Automatic Lobe Segmentation Using Attentive Cross Entropy And End-to-End Fissure Generation

Automatic lung lobe segmentation algorithm is of great significance for the diagnosis and treatment of lung diseases, however, which has great challenges due to the incompleteness of pulmonary fissure in lung CT images and the large variability of pathological features. Therefore, we propose a new automatic lung lobe segmentation method, which can be applied to any learning-based segmentation models. Our method is based on the assumption that accurate prediction of pulmonary fissure plays a dominant role in the whole lung lobe segmentation task, therefore we urge the model to pay attention to the area around the pulmonary fissure during the training process, which is realized by a task-specific loss function and end-to-end lung fissure generation method. We achieve 97.83% and 94.75% dice scores on our private dataset STLB and public LUNA16 dataset respectively.