Skip to main content
. 2021 Mar 19;25(6):1864–1872. doi: 10.1109/JBHI.2021.3067465

Fig. 2.

Fig. 2.

The proposed network architecture for COVID-19 lesion localization. The encoder-decoder and the patch discriminator work together to remove potential COVID-19 lesions from input CT slices. Shallow layers of the pre-trained ResNet-18 are used to perform low-level feature match between input and output to increase the image quality of the generated CT slices. Another ResNet-18 is used to classify the localization maps.