LUCID: Learned Undersampling-Adaptive Consistency-Guided Inference with Deterministic Flow Matching for Sparse-View CT Reconstruction
Learn to reconstruct sparse-view CT images using LUCID, a novel method that combines undersampling-adaptive consistency-guided inference with deterministic flow matching, to reduce streak artifacts and improve image quality
- Implement LUCID using deep learning frameworks such as PyTorch or TensorFlow to reconstruct sparse-view CT images
- Apply undersampling-adaptive consistency-guided inference to reduce streak artifacts and improve image quality
- Use deterministic flow matching to refine the reconstruction and reduce anatomically inconsistent hallucination-like structures
- Evaluate the performance of LUCID using metrics such as peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM)
- Compare the results of LUCID with existing supervised and generative methods for sparse-view CT reconstruction
Researchers and engineers working on medical imaging and computer vision can benefit from this method to improve the quality of reconstructed CT images, especially in situations where radiation dose and scanning time need to be minimized
💡 LUCID combines undersampling-adaptive consistency-guided inference with deterministic flow matching to improve the quality of reconstructed CT images
📸 Improve sparse-view CT reconstruction with LUCID, a novel method that reduces streak artifacts and improves image quality 📊
Key Takeaways
Learn to reconstruct sparse-view CT images using LUCID, a novel method that combines undersampling-adaptive consistency-guided inference with deterministic flow matching, to reduce streak artifacts and improve image quality
Full Article
Abstract:
arXiv:2606.16212v1 Announce Type: cross Abstract: Sparse-view CT reduces radiation dose and scanning time by acquiring fewer projection views, but angular undersampling makes reconstruction severely ill-posed, causing streak artifacts, structural blurring, and loss of fine details. Existing supervised methods are often tied to specific sampling settings, whereas generative methods may introduce anatomically inconsistent hallucination-like structures under severe undersampling. We propose Lucid,
DeepCamp AI