Research map / Large language models / Retrieval and knowledge
Error Correction in Radiology Reports: A Knowledge Distillation-Based Multi-Stage Framework
Jinge Wu, Zhaolong Wu, Ruizhe Li, Tong Chen, Abul Hasan, Yunsoo Kim, Jason Pui-Yin Cheung, Teng Zhang, Honghan Wu
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 EXKR external knowledge retrieval (ablation) | Llama3-8B · MIMIC-CXR radiology error benchmark detection | accuracy (%): 37.79 -> 58.14 | Improved |
| add MKGD knowledge graph distillation alone (ablation) | Llama3-8B · MIMIC-CXR radiology error benchmark detection | accuracy (%): 37.79 -> 37.92 | No clear effect |
| add MKGD on top of EXKR (ablation) | Llama3-8B · MIMIC-CXR radiology error benchmark detection | accuracy (%): 58.14 -> 62.27 | Improved |
| add EXKR or MKGD for correction (ablation) | Llama3-8B · MIMIC-CXR radiology error benchmark correction | AggNLG: 94.29 -> 94.43 | No clear effect |
| add dual-knowledge infusion (MKGD+EXKR) to staged baseline | all six LLMs · MIMIC-CXR radiology error benchmark detection | accuracy (%): 57.43 -> 72.48 (avg) | Improved |
| add dual-knowledge infusion to staged baseline | MMedLM2 · MIMIC-CXR radiology error benchmark correction | AggNLG: 58.27 -> 53.50 | Hurt |
| add dual-knowledge infusion to staged baseline | Llama3-Aloe · MIMIC-CXR radiology error benchmark correction | AggNLG: 90.17 -> 74.77 | Hurt |
| use staged inference instead of end-to-end (ablation) | six LLMs · MIMIC-CXR radiology error benchmark correction | AggNLG: 72.39 -> 82.27 (avg) | Improved |
| use MKGD+EXKR instead of simple RAG | six LLMs · MIMIC-CXR radiology error benchmark detection | accuracy (%): 62.42 -> 72.48 (avg) | Improved |
| use MKGD+EXKR instead of simple RAG | six LLMs · MIMIC-CXR radiology error benchmark processing time | seconds: 26.64 -> 17.65 (avg) | Improved |
| add dual-knowledge infusion to staged baseline | Phi3-medium · MIMIC-CXR radiology error benchmark localization | accuracy (%): 69.73 -> 63.44 | Hurt |
Closest papers by meaning
- 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 0.86)
- CURE: Curriculum-guided Multi-task Training for Reliable Anatomy Grounded Report Generation (CVPR 2026 · similarity 0.86)
- Instance-level Expert Knowledge and Aggregate Discriminative Attention for Radiology Report Generation (CVPR 2024 · similarity 0.86)
- Bootstrapping Large Language Models for Radiology Report Generation (AAAI 2024 · similarity 0.86)
- DART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation (CVPR 2025 · similarity 0.85)
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.
