Image Segmentation For pneumothorax disease Based On based on Nested Unet Model

Pneumothorax on lung can be caused by a blunt chest injury, damage from underlying lung disease or Covid-19 Virus. Using CT scanning to examine high-risk people is an important task for many doctors and hospital. With the development of machine learning techniques, computer-aided diagnosis is widely used in pneumothorax detection. In this paper, we proposed a nested Unet model with a backbone of EfficientNet. This model used many skip pathway connections in many layers to reduce the semantic gap between networks. We choose dice loss as our experiment metrics, which is widely used in segmentation task. The lower Dice loss is, the better performance the model has. Compared with the simple Unet model or the other models, the experiments show that our model has better performance.