Long-short diffeomorphism memory network for weakly-supervised ultrasound landmark tracking

Zhihua Liu, Bin Yang, Yan Shen, Xuejun Ni, Sotirios A. Tsaftaris, Huiyu Zhou

Abstract: Highlights•A novel multi-task long-short diffeomorphism memory network for ultrasound landmark analysis.•An expectation maximization memory align mechanism to iteratively update both long and short deformation utilizing the physical relation.•End-to-end weakly-supervised training with few landmark annotations for tracking and zero annotation for deformation learning.•LSDM can achieve better or competitive landmark tracking performance with high interpretation and strong generalization.