Research map / Label-efficient and robust learning / Self-supervised and contrastive
Large-Scale 3D Medical Image Pre-Training With Geometric Context Priors
Linshan Wu, Jiaxin Zhuang, Hao 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 intra-volume contrast pre-training (ablation) | Swin-B · TotalSegmentator | DSC (%): 80.97 -> 81.38 | Improved |
| add inter-volume contrast on top of intra-volume contrast (ablation) | VoCo Swin-B · TotalSegmentator | DSC (%): 81.38 -> 82.07 | Improved |
| add inter-volume contrast on top of intra-volume contrast (ablation) | VoCo Swin-B · CC-CCII | accuracy (%): 92.85 -> 93.64 | Improved |
| add semi-supervised learning to self-supervised contrast (omni-supervised) (ablation) | VoCo Swin-B · TotalSegmentator | DSC (%): 82.07 -> 84.84 | Improved |
| add semi-supervised learning to self-supervised contrast (omni-supervised) (ablation) | VoCo Swin-B · CC-CCII | accuracy (%): 93.64 -> 93.80 | Improved |
| add self-supervised contrast to pure semi-supervised pre-training (ablation) | VoCo Swin-B · CC-CCII | accuracy (%): 91.98 -> 93.80 | Improved |
| add VoCo pre-training over scratch | Swin-B/nnUNet · 7 segmentation datasets | DSC (%): +3.12% | Improved |
| add VoCo pre-training over scratch | Swin-B/nnUNet · 24 downstream segmentation tasks | DSC (%): +4.42% | Improved |
| add VoCo pre-training over scratch | nnUNet/Swin-B · MSD 10-Task | DSC (%): +2.98% | Improved |
| scale pre-training data from 10K to 160K (ablation) | VoCo · TotalSegmentator | DSC (%) | No clear effect |
| use larger model (VoCo-H over smaller) on small registration dataset (ablation) | VoCo · OASIS | DSC (%) | Hurt |
| add VoCo pre-training on PET AutoPET | nnUNet · AutoPET | DSC (%): 35.84 -> 33.02 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VoCo-v1 — VoCo: A Simple-yet-Effective Volume Contrastive Learning Framework for 3D Medical Image Analysis (CVPR 2024) · better on 1 of 1 shared result
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
- VoCo: A Simple-yet-Effective Volume Contrastive Learning Framework for 3D Medical Image Analysis (CVPR 2024 · similarity 0.94)
- Autoregressive Sequence Modeling for 3D Medical Image Representation (AAAI 2025 · similarity 0.88)
- Training Like a Medical Resident: Context-Prior Learning Toward Universal Medical Image Segmentation (CVPR 2024 · similarity 0.87)
- How Well Do Supervised 3D Models Transfer to Medical Imaging Tasks? (ICLR 2024 · similarity 0.87)
- Time-to-Event Pretraining for 3D Medical Imaging (ICLR 2025 · similarity 0.86)
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.86)
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.
