Research map / Vision-language and multimodal / Human and embodied modeling
BiOTPrompt: Bidirectional Optimal Transport Guided Prompting for Disease Evolution-aware Radiology Report Generation
Tengfei Liu, Yijian Fan, Boyue Wang, Yongli Hu, Mingjie Li, Jinghua Li, Junbin Gao, Xiaojun Chang, Zhihui Li, Baocai Yin
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 bidirectional OT (vs unidirectional alternative) (ablation) | BiOTPrompt · Longitudinal-MIMIC test | CE F1: 0.388 -> 0.417 | Improved |
| add Dual-Flow Asymmetry Analyzer (ablation) | BiOTPrompt · Longitudinal-MIMIC test | CE F1: 0.382 -> 0.417 | Improved |
| add vision-language consistency constraint (ablation) | BiOTPrompt · Longitudinal-MIMIC test | CE F1: 0.376 -> 0.417 | Improved |
| add classification losses (Lvis-cls + Ltext-cls) (ablation) | BiOTPrompt · Longitudinal-MIMIC test | CE F1: 0.384 -> 0.417 | Improved |
| add KL divergence loss (ablation) | BiOTPrompt · Longitudinal-MIMIC test | CE F1: 0.399 -> 0.417 | Improved |
| add KL divergence loss (ablation) | BiOTPrompt · Longitudinal-MIMIC test | ROUGE-L: 0.284 -> 0.285 | Improved |
| use fixed threshold instead of Top-K (ablation) | asymmetric region selection · Longitudinal-MIMIC test | CIDEr: 0 230 -> 0.236 | Improved |
| use soft sigmoid threshold instead of fixed threshold (ablation) | asymmetric region selection · Longitudinal-MIMIC test | CIDEr: 0.236 -> 0.203 | Hurt |
| use learnable scalar threshold instead of fixed threshold (ablation) | asymmetric region selection · Longitudinal-MIMIC test | CIDEr: 0.236 -> 0.228 | Hurt |
| vary asymmetry threshold delta from 0.04 to 0.09 (ablation) | Dual-Flow Asymmetry Analyzer · Longitudinal-MIMIC test | F1: F1 peaks around 0.05 (0.417); ROUGE-L 0.273 to 0.285 | Mixed |
| use prior images via BiOT prompts instead of R2GenGPT concatenation | R2GenGPT · Longitudinal-MIMIC test | CE F1: 0.316 -> 0.417 | Improved |
Closest papers by meaning
- HC-LLM: Historical-Constrained Large Language Models for Radiology Report Generation (AAAI 2025 · similarity 0.89)
- DART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation (CVPR 2025 · similarity 0.88)
- S2D-Align: Shallow-to-Deep Auxiliary Learning for Anatomically-Grounded Radiology Report Generation (AAAI 2026 · similarity 0.88)
- Bootstrapping Large Language Models for Radiology Report Generation (AAAI 2024 · similarity 0.87)
- LLM-RG4: Flexible and Factual Radiology Report Generation Across Diverse Input Contexts (AAAI 2025 · similarity 0.87)
- Reporting Practice Matters: The Impact of Reference Choice on Chest X-ray Report Evaluation (NeurIPS 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.
