CT Images Segmentation of Lungs with COVID-19 Infection Using Mask R-CNN

The coronavirus (COVID-2019) pandemic has caused a catastrophic effect on health and global economy. The most common standard for confirming the virus relies on RT-PCR tests. As a complement to RT-PCR, Computed tomography (CT) can be used for diagnosing COVID-19. We describe the R-CNN (area-based torsional neural network) approach to segmentation of CT images of the lungs of people with COVID-19 using a variety of augmentation methods. The class imbalance problem leads to inefficient training, which makes model degenerated. In this paper, we have used a method based on Mask R-CNN to segment Left lung, right lung, Covid-19 infection. In our model, the Focal Loss function is used to suppress well-classified examples. The model is tested on COVID-19-CT-Seg-20cases dataset and the results showed that the accuracy reaches 87.93%. Compared with the smooth loss function in Mask R-CNN it improves by 5%. Therefore, this model will aid health professionals to fasten the screening and validation of the initial assessment towards COVID-19 patients.