Research map / Image generation / Text-to-image diffusion
GenerateCT: Text-Conditional Generation of 3D Chest CT Volumes
Ibrahim Ethem Hamamci, Sezgin Er, Anjany Sekuboyina, Enis Simsar, Alperen Tezcan, Ayse Gulnihan Simsek, Sevval Nil Esirgun, Furkan Almas, Irem Dogan, Muhammed Furkan Dasdelen, Chinmay Prabhakar, Hadrien Reynaud and 4 more
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 |
|---|---|---|---|
| use a 3D generation architecture instead of slice-wise 2D generation | GenerateCT (3SCM) · CT-RATE test | FVD (I3D): 3557.7 -> 1092.3 | Improved |
| replace Phenaki text-to-video baseline with the cascaded CT-ViT plus diffusion pipeline | GenerateCT · CT-RATE test | FID: 104.3 -> 55.8 | Improved |
| add one more super-resolution diffusion step (2SCM to 3SCM) (ablation) | GenerateCT · CT-RATE test | FVD (I3D): 1661.4 -> 1092.3 | Improved |
| add a further super-resolution step (3SCM to 4SCM) (ablation) | GenerateCT · CT-RATE test | FID: 55.8 -> 71.3 | Hurt |
| use CT-ViT reconstructed volumes rather than original downsampled volumes as diffusion input | GenerateCT · CT-RATE test | image quality | Improved |
| train classifier on real plus equal number of GenerateCT synthetic volumes | CT-Net multi-abnormality classifier · CT-RATE test, 18 abnormalities | mean AP: 11% improvement in mean AP (baseline 0.254) | Improved |
| train classifier on real plus equal number of GenerateCT synthetic volumes | CT-Net multi-abnormality classifier · CT-RATE test, 18 abnormalities | mean AUROC: 6% increase (baseline 0.631) | Improved |
| train classifier only on 100,000 synthetic volumes instead of 20,000 real volumes | CT-Net multi-abnormality classifier · CT-RATE test, 18 abnormalities | mean AP: 8% increase in mean AP; 4% increase in mean AUROC | Improved |
| train only on synthetic volumes from unseen prompts instead of real RadChestCT data | CT-Net multi-abnormality classifier · RadChestCT | mean AUROC: 0.613 -> 0.536 | Hurt |
| add real RadChestCT data to synthetic unseen-prompt data for classifier training | CT-Net multi-abnormality classifier · RadChestCT | mean AP: 0.177 -> 0.190 | Improved |
Closest papers by meaning
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.88)
- Trade-Offs in Fine-Tuned Diffusion Models between Accuracy and Interpretability (AAAI 2024 · similarity 0.86)
- Tokenizer-Generator Coupling in Medical Image Generation (NeurIPS 2026 · similarity 0.85)
- Autoregressive Sequence Modeling for 3D Medical Image Representation (AAAI 2025 · similarity 0.85)
- Large-scale and Fine-grained Vision-language Pre-training for Enhanced CT Image Understanding (ICLR 2025 · similarity 0.85)
- Bootstrapping Chest CT Image Understanding by Distilling Knowledge from X-ray Expert Models (CVPR 2024 · 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.
