Research map / Detection and segmentation / CNN features
Unsupervised Local Discrimination for Medical Images
Huai Chen, Renzhen Wang, Xiuying Wang, Jieyu Li, Qu Fang, Hui Li, Jianhao Bai, Qing Peng, Deyu Meng, Lisheng 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 |
|---|---|---|---|
| add hypersphere mixup to patch discrimination (ablation) | Pd · Color fundus segmentation | DSC (%): 84.19 -> 84.65 | Improved |
| add hypersphere mixup to patch discrimination (ablation) | Pd · Chest X-ray segmentation | DSC (%): 79.98 -> 81.81 | Improved |
| add region discrimination to Pd+mix (ablation) | Pd+mix · Color fundus segmentation | DSC (%): 84.65 -> 85.12 | Improved |
| add region discrimination to Pd+mix (ablation) | Pd+mix · Chest X-ray segmentation | DSC (%): 81.81 -> 82.11 | Improved |
| replace IPMI21 center-distance loss with region discrimination loss | IPMI21 · Chest X-ray pneumothorax | DSC (%): 41.03 -> 43.20 | Improved |
| replace IPMI21 with region discrimination | IPMI21 · Color fundus hard exudates | DSC (%): 65.89 -> 67.05 | Improved |
| pre-train with Rd instead of training from scratch | tiny VGG16 U-Net · Chest X-ray pneumothorax | DSC (%): 34.45 -> 43.20 | Improved |
| use region discrimination instead of supervised pre-training | supervised pre-trained encoder · Color fundus segmentation | DSC (%): 82.64 -> 85.12 | Improved |
| increase cluster number from 4 to 16 (ablation) | Rd · Color fundus segmentation | DSC (%): 84.51 -> 85.12 | Improved |
| increase cluster number from 4 to 16 (ablation) | Rd · Chest X-ray segmentation | DSC (%): 82.33 -> 82.11 | Hurt |
| add region discrimination with cluster number C=4 versus C=0 (ablation) | Pd+mix · Chest X-ray pneumothorax | DSC (%): 42.36 -> 44.34 | Improved |
| increase patch-out size from (8,8) to (16,16) (ablation) | Pd+mix · Color fundus segmentation | DSC (%) | Hurt |
Closest papers by meaning
- Contrastive learning of global and local features for medical image segmentation with limited annotations (NeurIPS 2020 · similarity 0.88)
- Keypoint-Augmented Self-Supervised Learning for Medical Image Segmentation with Limited Annotation (NeurIPS 2023 · similarity 0.87)
- Mine yOur owN Anatomy: Revisiting Medical Image Segmentation With Extremely Limited Labels (TPAMI 2024 · similarity 0.87)
- Neighbor Does Matter: Density-Aware Contrastive Learning for Medical Semi-supervised Segmentation (AAAI 2025 · similarity 0.86)
- ConDSeg: A General Medical Image Segmentation Framework via Contrast-Driven Feature Enhancement (AAAI 2025 · similarity 0.85)
- DyCON: Dynamic Uncertainty-aware Consistency and Contrastive Learning for Semi-supervised Medical Image Segmentation (CVPR 2025 · 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.
