Lung segmentation and automatic detection of COVID-19 using radiomic features from chest CT images

Highlights：

• A classification model using statistical analysis with high interpretability is proposed for differentiating COVID-19 infection from community-acquired pneumonia and healthy controls. The model using texture features from our segmentation results achieved an AUC of 0.9470, a sensitivity of 0.9500, and a specificity of 0.9270.

• The proposed approach achieves promising results in both lung segmentation and automatic detection of COVID-19. It has a great promise for clinical use in facilitating automatic diagnosis of COVID-19 infection on chest CT images.

摘要

•A new deep-learning-based method which integrates a 3D V-Net with shape priors is proposed to extract pulmonary parenchyma from chest CT images. Compared with manually delineated lung contours, the proposed segmentation method achieved a dice similarity coefficient of 0.9796, a sensitivity of 0.9840, a specificity of 0.9954 and a mean surface distance error of 0.0318 mm.•A classification model using statistical analysis with high interpretability is proposed for differentiating COVID-19 infection from community-acquired pneumonia and healthy controls. The model using texture features from our segmentation results achieved an AUC of 0.9470, a sensitivity of 0.9500, and a specificity of 0.9270.•The proposed approach achieves promising results in both lung segmentation and automatic detection of COVID-19. It has a great promise for clinical use in facilitating automatic diagnosis of COVID-19 infection on chest CT images.