Artificial intelligence for rapid identification of the coronavirus disease 2019 (COVID-19)

Xueyan Mei; Hao-Chih Lee; Kaiyue Diao; Mingqian Huang; Bin Lin; Chenyu Liu; Zongyu Xie; Yixuan Ma; Philip M. Robson; Michael Chung; Adam Bernheim; Venkatesh Mani; Claudia Calcagno; Kunwei Li; Shaolin Li; Hong Shan; Jian Lv; Tongtong Zhao; Junli Xia; Qihua Long; Sharon Steinberger; Adam Jacobi; Timothy Deyer; Marta Luksza; Fang Liu; Brent P. Little; Zahi A. Fayad; Yang Yang

Pertinent images contain pulmonary tissue and a potential parenchymal abnormality. For the selection of pertinent slices, image segmentation was used to detect parenchymal tissue. The raw CT images all had a 512 x 512 matrix storing CT intensities in Hounsfield Units (HU). A standard lung window (width (w)=1500 HU and level (l)=-600 HU) was used to normalize each slice to pixel intensities between 0 and 255. We segmented CT images into two parts, body and lung. The body part was segmented by finding the largest connected component consisting of pixels with an intensity greater than 175. The segmented connected component was filled into a solid region. The lung region was defined as the pixels with intensity less than 175 that fall within the segmented body part. Small regions with less than 64 pixels were removed, as they are typically segmented due to random noise. The lung region was enlarged by 10 pixels to fully include the pleural boundary. We discard images if the size of the lung was smaller than 20% of the size of the body part.