Research map / Detection and segmentation / Open-set and few-shot detection
Delving Aleatoric Uncertainty in Medical Image Segmentation via Vision Foundation Models
Ruiyang Li, Fang Liu, Licheng Jiao, Xinglin Xie, Jiayao Hao, Shuo Li, Xu Liu, Jingyi Yang, Lingling Li, Puhua Chen, Wenping Ma
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 |
|---|---|---|---|
| filter 10% highest-AUV samples using MedSAM2 features | nnU-Net · LiTS | Dice (%): 82.96 -> 85.45 | Improved |
| use MedSAM2 instead of task-specific nnU-Net as feature extractor for AUV (ablation) | AUV data filtering · Five datasets mean | Mean Dice (%): 75.17 -> 76.45 | Improved |
| use CLIP-Driven as feature extractor (ablation) | AUV data filtering · Five datasets mean | Mean Dice (%): 75.03 | No clear effect |
| filter using data variance instead of foundation-model AUV (ablation) | nnU-Net · Five datasets mean | Mean Dice (%): 75.10 -> 75.39 | No clear effect |
| add DUO to nnU-Net | nnU-Net · Five datasets mean | Mean Dice (%): 75.10 -> 75.69 | Improved |
| add DUO to Swin-UNETR | Swin-UNETR · Five datasets mean | Mean Dice (%): 73.09 -> 73.64 | Improved |
| add DUO to U-Mamba | U-Mamba · Five datasets mean | Mean Dice (%): 74.13 -> 74.66 | Improved |
| add re-weighting (RW) by semantic perception scale (ablation) | DUO · LiTS tumor | Tumor Dice (%): 69.87 -> 71.66 | Improved |
| add estimated noise (EN) label denoising (ablation) | DUO · LiTS tumor | Tumor Dice (%): 71.37 -> 71.66 | Improved |
| use singular value energy S instead of Fisher/MD/EAOA (ablation) | AUV data filtering · LiTS tumor, 90% data | Tumor Dice (%): 70.41 -> 74.28 | Improved |
| normalize input with foreground mask (strategy c) instead of raw image (a) (ablation) | AUV data filtering · LiTS tumor, 90% data | Tumor Dice (%): 73.35 -> 74.28 | Improved |
| filter 5% instead of 10% of samples (ablation) | AUV data filtering · LiTS tumor | Tumor Dice (%): 74.28 -> 70.03 | Hurt |
Closest papers by meaning
- DyCON: Dynamic Uncertainty-aware Consistency and Contrastive Learning for Semi-supervised Medical Image Segmentation (CVPR 2025 · similarity 0.88)
- Multi-view Evidential Learning-based Medical Image Segmentation (AAAI 2025 · similarity 0.88)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.87)
- ConDSeg: A General Medical Image Segmentation Framework via Contrast-Driven Feature Enhancement (AAAI 2025 · similarity 0.87)
- Training Like a Medical Resident: Context-Prior Learning Toward Universal Medical Image Segmentation (CVPR 2024 · similarity 0.86)
- ValUES: A Framework for Systematic Validation of Uncertainty Estimation in Semantic Segmentation (ICLR 2024 · 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.
