Research map / Label-efficient and robust learning / Semi-supervised learning
MagicNet: Semi-Supervised Multi-Organ Segmentation via Magic-Cube Partition and Recovery
Duowen Chen, Yunhao Bai, Wei Shen, Qingli Li, Lequan Yu, Yan Wang
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 |
|---|---|---|---|
| Use MagicNet's cube partition-and-recovery and pseudo-label blending | V-Net semi-supervised segmentation · MACT 10% labeled | Avg. DSC: 72.18 -> 79.31 | Improved |
| Add cross-image partition-and-recovery (ablation) | Mean Teacher on MACT 10% labeled · MACT 10% labeled | Avg. DSC: 71.88 -> 78.13 | Improved |
| Add within-image partition-and-recovery to Cross (ablation) | MagicNet · MACT 10% labeled | Avg. DSC: 78.13 -> 78.83 | Improved |
| Add location reasoning to Cross (ablation) | MagicNet · MACT 10% labeled | Avg. DSC: 78.13 -> 78.56 | Improved |
| Add location reasoning to Cross + In (ablation) | MagicNet · MACT 10% labeled | Avg. DSC: 78.83 -> 78.87 | No clear effect |
| Add cube-wise pseudo-label blending (ablation) | Cross + In + Loc · MACT 10% labeled | Avg. DSC: 78.87 -> 79.31 | Improved |
| Keep relative cube positions rather than scrambling them (ablation) | Cross-image partition-and-recovery · MACT 10% labeled | DSC: 69.53 -> 73.84; 71.91 -> 78.13 | Improved |
| Apply cross-image augmentation to labeled and unlabeled data rather than unlabeled data only (ablation) | Cross-image partition-and-recovery · MACT 10% labeled | DSC: 73.84 -> 78.13 | Improved |
| Blend teacher predictions with cube-wise representations instead of teacher-only supervision (ablation) | Unlabeled-data pseudo-labeling · MACT 10% labeled | DSC: 78.47 -> 79.31 | Improved |
| Use mutual supervision instead of cube-wise blending (ablation) | Unlabeled-data pseudo-labeling · MACT 10% labeled | DSC: 79.31 -> 70.45 | Hurt |
| Use MagicNet augmentation instead of CutMix, CutOut, or MixUp | V-Net semi-supervised segmentation · MACT 10% labeled | DSC: 73.80 -> 79.31; 72.47 -> 79.31; 72.13 -> 79.31 | Improved |
| Use N=3 rather than N=2 cube divisions (ablation) | MagicNet · MACT 10% labeled | DSC: 78.68 -> 79.31 | Improved |
Later papers that compared with this method
2 later accepted papers report MagicNet as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Divide, Conquer, and Aggregate: Asymmetric Experts for Class-Imbalanced Semi-Supervised Medical Image Segmentation (CVPR 2026 · better on 2 of 2 shared results)
- Gradient-Aware for Class-Imbalanced Semi-supervised Medical Image Segmentation (ECCV 2024 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning From Partially Labeled Data for Multi-Organ and Tumor Segmentation (TPAMI 2023 · similarity 0.89)
- CCQ: Cross-Class Query Network for Partially Labeled Organ Segmentation (AAAI 2023 · similarity 0.89)
- DoDNet: Learning To Segment Multi-Organ and Tumors From Multiple Partially Labeled Datasets (CVPR 2021 · similarity 0.89)
- CAT: Coordinating Anatomical-Textual Prompts for Multi-Organ and Tumor Segmentation (NeurIPS 2024 · similarity 0.85)
- A Semantic Knowledge Complementarity based Decoupling Framework for Semi-supervised Class-imbalanced Medical Image Segmentation (CVPR 2025 · similarity 0.85)
- Combinatorial CNN-Transformer Learning with Manifold Constraints for Semi-supervised Medical Image Segmentation (AAAI 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.
