Anatomy-Anchored Self-Supervision: Distilling Vision Foundation Models for Invariant Ultrasound Representation
Abstract
This work proposes AnaUS, an anatomy-level self-supervised pretraining framework for ultrasound representation learning.
Type
Publication
International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2026)
AnaUS discovers unlabeled anatomical structures with an LP-SAM driven latent prompt engine and combines anatomy disentanglement alignment with core-region prediction for robust ultrasound representation learning.
