, series = {Proceedings of Machine Learning Research}, month = {22 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v184/liu22a/liu22a.pdf}, url = {https://proceedings.mlr.press/v184/liu22a.html}, abstract = {IBy using a recently released chest X-ray (CXR) image database for COVID-19 positive cases along with Normal, Lung Opacity (Non-COVID lung infection), and Viral Pneumonia images, this study compares the performance of SOTA deep learning models in detecting COVID-19 CXR images. Pre-trained deep learning models are retrained under several combinations of optimizers, learning rate schedulers, and loss functions. Our study shows that these SOTA deep learning models perform well if the models and parameters are selected meticulously. Overall, EfficientNet is superior to others especially across different optimizers. Regarding the loss function, the integration of cosine embedding similarity and cross entropy is slightly better than cross entropy itself while we adopt the SGD optimizer. In terms of optimizer, SGD constantly performs well while Adam and AdamW are unstable across different models.} }
Endnote
%0 Conference Paper %T A Comparison Study to Detect COVID-19 Chest X-Ray Images with SOTA Deep Learning Models %A Qingzhong Liu %A Zhongxue Chen %A Henry C. Liu %B Proceedings of the 1st Workshop on Healthcare AI and COVID-19, ICML 2022 %C Proceedings of Machine Learning Research %D 2022 %E Peng Xu %E Tingting Zhu %E Pengkai Zhu %E David A. Clifton %E Danielle Belgrave %E Yuanting Zhang %F pmlr-v184-liu22a %I PMLR %P 146--153 %U https://proceedings.mlr.press/v184/liu22a.html %V 184 %X IBy using a recently released chest X-ray (CXR) image database for COVID-19 positive cases along with Normal, Lung Opacity (Non-COVID lung infection), and Viral Pneumonia images, this study compares the performance of SOTA deep learning models in detecting COVID-19 CXR images. Pre-trained deep learning models are retrained under several combinations of optimizers, learning rate schedulers, and loss functions. Our study shows that these SOTA deep learning models perform well if the models and parameters are selected meticulously. Overall, EfficientNet is superior to others especially across different optimizers. Regarding the loss function, the integration of cosine embedding similarity and cross entropy is slightly better than cross entropy itself while we adopt the SGD optimizer. In terms of optimizer, SGD constantly performs well while Adam and AdamW are unstable across different models.
APA
Liu, Q., Chen, Z. & Liu, H.C.. (2022). A Comparison Study to Detect COVID-19 Chest X-Ray Images with SOTA Deep Learning Models. Proceedings of the 1st Workshop on Healthcare AI and COVID-19, ICML 2022, in Proceedings of Machine Learning Research 184:146-153 Available from https://proceedings.mlr.press/v184/liu22a.html.

Related Material