GLMKD: Joint global and local mutual knowledge distillation for weakly supervised lesion segmentation in histopathology images

Hangbei Cheng, Xueyu Liu, Jun Zhang, Xiaorong Dong, Xuetao Ma, Yansong Zhang, Hao Meng, Xing Chen, Guanghui Yue, Yidi Li, Yongfei Wu

Abstract: Highlights•Present a weakly-supervised framework for lesion segmentation based on global-local information.•Combine global supervisory information and multi-scale MIL via dual-stream knowledge distillation.•Propose MMIL model effectively captures subtle lesion variations in lesions across scales.•A shape transfer loss function improves global-local knowledge exchange.