Weakly Supervised Segmentation of COVID19 Infection with Scribble Annotation on CT Images

Highlights：

• An uncertainty-aware mean teacher framework is incorporated into the proposed method to guide the model training, encouraging the segmentation predictions to be consistent under different perturbations for an input image. With the pixel level uncertainty measure on the predictions of the teacher model, the student model is guided with reliable supervision. We further regularize the model with a transformation-consistent strategy, which is beneficial for the segmentation task and makes our approach easier to deal with the segmentation of irregular lesion areas.

• We evaluated the proposed method on three datasets and compared it with other advanced approaches. The results demonstrated the superiority of the proposed method.

摘要

•We propose a weakly supervised COVID-19 infection segmentation method with scribble supervision. To the best of our knowledge, it is the first work adopting scribble-level supervision in COVID-19 segmentation.•An uncertainty-aware mean teacher framework is incorporated into the proposed method to guide the model training, encouraging the segmentation predictions to be consistent under different perturbations for an input image. With the pixel level uncertainty measure on the predictions of the teacher model, the student model is guided with reliable supervision. We further regularize the model with a transformation-consistent strategy, which is beneficial for the segmentation task and makes our approach easier to deal with the segmentation of irregular lesion areas.•We evaluated the proposed method on three datasets and compared it with other advanced approaches. The results demonstrated the superiority of the proposed method.