Harnessing Magnitude-Only and Complex Measurements for Improved Dynamic MRI Reconstruction with Learned Priors

UMPIRE-Net: Unrolled Magnitude-Phase Regularization Network for Accelerated MRI

MRI reconstruction from undersampled k-space measurements is an ill-posed inverse problem. Physics-driven deep learning (PD-DL) methods have shown strong performance for this task by combining the MRI forward model with learned image regularization within algorithm-unrolling frameworks. However, most existing PD-DL methods reconstruct complex-valued images directly, thereby implicitly coupling magnitude and phase within a single learned representation. This coupled regularization may be suboptimal in reconstruction settings where accurate phase modeling plays an important role, such as partial Fourier (PF) imaging, where recovery of the omitted asymmetric k-space measurements depends on the underlying image phase. In such scenarios, explicit modeling of magnitude and phase as separate components may reduce the reliance on externally estimated or predefined phase information. To this end, we propose UMPIRE-Net (Unrolled Magnitude-Phase In REgularization Network), a PD-DL method that introduces separate learned regularizers for magnitude and phase components, together with a novel data-fidelity formulation that enforces measurements consistency. We evaluate UMPIRE-Net for accelerated MRI with PF across different datasets and acceleration factors. Experimental results demonstrate that our proposed method improves reconstruction quality compared with a conventional complex-valued PD-DL baseline, yielding sharper images and reduced artifacts. Code available at: https://github.com/MahdiSaberii/UMPIRE-Net

A Bilevel Sensitivity-Corrected Reconstruction Framework with Deep Priors for Parallel MRI

Enhancing Implicit Neural Representations with Image Feature Embedding for Unsupervised Cardiac Cine MRI Reconstruction

Cardiac cine Magnetic Resonance Imaging (MRI) is a critical diagnostic tool that provides dynamic insights for radiologists. To accelerate acquisition, under-sampled k-space data is often used, requiring reconstruction methods that combine coil sensitivity encoding with prior information to recover missing data. Deep learning approaches have gained more attention for leveraging data-adaptive priors. While supervised learning approaches are a common choice, they depend on fully sampled reference data, which is not always available. Unsupervised methods eliminate the need for fully sampled reference data, which can be advantageous in cardiac cine MRI reconstruction. Among them, implicit neural representations (INRs) have shown great potential due to their simple architecture and good quality reconstructions. In this work, we propose an image-domain dual-branch INR framework, termed I-FP-INR, which extends the original INR design by introducing an additional feature-processing branch. This design aims to extract complementary feature embeddings to enhance the overall representation, thereby benefiting reconstruction. Extensive evaluations on both public datasets and in-house data show consistent improvements over baseline methods in reconstruction quality, with strong robustness across varied scenarios.

High-dimensional Embedding Prior for Noisy K-space Domain MRIReconstruction

Magnetic resonance imaging (MRI) reconstruction under realistic acquisition conditions can be fundamentally viewed as estimating the underlying k-space distribution from incomplete and noise-corrupted measurements. While diffusion models have recently shown strong potential as generative prior for inverse problems,existingapproachesstruggletohandlenoisyreconstruction settings, especially when operating directly in k-space domain. In this work, we propose a unified high-dimensional k-space reconstruction framework tailored for noisy inverse problems, whichenhancesdiffusion-based solversthroughrepresentation lifting.Ratherthanmodifyingthe underlying optimization procedures, the proposed framework augments the data representation space, enabling existing diffusion-based solvers to operate on enriched k-space embeddings with improved expressiveness. Extensive experiments on both in-house and public datasets across varying noise levels and undersampled factors demonstrate that the proposed frame work consistently improves reconstruction quality for multiple diffusion-based inverse solvers. Notably, the largest gains are observed in high-noise regimes, which is consistent with our theoretical analysis of error propagation under high-dimensional representation. These results suggest that high-dimensional representation provides a general and model-agnostic mechanism for improving diffusion-based MRI reconstruction in noisy settings, offering a new perspective on robust k-space generative modeling for practical inverse problems. The code will be available at https://github.com/yqx7150/HEP-MRIRec.

Dual-Domain Cross-Prompt Learning for Efficient Sparse-View CT.

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.

Learned spherical ADMM for Rician MRI denoising

Rician noise corruption in magnetic resonance imaging poses a challenging restoration problem due to its nonlinearity and signal dependence. In this paper, we propose a novel learned spherical alternating direction method of multipliers (LSADMMs) for effective Rician noise removal under spherical constraints. Grounded in a sphere-constrained variational formulation, the proposed architecture unfolds the iterations of a proximal linearized alternating direction method of multiplier solver into a deep neural network. Structurally, LSADMM alternates between lightweight, learnable gradient descent modules and fixed, physics-based operators. To enable blind denoising, we incorporate a noise-level estimation prefix that provides adaptive guidance across noise levels. Notably, a parameter-free spherical projection layer is incorporated to strictly enforce the geometric constraint of the Rician degradation model, ensuring that intermediate iterates remain physically consistent. Theoretically, we establish the boundedness of the unfolded iterates and prove the layer-wise stability of the unrolled dynamics. Extensive numerical experiments on synthetic and real-world datasets demonstrate that LSADMM achieves competitive restoration performance with a lightweight architecture, requiring substantially fewer parameters than conventional end-to-end deep learning methods.