Research map / Vision-language and multimodal / CLIP and zero-shot
PRIOR: Prototype Representation Joint Learning from Medical Images and Reports
Pujin Cheng, Li Lin, Junyan Lyu, Yijin Huang, Wenhan Luo, Xiaoying Tang
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 local alignment module to global alignment (ablation) | PRIOR · SIIM classification 1% labeled | AUC-ROC (%): 82.01 -> 84.86 | Improved |
| add cross-modality conditional reconstruction (ablation) | PRIOR · SIIM classification 1% labeled | AUC-ROC (%): 84.86 -> 85.80 | Improved |
| add sentence prototype memory bank to LAM+CCR (ablation) | PRIOR · SIIM classification 1% labeled | AUC-ROC (%): 85.80 -> 87.27 | Improved |
| add sentence prototype memory bank to LAM without CCR (ablation) | PRIOR · SIIM classification 1% labeled | AUC-ROC (%): 84.86 -> 86.81 | Improved |
| add all components on 1% SIIM segmentation (ablation) | PRIOR · SIIM segmentation 1% labeled | Dice: 17.88 -> 20.43 | Improved |
| use PRIOR instead of MGCA on RSNA classification | classification · RSNA classification | AUC-ROC (%): 89.30 -> 89.22 (100%) | No clear effect |
| use PRIOR instead of MGCA on SIIM 10% classification | classification · SIIM classification 10% labeled | AUC-ROC (%): 89.68 -> 89.13 | Hurt |
| use PRIOR for zero-shot classification instead of GLoRIA | zero-shot · CheXpert zero-shot 5x200 | accuracy (%): 23.20 -> 34.90 | Improved |
| use PRIOR for image-to-text retrieval instead of GLoRIA | retrieval · CheXpert image-to-text retrieval 5x200 | Precision@1 (%): 29.3 -> 40.2 | Improved |
| use medical report pretraining for 1% RSNA detection | detection · RSNA detection 1% labeled | mAP(0.5:0.95): 0.20 vs below 0.1 for most methods | No clear effect |
| use report-based VLP pretraining versus SSL for RSNA detection | detection · RSNA detection | mAP(0.5:0.95) | Mixed |
| fine-tune more ResNet blocks beyond linear evaluation | evaluation protocol · CheXpert 1% | AUC-ROC | Improved |
Later papers that compared with this method
2 later accepted papers report PRIOR as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- MedUnifier: Unifying Vision-and-Language Pre-training on Medical Data with Vision Generation Task using Discrete Visual Representations (CVPR 2025 · better on 3 of 3 shared results)
- MLIP: Enhancing Medical Visual Representation with Divergence Encoder and Knowledge-guided Contrastive Learning (CVPR 2024 · better on 4 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Joint Learning of Localized Representations from Medical Images and Reports (ECCV 2022 · similarity 0.89)
- MedUnifier: Unifying Vision-and-Language Pre-training on Medical Data with Vision Generation Task using Discrete Visual Representations (CVPR 2025 · similarity 0.87)
- MLIP: Enhancing Medical Visual Representation with Divergence Encoder and Knowledge-guided Contrastive Learning (CVPR 2024 · similarity 0.87)
- Unlocking the Power of Spatial and Temporal Information in Medical Multimodal Pre-training (ICML 2024 · similarity 0.87)
- Alignment, Mining and Fusion: Representation Alignment with Hard Negative Mining and Selective Knowledge Fusion for Medical Visual Question Answering (CVPR 2025 · similarity 0.87)
- Enhancing Medical Large Vision-Language Models via Alignment Distillation (AAAI 2026 · 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.
