Research map / Detection and segmentation / Weakly supervised detection and segmentation
Label-Efficient Hybrid-Supervised Learning for Medical Image Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 DII to weight weakly annotated instances individually (ablation) | Hybrid-supervised segmentation · CVC-EndoSceneStill | Dice: 69.77 -> 72.52 | Helped |
| Add DCR without DII (ablation) | Hybrid-supervised segmentation · CVC-EndoSceneStill | Dice: 69.77 -> 76.32 | Helped |
| Combine DII and DCR (ablation) | Hybrid-supervised segmentation · CVC-EndoSceneStill | Dice: 69.77 -> 81.15 | Helped |
| Add DII sampling tutorial to DII + DCR (ablation) | Hybrid-supervised segmentation · CVC-EndoSceneStill | Dice: 81.15 -> 82.56 | Helped |
| Replace foreground paste with CutMix (ablation) | DCR · CVC-EndoSceneStill | Dice: 82.56 -> 81.37 | Hurt |
| Replace pseudo regularization with KL divergence (ablation) | DCR · CVC-EndoSceneStill | Dice: 82.56 -> 76.83 | Hurt |
| Increase the strongly annotated subset from 28 to 55 samples (ablation) | Full method · CVC-EndoSceneStill | Dice: 80.36 -> 82.56 | Helped |
| Increase the strongly annotated subset from 55 to 110 samples (ablation) | Full method · CVC-EndoSceneStill | Dice: 82.56 -> 82.97 | Helped |
| Increase the strongly annotated subset from 110 to 220 samples (ablation) | Full method · CVC-EndoSceneStill | Dice: 82.97 -> 82.99 | Helped |
| Use the proposed hybrid-supervised method instead of the strongest hybrid baseline | Hybrid-supervised segmentation · CVC-EndoSceneStill | Dice: 76.41 -> 82.56 | Helped |
| Use the proposed hybrid-supervised method instead of the strongest hybrid baseline | Hybrid-supervised segmentation · AS-OCT Mean | Dice: 92.26 -> 93.10 | Helped |
Closest papers by meaning
- Every Annotation Counts: Multi-Label Deep Supervision for Medical Image Segmentation (CVPR 2021 · similarity 0.90)
- Medical Image Segmentation with Minimal Labeling Effort: How Far Can We Push the Limits? (AAAI 2026 · similarity 0.90)
- Rethinking Box Supervision: Bias-Free Weakly Supervised Medical Segmentation (CVPR 2026 · similarity 0.90)
- WISH: Weakly Supervised Instance Segmentation using Heterogeneous Labels (CVPR 2025 · similarity 0.88)
- SGTC: Semantic-Guided Triplet Co-training for Sparsely Annotated Semi-Supervised Medical Image Segmentation (AAAI 2025 · similarity 0.88)
- Annotation Ambiguity Aware Semi-Supervised Medical Image Segmentation (CVPR 2025 · similarity 0.87)
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.