Research map / Vision-language and multimodal / Multimodal LLMs
Region-Aware Multimodal Large Language Model via SlowFast Tokenization and Pseudo-Mask Guidance for 3D CT Report Generation
Sunggu Kyung, Jinyoung Seo, Hyunseok Lim, Dongyeong Kim, Hyungbin Park, Jimin Sung, Wooyoung Jo, Yoojin Nam, Namkug Kim
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 Lesion Attribute Extractor (lesion size/location text prompts) to R+Mask (ablation) | MedRegion-CT · RadGenome-Chest CT (internal, n=1551) 30% train | CRG: 0.3699 -> 0.3861 | Improved |
| add Lesion Attribute Extractor to R+Mask (ablation) | MedRegion-CT · AMC zero-shot | GPT-CA: 0.2323 -> 0.2564 | Improved |
| add Mask-Driven Visual Extractor to R+Attr (ablation) | MedRegion-CT · RadGenome-Chest CT (internal, n=1551) 30% train | GPT-CA: 0.3807 -> 0.3839 | Improved |
| add Mask-Driven Visual Extractor to R+Attr (ablation) | MedRegion-CT · AMC zero-shot | GPT-CA: 0.2530 -> 0.2564 | Improved |
| add Mask + Attr to Region-based SlowFast Tokenizer (ablation) | MedRegion-CT · RadGenome-Chest CT (internal, n=1551) 30% train | GREEN: 0.3930 -> 0.4025 | Improved |
| use Region-based SlowFast Tokenizer instead of conventional SlowFast (LITA) (ablation) | MedRegion-CT · AMC zero-shot | GPT-CA: 0.2241 -> 0.2244 | Improved |
| use Region-based SlowFast Tokenizer instead of M3D 3D ViT encoder (ablation) | MedRegion-CT · RadGenome-Chest CT (internal, n=1551) 30% train | CRG: 0.3487 -> 0.3613 | Improved |
| add Mask extractor to R+Attr on ROUGE (ablation) | MedRegion-CT · AMC zero-shot | ROUGE: 0.1628 (R+Mask) vs 0.1577 (full) | Mixed |
| distort organ masks (ablation) | MedRegion-CT · segmentation noise test | BLEU: 0.3273 -> 0.2252 | Hurt |
| distort lesion masks (ablation) | MedRegion-CT · segmentation noise test | BLEU: 0.3273 -> 0.2986 | Hurt |
| use MedRegion-CT instead of Med3DVLM | MedRegion-CT · RadGenome-Chest CT (internal, n=1551) | GREEN: 0.3881 -> 0.4555 | Improved |
| use MedRegion-CT instead of RadFM | MedRegion-CT · AMC (external zero-shot, n=500) | BLEU: 0.1464 -> 0.1275 | Hurt |
Closest papers by meaning
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.89)
- Interpretable Bilingual Multimodal Large Language Model for Diverse Biomedical Tasks (ICLR 2025 · similarity 0.88)
- MedMO: Grounding and Understanding Multimodal Large Language Model for Medical Images (CVPR 2026 · similarity 0.87)
- HC-LLM: Historical-Constrained Large Language Models for Radiology Report Generation (AAAI 2025 · similarity 0.86)
- MEPNet: Medical Entity-Balanced Prompting Network for Brain CT Report Generation (AAAI 2025 · similarity 0.86)
- Large-scale and Fine-grained Vision-language Pre-training for Enhanced CT Image Understanding (ICLR 2025 · 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.
