A weakly supervised deep active contour model for nodule segmentation in thyroid ultrasound images

Zhizhou Li, Shichong Zhou, Cai Chang, Yuanyuan Wang, Yi Guo

Abstract: Highlights•We propose a Weakly Supervised Deep Active Contour Model for thyroid ultrasound images to alleviate labeling burden.•A boundary gradient similarity loss function is introduced to a contour deformation network to optimize initial contour.•The detailed boundary information in the level-set is considered to optimize the segmentation at nodule boundaries.•We achieve state-of-the-art in the thyroid ultrasound image segmentation problem using weakly annotations.