Research map / Label-efficient and robust learning / Domain adaptation and generalization
Multi-Channel Uncertainty-Weighted Score Matching for Conditional Diffusion in Medical UDA
CHEN LI, Meilong Xu, Xiaoling Hu, Weimin Lyu, Chao Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add multi-channel uncertainty-weighted score matching for CDM training (ablation) | Ours-PL · BraTS (T2->T1CE) | Dice (avg, %): 52.88 -> 58.16 | Improved |
| add pixel-wise high-confidence collapse threshold (ablation) | Ours-PL · BraTS (T2->T1CE) | Dice (avg, %): 48.66 -> 58.16 | Improved |
| add CDM-generated labeled target images to pseudo-labeling (ablation) | mean-teacher pseudo-labeling · BraTS (T2->T1CE) | Dice (avg, %): 52.55 -> 58.16 | Improved |
| add real source images to generated data for segmenter training (ablation) | Ours-PL · BraTS (T2->T1CE) | Dice (avg, %): 49.51 -> 58.16 | Improved |
| use two ranked confidence maps (k=2) instead of one (ablation) | UPDiff-UDA · BraTS (T2->T1CE) | Dice (avg, %): 52.88 -> 58.16 | Improved |
| replace random Bezier curve with feature-guided Bezier adaptation (ablation) | initial segmenter · BraTS (T2->T1CE) | Dice (x10^2): 41.88 -> 48.45 | Improved |
| replace histogram matching with Bezier adaptation (ablation) | initial segmenter · BraTS (T2->T1CE) | Dice (x10^2): 11.98 -> 48.45 | Improved |
| use Bezier adaptation instead of CyCADA style transfer end-to-end (ablation) | UPDiff-UDA · BraTS (T2->T1CE) | Dice (avg, %): 48.76 -> 58.16 | Improved |
| raise confidence threshold delta from 0.7 to 0.8 (vs 0.9) (ablation) | Ours-PL · BraTS (T2->T1CE) | Dice (avg, %): 54.10 -> 58.16 | Improved |
| use ResNet34 instead of ResNet18 backbone (ablation) | Ours-PL · BraTS (T2->T1CE) | Dice (avg, %): 53.46 -> 58.16 | Improved |
| use ResNet50 instead of ResNet34 backbone (ablation) | Ours-PL · BraTS (T2->T1CE) | Dice (avg, %): 58.16 -> 57.17 | No clear effect |
| add generated target data to ADVENT | ADVENT · Abdominal Multi-Organ (MRI->CT) | Dice (avg, %): 44.08 -> 82.38 | Improved |
Closest papers by meaning
- A Unified Degradation-Robust Approach to SSL and UDA for 3D Medical Images (AAAI 2025 · similarity 0.89)
- MAPSeg: Unified Unsupervised Domain Adaptation for Heterogeneous Medical Image Segmentation Based on 3D Masked Autoencoding and Pseudo-Labeling (CVPR 2024 · similarity 0.87)
- DyCON: Dynamic Uncertainty-aware Consistency and Contrastive Learning for Semi-supervised Medical Image Segmentation (CVPR 2025 · similarity 0.87)
- MedFlowSeg: Flow Matching for Medical Image Segmentation with Frequency-Aware Attention (NeurIPS 2026 · similarity 0.86)
- Steady Progress Beats Stagnation: Mutual Aid of Foundation and Conventional Models in Mixed Domain Semi-Supervised Medical Image Segmentation (CVPR 2025 · similarity 0.86)
- Constructing and Exploring Intermediate Domains in Mixed Domain Semi-supervised Medical Image Segmentation (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
