This article was originally published here
Phys Eng Sci Med. 2022 Mar 14. doi: 10.1007/s13246-022-01110-w. Online ahead of print.
COVID-19 is a deadly epidemic that has been declared a public health emergency of international concern. The massive damage of the disease on public health, social life and the world economy increases the importance of alternative methods of rapid diagnosis and monitoring. The RT-PCR test, which is considered the gold standard in diagnosing the disease, is complicated, expensive, time consuming, prone to contamination and can give false negative results. These disadvantages reinforce the trend towards medical imaging techniques such as computed tomography (CT). Typical visual signs such as ground glass opacity (GGO) and consolidation of CT images allow for quantitative assessment of the disease. In this context, it aims at the segmentation of infected lung CT images with the residual network-based DeepLabV3+, which is a redesigned convolutional neural network (CNN) model. In order to assess the robustness of the proposed model, three different segmentation tasks like task-1, task-2 and task-3 were applied. Task-1 represents binary segmentation as lung (infected and uninfected tissues) and background. Task-2 represents multi-class segmentation as lung (uninfected tissue), COVID (GGO, consolidation, and pleural effusion irregularities are brought together under one roof), and background. Finally, the segmentation in which each lesion type is considered as a separate class is defined as Task-3. The COVID-19 imaging data for each segmentation task consists of 100 single-slice CT scans from over 40 diagnosed patients. Model performance was assessed using Dice’s similarity coefficient (DSC), intersection over union (IoU), sensitivity, specificity, and precision by performing a fivefold cross-validation. . The average DSC performance for three different segmentation tasks was obtained at 0.98, 0.858 and 0.616, respectively. The experimental results demonstrate that the proposed method has robust performance and great potential in the assessment of COVID-19 infection.