Researchers have developed a new method for generating 3D CT scans from radiology reports, addressing limitations in existing text-to-CT approaches. The proposed technique utilizes a generation-oriented 3D-CLIP encoder trained with structured hard negatives, which enhances contrastive difficulty without increasing memory costs. This encoder conditions a latent diffusion model operating directly in 3D latent space, avoiding artifacts from super-resolution pipelines and improving semantic controllability. The method achieves state-of-the-art performance on the CT-RATE benchmark for image fidelity and factual correctness, while also being more efficient in terms of inference time and GPU memory. AI
IMPACT Improves the accuracy and efficiency of generating 3D medical imaging from text, potentially aiding radiologists and researchers.
RANK_REASON The cluster contains an academic paper detailing a new method for text-to-3D CT generation. [lever_c_demoted from research: ic=1 ai=1.0]
- 3D-CLIP
- alphaXiv
- arXiv
- CatalyzeX
- CORE Recommender
- CT-RATE
- DagsHub
- Daniele Molino
- Gotit.pub
- Hugging Face
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
