This repository contains the official implementation of RC-Net, proposed in our paper "Rethinking Region Consistency in Semi-supervised 3D Medical Image Segmentation".
RC-Net introduces a novel approach to improve weak-to-strong consistency methods, emphasizing the foreground-background paradigm to enhance segmentation performance, particularly in challenging medical imaging tasks. Our method effectively reduces edge uncertainty and improves robustness in texture-blurred scenarios.
The code for RC-Net will be made publicly available upon acceptance of the paper. Stay tuned for updates!