Abstract—The spread of the coronavirus disease 2019 (COVID-19) epidemic has resulted in massive loss of life and economic loss all over the world. Computed Tomography (CT) has been proven to help us screen out patents more effectively and accurately, but there are not enough experts to interpret CT images. Reliable automatic Artificial Intelligence (AI) recognition systems for COVID-19 CT scans are incredibly vital due to this situation. Moreover, the backbone network for feature extraction, e.g. the deep Residual Network (ResNet), plays a fundamental role in these advanced systems. However, the original design of residual connection way can hardly combine the local extracted features, which is important for the interpretation of CT images. In recent years, the relationship between ResNets and the dynamical system has drawn wide attention, and to combine extracted features, our study explores the high order numerical differential formula from this perspective: merging several standard residual blocks into one as advanced high-order scheme in several aspects with no extra parameters. Compared to the original residual connection method, the network with our proposed sixth-order residual block achieves 83.11% accuracy on the CCAP dataset and has better stability and convergence speed performance without adding parameters.
Cite: Hao Huang and Sei-ichiro Kamata, "Classification of COVID-19 on Chest CT Scans with Higher Order Residual Network," International Journal of Pharma Medicine and Biological Sciences, Vol. 11, No. 4, pp. 70-75, October 2022. doi: 10.18178/ijpmbs.11.4.70-75

Copyright © 2022 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.