Jul 2026· IEEE Transactions on Medical Imaging· Vol PP· 0 citations
Medicine
TL;DR
This work designs an implicit pixel-wise learnable step size to adapt to the spatial gradient heterogeneity of CT images and develops a cross-prompt guiding mechanism to enable inter-domain prompt interaction, which facilitates efficient prompt generation and enhances the convergence stability of the model.
Abstract
Sparse-view computed tomography (CT) effectively reduces radiation exposure, yet it degrades image signal-to-noise ratio (SNR) and compromises the reliability of clinical diagnosis. Deep unrolling networks, which integrate the merits of optimization-based and data-driven paradigms, have achieved promising performance for sparse-view CT reconstruction. However, existing learned data consistency (DC) and prompt-based reconstruction methods only capture simplistic image priors and rely on elaborately designed regularizers as well as excessive unrolled iterations, leading to heavy computational overhead and over-smoothed results. In this work, we integrate prompt learning into unrolled gradient descent networks and propose a Dual-domain Cross-Prompt Learning (DCPL) framework to address these limitations. Specifically, we first design an implicit pixel-wise learnable step size to adapt to the spatial gradient heterogeneity of CT images. We then integrate learnable prompts separately into the data-fidelity and regularization terms during unrolled iterations, enabling the model to adaptively capture intrinsic CT anatomical and noise priors. Furthermore, a cross-prompt guiding mechanism is developed to enable inter-domain prompt interaction, which facilitates efficient prompt generation and enhances the convergence stability of the model. Extensive experiments on multiple clinical benchmarks under both in-domain and cross-domain settings demonstrate that our DCPL achieves consistent improvements in artifact suppression and fine structural preservation, even under extremely sparse-view sampling. Notably, our method delivers robust reconstruction quality with significantly fewer parameters, higher inference efficiency, and with only a few unrolled iterations.
Introduction Sparse-view computed tomography (CT) reconstruction is crucial for clinical diagnostics, as reducing radiation exposure is essential to minimize risks to patients. Existing dual-domain reconstruction methods leverage both image and projection domains but often process them sequentially, overlooking their implicit correlations. Methods To address this limitation, we propose Cross-Domain TransNet, a Transformer-based dual-domain framework for sparse-view CT reconstruction. The proposed model captures long-range dependencies within each domain and integrates image and sinogram representations through a hybrid self-attention mechanism. In addition, a Convolution Fusion Layer (CFL) is introduced to enhance feature interactions and facilitate more effective utilization of dual-domain information. Results Extensive experiments on the NIH-AAPM dataset demonstrate the superior performance and generalization capability of the proposed method under various sparse-view settings. The results show that Cross-Domain TransNet consistently improves reconstruction quality, effectively suppresses noise, and reduces artifacts, outperforming both conventional reconstruction algorithms and state-of-the-art deep learning approaches. Conclusion Cross-Domain TransNet provides an effective and robust solution for sparse-view CT reconstruction. By fully exploiting complementary information from both image and projection domains, the proposed framework enhances diagnostic image quality while supporting radiation dose reduction.
Junling Wang, Chunhua Zou, Hongjie Yang et al.· Frontiers in Nuclear Medicin...· 0 citations
Objective. Ultra-low-dose CT (ULDCT) can be achieved by reducing the tube current and employing sparse-view projections, thereby improving patient safety by lowering radiation exposure. However, this strategy inevitably introduces severe aliasing artifacts and increased noise, leading to substantial degradation of image quality. To simultaneously address undersampling-induced artifacts and noise contamination, we propose a dual-domain dual-branch residual-learning network (D 3R-Net) for high-fidelity ULDCT reconstruction. Approach. The proposed framework first performs edge-preserving sinogram restoration using an improved directional cubic convolution (iDCC) interpolation method, followed by a U-Net optimized with an inner-structure gradient loss to preserve critical edge-gradient information. In the image domain, a dual-branch structure-infiltrated guidance network (DB-SiGN) is designed to extract low- and high-frequency information from the refined reconstruction and the original noisy projections, respectively. The gradient features extracted from the low-frequency branch are used to guide the high-frequency branch, enabling more effective discrimination between true anatomical structures and noise/artifacts. Both branches learn residual mappings between the refined reconstruction and the corresponding normal-dose CT (NDCT) image, and their outputs are adaptively fused through spatial attention weighting to produce the final reconstruction. Main results. Experimental results on both simulated dose-reduction datasets and real CBCT data demonstrate that D 3R-Net consistently outperforms competing methods in terms of quantitative metrics and visual image quality across all evaluated scenarios. In addition, the proposed method achieves superior and more robust downstream segmentation performance, with reconstructed images exhibiting the highest consistency with NDCT references. Significance. D 3R-Net establishes a robust and interpretable dual-domain reconstruction framework for ULDCT imaging. By effectively suppressing noise and aliasing artifacts while preserving fine anatomical structures, the proposed method provides a promising solution for safe, reliable, and clinically deployable ULDCT reconstruction.
Jia-Bing Xiang, Yu-Hang Yang, Yan-Xin Wang et al.· Physics in Medicine and Biol...· 0 citations
Sparse-view computed tomography (CT) reduces radiation dose by acquiring fewer projection views, but the resulting inverse problem is highly ill-posed and often produces severe streak artifacts. Existing deep reconstruction methods have achieved promising performance, yet many rely on first-order updates or large regularization networks, which can be less effective in ill-conditioned settings. We propose \textbf{CG-GLORE}, a compact deep unrolling framework inspired by second-order optimization for sparse-view CT reconstruction. Each unrolled stage uses a CG-solved linear system based on a structured Hessian surrogate: it retains the physics-induced curvature of the data-fidelity term while using an identity approximation for the learned regularization term. Thus, the method is second-order-inspired rather than an exact Newton method for the full learned objective. To model image priors, we design a Global-Local Regularization Network (GLORE), which combines convolutional local feature extraction with a Long-Range Dependency Representation module based on sparse patchification and Nystr\"{o}m attention. This design captures anatomical details and non-local dependencies while maintaining practical complexity. Experiments on AAPM and DeepLesion under multiple sparse-view and noise settings show that CG-GLORE achieves strong quantitative performance, stable convergence, lower noise power, and improved visual fidelity compared with representative reconstruction methods.
Tran Xuan Hieu Le, D. C. Bui, V. Le et al.· 0 citations
This paper proposes K-NeAS, a unified and scalable architecture for automated, multi-material surface reconstruction that replaces independent material networks with a shared latent backbone and introduces a fully differentiable $K$-material sequential soft selector to model an arbitrary number of overlapping tissues.
Daksh K. Shah, Emmanouil Nikolakakis, Razvan V. Marinescu· 0 citations
ObjectiveSparse View Computed Tomography (SVCT) is an effective way to reduce radiation dose. However, missing information of projection leads to noise and artifacts. Current SVCT reconstruction networks cannot effectively utilize the correlation information of images from different energies, and linear interpolation in the projection domain brings more secondary artifacts. A dual-domain approach involving data complementation and correction is proposed as a potential solution in this work.ApproachWe introduce the MHSG-Model for sparse-view spectral CT, designed for integrated processing in both the projection and image domains. This model consists of a Projection Domain Generative Model (PDGM) and an Image Domain Correction Module (IDCM). The PDGM employs a generative model integrated with a selective State Space Model (SSM) to recover missing projection data, reducing secondary artifacts and computational complexity. The IDCM utilizes a novel cross-attention mechanism, incorporating a multi-scale attention for high-energy images and a high-frequency attention for low-energy images, to further enhance reconstruction quality.Main resultsThe MHSG-Model was launched and validated on an abdominal slice simulation dataset and a sparse view spectral CT dataset of the AAPM DL Challenge. The proposed network achieved superior performance metrics, with an RMSE of 0.0111, MAE of 15.91, SSIM of 0.9674, and PSNR of 39.73 dB. The experimental results demonstrate that MHSG-Model has a good performance on preserving detail information and removing artifacts, which show the potential to be applied in clinical sparse view spectral CT reconstruction.
Qi-Wei Li, Zaifeng Shi, Fanning Kong et al.· Journal of X-Ray Science and...· 0 citations
A model-driven bilevel optimization framework that couples SENSE-based image reconstruction with SPIRiT-based k-space calibration through shared CSMs, and introduces a deep-prior-guided regularization strategy that preserves the structure of classical linear regularizers while adaptively learning spatially varying regularization weights from denoised intermediate reconstructions.
Weipeng Chen, Yan-Ran Li, Raymond H. Chan et al.· Journal of Mathematical Imag...· 0 citations