Awesome Computed Tomography EN 在 GitHub 上加星标

深度展开与基于模型的学习

深度展开是指把经典迭代重建算法转换成神经网络:固定迭代次数,并从数据中学习其中的部分组件,比如步长、正则项或整个近端算子。每一层都调用 CT 的前向投影和反投影算子,因此网络在结构上就保证了数据一致性,同时学到比手工设计丰富得多的先验。

典型例子展开的是梯度下降(学习梯度下降)、原始-对偶算法(Learned Primal-Dual)、ISTA/FISTA、ADMM 或专家场方案(LEARN)。相关的基于模型的方法还包括即插即用(plug-and-play)和去噪正则化(RED),它们保留迭代求解器、插入训练好的去噪器,以及在正弦图与图像之间交替的双域网络。

在稀疏视角、有限角、低剂量和金属伪影问题上,展开网络是精度最高的监督方法之一,并且往往比纯粹的图像到图像网络更能适应不同的采集设置。代价是显存和训练时间,因为每一层都要调用投影算子;显存高效的展开和三维展开方案正是为此而提出的。

在可检索的目录中浏览全部 68 篇「深度展开与基于模型的学习」论文

入门推荐

如果你刚接触这个主题,可以先读这几篇;由维护者挑选。

论文

共 68 篇,按时间倒序列出最新的 30 篇。

论文发表出处年份链接
CG-GLORE: A Conjugate Gradient-Based Global-Local Regularization Network for Sparse-View CT ReconstructionBMVC2026paper
Resolution-Agnostic Neural Operators for Multi-Rate Sparse-View CTECCV2026paper · code
Dose-Aware Cold Diffusion with Physics Consistency for Generalizable Low-Dose CT ReconstructionIJCNN2026paper
Dual-domain U-Nets with embedded back projection operators for motion-resolved 4D CBCT reconstruction预印本2026paper
Efficient Unrolled Networks for Large-Scale 3D Inverse ProblemsCVPR2026paper · code
GB-SVFBP: Gaussian-Based Shift-Variant FBP neural network预印本2026paper
HARU-Net: Hybrid Attention Residual U-Net for Edge-Preserving Denoising in Cone-Beam Computed TomographyBiomed. Phys. Eng. Express2026paper
Learning Where and What to Lift for Bi-planar X-ray-to-CT Reconstruction预印本2026paper
LUCID: Learned Undersampling-Adaptive Consistency-Guided Inference with Deterministic Flow Matching for Sparse-View CT Reconstruction预印本2026paper
Reconstruct Anything Model: A Lightweight General Model for Computational ImagingICLR2026paper · code
Toward a Foundation Plug-and-Play Prior for Computed Tomography Reconstruction via a Multimodal Diffusion ModelSC Workshops2026paper
A Learnt Half-Quadratic Splitting-Based Algorithm for Fast and High-Quality Industrial Cone-beam CT Reconstruction预印本2025paper
Learning Cocoercive Conservative Denoisers via Helmholtz Decomposition for Poisson Inverse ProblemsNeurIPS2025paper · code
Data-Efficient Limited-Angle CT Using Deep Priors and RegularizationLNCS2025paper
EAGLE: An Edge-Aware Gradient Localization Enhanced Loss for CT Image ReconstructionJ. Med. Imaging2025paper · code
End-to-End Deep Learning for Interior Tomography with Low-Dose X-ray CTPMB2025paper
Equivariant Multiscale Learned Invertible Reconstruction for Cone Beam CT: From Simulated to Real Data预印本2025paper
FIND-Net -- Fourier-Integrated Network with Dictionary Kernels for Metal Artifact ReductionMICCAI2025paper · code
MInDI-3D: Iterative Deep Learning in 3D for Sparse-view Cone Beam Computed TomographyIEEE Access2025paper
Semantic contrastive learning for orthogonal X-ray computed tomography reconstructionFully3D2025paper
Plug-and-Play Learned Proximal Trajectory for 3D Sparse-View X-Ray Computed TomographyECCV2024paper
SOUL-Net: A Sparse and Low-Rank Unrolling Network for Spectral CT Image ReconstructionTNNLS2024paper · code
Deep Embedding-Attention-Refinement for Sparse-View CT ReconstructionTIM2023paper
DREAM-Net: Deep Residual Error Iterative Minimization Network for Sparse-View CT ReconstructionJBHI2023paper
FreeSeed: Frequency-Band-Aware and Self-Guided Network for Sparse-View CT ReconstructionMICCAI2023paper · code
LEARN++: Recurrent Dual-Domain Reconstruction Network for Compressed Sensing CTTRPMS2023paper
SPIE-DIR: Self-Prior Information Enhanced Deep Iterative Reconstruction Using Two Complementary Limited-Angle Scans for DECTTIM2023paper
Adaptive Convolutional Dictionary Network for CT Metal Artifact ReductionIJCAI2022paper · code
Stabilizing deep tomographic reconstruction - Part A. Hybrid framework and experimental resultsPatterns2022paper
Stabilizing deep tomographic reconstruction - Part B. Convergence analysis and adversarial attacksPatterns2022paper

在可检索的目录中浏览全部 68 篇「深度展开与基于模型的学习」论文

数据集、基准与工具库

目录中任务或标签与本主题相符的条目。

指南

涵盖本主题的指南。

其他主题