Abstract: Highlights•Proposing a novel dual consistency network for semi-supervised medical image segmentation.•Model-level and task-level consistency complement each other.•Enforcing the model to output more accurate predictions in ambiguous areas and at segmentation boundaries.
Loading