Research map / Label-efficient and robust learning / Self-supervised and contrastive
UniMiSS: Universal Medical Self-Supervised Learning via Breaking Dimensionality Barrier
Yutong Xie, Jianpeng Zhang, Yong Xia, Qi Wu
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 volume-slice consistency to 3D pre-training (ablation) | UniMiSS · BCV val (100% labels) | Dice: 80.94 -> 82.36 | Improved |
| add volume-slice consistency to 3D pre-training (ablation) | UniMiSS · RICORD val (100% labels) | accuracy: 80.96 -> 82.43 | Improved |
| add volume-slice consistency to 3D pre-training (ablation) | UniMiSS · BCV val (20% labels) | Dice: 72.08 -> 74.56 | Improved |
| add volume-slice consistency to 3D pre-training (ablation) | UniMiSS · RICORD val (20% labels) | accuracy: 69.87 -> 72.46 | Improved |
| set iteration interval to 2 instead of 1 (ablation) | UniMiSS · BCV val | Dice: 82.70 -> 83.56 | Improved |
| set iteration interval to 2 instead of 3 (ablation) | UniMiSS · BCV val | Dice: 83.28 -> 83.56 | Improved |
| set iteration interval to 2 instead of 1 (ablation) | UniMiSS · ChestXR | accuracy: 96.65 -> 97.57 | Improved |
| add switchable patch embedding instead of interpolation plus linear embedding (ablation) | MiT with UniMiSS pre-training · BCV offline test | Dice: 76.65 -> 84.99 | Improved |
| add switchable patch embedding instead of interpolation plus linear embedding (ablation) | MiT random init · BCV offline test | Dice: 73.31 -> 79.93 | Improved |
| add 2D X-rays to 3D CT pre-training (UniMiSS vs DINO on 3D only) | Transformer SSL · BCV (100% labels) | Dice: 82.61 -> 84.99 | Improved |
| add 2D X-rays to 3D CT pre-training (UniMiSS vs DINO on 3D only) | Transformer SSL · RICORD (100% labels) | accuracy | Improved |
| use 2D-3D cross-dimension pre-training instead of 2D-only DINO | Transformer SSL · JSRT (100% labels) | Dice: 93.03 -> 94.08 | Improved |
Later papers that compared with this method
2 later accepted papers report UniMiSS as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- AnatoMask: Enhancing Medical Image Segmentation with Reconstruction-guided Self-masking (ECCV 2024 · better on 1 of 1 shared result)
- Anatomical Invariance Modeling and Semantic Alignment for Self-supervised Learning in 3D Medical Image Analysis (ICCV 2023 · 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
- UniMiSS+: Universal Medical Self-Supervised Learning From Cross-Dimensional Unpaired Data (TPAMI 2024 · similarity 0.96)
- Revisiting MAE Pre-training for 3D Medical Image Segmentation (CVPR 2025 · similarity 0.87)
- A Unified Degradation-Robust Approach to SSL and UDA for 3D Medical Images (AAAI 2025 · similarity 0.85)
- Autoregressive Sequence Modeling for 3D Medical Image Representation (AAAI 2025 · similarity 0.85)
- MAPSeg: Unified Unsupervised Domain Adaptation for Heterogeneous Medical Image Segmentation Based on 3D Masked Autoencoding and Pseudo-Labeling (CVPR 2024 · similarity 0.84)
- Unleashing the Potential of SAM for Medical Adaptation via Hierarchical Decoding (CVPR 2024 · similarity 0.84)
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.
