Customized T-time inner sampling network with uncertainty-aware data augmentation strategy for multi-annotated lesion segmentation

Xi Zhou, Xinxin Wang, Haiqin Ma, Jianjian Zhang, Xiaomei Wang, Xiuxiu Bai, Li Zhang, Jia Long, Jiakuan Chen, Hongbo Le, Wenjie He, Shen Zhao, Jun Xia, Guang Yang

Abstract: Highlights•Custom t-time network for lesion segmentation generates hypotheses in one pass.•Uncertainty defined for medical images with multiple annotations or predictions.•Strategy augments data based on uncertainty probability.