Research map / Vision-language and multimodal / Human and embodied modeling
Versatile Vision-Language Model for 3D Computed Tomography
Jiayu Lei, Ziqing Fan, Yanyong Zhang, Weidi Xie, Ya Zhang, Yanfeng Wang
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 voxel-text (segmentation) alignment to image-text contrastive pretraining (ablation) | hybrid ResNet-ViT encoder · segmentation, diagnosis, retrieval | average Dice / AUC: Average Dice +8.57%, AUC +2.09%, Average IR@10 +2.75% | Improved |
| add category description alignment to ITC (ablation) | hybrid ResNet-ViT encoder · Rad-ChestCT / CT-ORG | AUC / DSC: AUC 54.73 -> 55.07; CT-ORG DSC 85.56 -> 90.78 | Improved |
| combine all three pretraining objectives (ITC, Seg., Clas.) (ablation) | hybrid ResNet-ViT encoder · Rad-ChestCT / CT-ORG | AUC / DSC: AUC 54.73 -> 57.44; CT-ORG DSC 85.56 -> 93.38 | Improved |
| use hybrid ResNet-ViT instead of ResNet3D (ablation) | vision encoder · diagnosis and segmentation | ACC / average DSC: ACC +5.29%, average DSC +7.40% | Improved |
| use hybrid ResNet-ViT instead of ViT3D (ablation) | vision encoder · diagnosis and segmentation | ACC / average DSC: ACC +5.38%, average DSC +4.52% | Improved |
| use hybrid instead of ViT3D for CT-RATE image-text retrieval (ablation) | vision encoder · CT-RATE retrieval | IR/TR recall | Hurt |
| add 2D slice data to 3D pretraining data (ablation) | hybrid ResNet-ViT encoder · segmentation and diagnosis | DSC / AUC: CT-ORG DSC 93.38 -> 94.74; Rad-ChestCT AUC 57.44 -> 57.22 | Mixed |
| multi-granular pretraining plus unified instruction tuning versus CT-CLIP | CTInstruct · Rad-ChestCT diagnosis | AUC (%): 52.17 -> 59.42 | Improved |
| CTInstruct versus M3D on RP3D captioning | CTInstruct · RP3D-Cap | ROUGE-1 / BLEU-1 / METEOR: ROUGE-1 18.65 vs 19.51 | No clear effect |
Closest papers by meaning
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.91)
- Large-scale and Fine-grained Vision-language Pre-training for Enhanced CT Image Understanding (ICLR 2025 · similarity 0.88)
- Enhancing vision-language models for medical imaging: bridging the 3D gap with innovative slice selection (NeurIPS 2024 · similarity 0.87)
- 3DMedAgent: Unified Perception-to-Understanding for 3D Medical Analysis (ICML 2026 · similarity 0.87)
- Toward a Vision-Language Foundation Model for Medical Data: Multimodal Dataset and Benchmarks for Vietnamese PET/CT Report Generation (NeurIPS 2025 · similarity 0.86)
- Knowledge Transfer Scaling Laws for 3D Medical Imaging (NeurIPS 2026 · 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.
