Research map / Vision-language and multimodal / CLIP and zero-shot
Multimodal Causality-Driven Representation Learning for Generalizable Medical Image Segmentation
Xusheng Liang, Lihua Zhou, Nianxin Li, Miao Xu, Ziyang Song, Dong Yi, Jinlin Wu, Jiawei Ma, Hongbin Liu, Zhen Lei, Jiebo Luo
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 MTRS and CDRL (full MCDRL) over CLIP baseline (ablation) | MCDRL · 5-site endoscopic DG average | mDice (%): 69.37 -> 88.46 | Improved |
| add Multimodal Target Region Selection to CDRL (ablation) | MCDRL · 5-site endoscopic DG average | mDice (%): 80.47 -> 88.46 | Improved |
| add Causal-Driven Representation Learning to MTRS (ablation) | MCDRL · 5-site endoscopic DG average | mDice (%): 78.71 -> 88.46 | Improved |
| use MCDRL instead of BiomedCoOp | CLIP ViT-L/14 · 5-site endoscopic DG average | Dice (%): 79.4 -> 81.6 | Improved |
| use larger CLIP backbone (ResNet-50 to ViT-L/14) | MCDRL · 5-site endoscopic DG average | Dice (%): 78.6 -> 81.6 | Improved |
| increase confounder dictionary size from 3 to 12 (ablation) | MCDRL on ViT-L/14 · 5-site average | Dice (%): 76.7 -> 81.6 | Improved |
| increase confounder dictionary size from 12 to 15 (ablation) | MCDRL on ViT-L/14 · 5-site average | Dice (%): 81.6 -> 81.4 | No clear effect |
| increase causal network depth from 1 to 3 layers (ablation) | MCDRL on ViT-L/14 · 5-site average | Dice (%): 78.7 -> 81.6 | Improved |
| increase causal network depth from 3 to 5 layers (ablation) | MCDRL on ViT-L/14 · 5-site average | Dice (%): 81.6 -> 82.0 | No clear effect |
| use MCDRL instead of baseline on nodule lesions | CLIP backbone · lesion types | mDice (%): 73.5 -> 84.9 | Improved |
Closest papers by meaning
- Make Foundation Models Trustworthy Again: Causal Fine-Adaptation for Medical Image Segmentation (AAAI 2026 · similarity 0.90)
- MedCLIPSeg: Probabilistic Vision-Language Adaptation for Data-Efficient and Generalizable Medical Image Segmentation (CVPR 2026 · similarity 0.87)
- Safe Generalization: Mitigating Catastrophic Forgetting in Single-Source Multi-Organ Segmentation via Collaborative Causal Learning (ECCV 2026 · similarity 0.87)
- Multi-view Evidential Learning-based Medical Image Segmentation (AAAI 2025 · similarity 0.87)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.86)
- Modality-agnostic Domain Generalizable Medical Image Segmentation by Multi-Frequency in Multi-Scale Attention (CVPR 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.
