Research map / Vision-language and multimodal / Multimodal LLMs
Seeing the Trees for the Forest: Rethinking Weakly-Supervised Medical Visual Grounding
Ta Duc Huy, Duy Anh Huynh, Yutong Xie, Yuankai Qi, Qi Chen, Phi Le Nguyen, Sen Kim Tran, Son Lam Phung, Anton van den Hengel, Zhibin Liao, Minh-Son To, Johan W. Verjans and 1 more
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 disease-aware prompting to global contrastive loss (ablation) | Global loss model · MS-CXR | Dice: 0.279 -> 0.346 | Improved |
| add local contrastive loss on top of DAP + global loss (ablation) | DAP · MS-CXR | CNR: 1.061 -> 1.254 | Improved |
| add disease-aware prompting to global loss compared with local loss alone (ablation) | DAP · MS-CXR | CNR: 1.061 vs 1.064 | No clear effect |
| prompt in feature space instead of pixel masking (ablation) | DAP · MS-CXR | CNR: 0.949 -> 1.254 | Improved |
| informative disease-aware prompt instead of learnable CoCoOp prompts (ablation) | DAP · MS-CXR | CNR: 1.042 -> 1.254 | Improved |
| informative disease-aware prompt instead of learnable visual prompt (ablation) | DAP · MS-CXR | CNR: 1.129 -> 1.254 | Improved |
| use Chefer et al. explainability instead of GradCAM (ablation) | DAP · RSNA Pneumonia | CNR: 1.462 -> 1.630 | Improved |
| use Chefer et al. explainability instead of LRP (ablation) | DAP · RSNA Pneumonia | CNR: 0.973 -> 1.630 | Improved |
| increase batch size from 128 to 512 (ablation) | DAP · MS-CXR (20% subset) | CNR: 1.027 -> 1.042 | Improved |
| lower learning rate from 1e-1 to 1e-3 (ablation) | DAP · MS-CXR (20% subset) | CNR: 0.917 -> 1.042 | Improved |
| raise local contrastive loss weight from 0.1 to 1 (ablation) | DAP · MS-CXR (20% subset) | Dice: 0.350 -> 0.343 | Hurt |
| prompt only the last layer instead of the first layer (ablation) | DAP · MS-CXR (20% subset) | Dice: 0.341 -> 0.350 | Improved |
Closest papers by meaning
- How Do Medical MLLMs Fail? A Study on Visual Grounding in Medical Images (ICLR 2026 · similarity 0.92)
- MedSG-Bench: A Benchmark for Medical Image Sequences Grounding (NeurIPS 2025 · similarity 0.91)
- MedReasoner: Reinforcement Learning Drives Reasoning Grounding from Clinical Thought to Pixel-Level Precision (AAAI 2026 · similarity 0.89)
- PathFLIP: Fine-grained Language-Image Pretraining for Versatile Computational Pathology (AAAI 2026 · similarity 0.88)
- Decomposing Disease Descriptions for Enhanced Pathology Detection: A Multi-Aspect Vision-Language Pre-training Framework (CVPR 2024 · similarity 0.88)
- RadZero: Similarity-Based Cross-Attention for Explainable Vision-Language Alignment in Chest X-ray with Zero-Shot Multi-Task Capability (NeurIPS 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.
