CAE-UNet: An Effective Automatic Segmentation Model for CT Images of COVID-19

Abstract

Since December 2019, COVID-19 has ravaged the world, severely affecting the quality of life and physical health of human society. Computed tomography (CT) imaging is an effective way to detect solid lung lesions as well as pulmonary ground-glass nodules and is an effective way to diagnose COVID-19. The automatic and accurate segmentation of COVID-19 lesion areas from CT images can determine the severity of the disease, which is essential for the diagnosis and treatment of COVID-19. A new model CAE-UNet(Combine-ASPP-ECA-UNet) is proposed in this paper for COVID-19 CT image segmentation based on UNet. The coding structure of UNet is replaced with the improved ResNet50 and incorporated with ECA attention module and atrous spatial pyramid pooling(ASPP). Fusing different sensory fields, global, local and spatial features to enhance the detail segmentation effect of the network. The experimental results on the CC-CCII show that the mIoU of the proposed CAE-UNet reaches 79.53%, which is better than some other mainstream methods. The proposed method achieves automatic and efficient segmentation of COVID-19 CT images.