UCAD: Uncertainty-Guided Contour-Aware Displacement for Semi-Supervised Medical Image Segmentation

by Chengbo Ding, Fenghe Tang, and Shaohua Kevin Zhou.

Introduction

This repo is the official implementation of UCAD: Uncertainty-Guided Contour-Aware Displacement for Semi-Supervised Medical Image Segmentation which is accepted to ISBI-2026.

Requirements

This repository is based on PyTorch 2.4.1, CUDA 12.8 and Python 3.8.20. All experiments in our paper were conducted on an NVIDIA GeForce RTX 4090 GPU with an identical experimental setting.

Usage

We provide code and model for ACDC and Synapse dataset.

Data could be got at ACDC and Synapse.

Note: Please adjust the arguments in these scripts according to specific experimental settings.

To train a model,

cd code
# You should generate superpixel segments first
python generate_masks_offline.py
# Training
CUDA_VISIBLE_DEVICES=0 python train_UCAD.py

To test a model,

CUDA_VISIBLE_DEVICES=0 python eval.py

Acknowledgements

Our code is largely based on BCP. Thanks for these authors for their valuable work, hope our work can also contribute to related research.

Questions

If you have any questions, welcome contact me at '[email protected]'