Keywords: AI/ML Image Reconstruction, Data Processing

Motivation: Deep unfolding neural networks had attained great success in solving tasks of magnetic resonance image (MRI) reconstruction.

Goal(s): However, minor perturbation in MR signals can result in significant distortions such as some artifacts of the reconstructed images via previous deep unfolding methods.

Approach: This paper proposes a deep equilibrium unfolding network based on adversarial learning to improve robustness of unfolding networks.

Results: Experiment results demonstrate that the proposed method obtains better reconstructed MR images compared with baseline-networks when some artifacts exist in under-sampled multi-channel k-space data.

Impact: We propose a robust method for MRI reconstruction against artefacts in k-space data.

Keywords: AI/ML Image Reconstruction, Data Processing

Motivation: Self-supervised learning-based MR image reconstruction learn a prior on the data distribution, but data in medical imaging settings are highly diverse, which consists of different corruptions or degradation factors. Existing methods need improvement in preserving details for undersampled image reconstruction with different degradation factors.

Goal(s): Our goal is to preserve details for undersampled image reconstruction with different degradation factors.

Approach: A detail-preserving self-supervised federated learning method is proposed to preserve details by employing personalized federated models to refine undersampled training data iteratively.

Results: Experiments show that promising results are achieved by proposed method, and details are preserved and refined for undersampled image reconstruction.

Impact: Detail-preserving self-supervised federated learning method can effectively preserve more details compared to self-supervised learning methods.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, Self-supervised, Reconstruction

Motivation: To improve self-supervised deep learning (DL) reconstruction for highly-accelerated acquisition regimes.

Goal(s): To introduce the concept of cyclic-consistency to improve self-supervised DL reconstruction for highly-accelerated MRI.

Approach: Cyclic-consistency data is formed by simulating new undersampled acquisitions from the neural network output, with a similar undersampling pattern distribution as the true one. Then reconstruction on these simulated data is trained to match acquired data at the true sampling locations, building cyclic consistency for network training. This is supplemented with a conventional self-supervised masking strategy.

Results: The proposed method significantly reduces artifacts at rate 6 and 8 fastMRI reconstruction, and 20-fold fMRI. 

Impact: Substantial reduction in aliasing artifacts is achieved at high acceleration rates using the proposed cyclic-consistent self-supervised learning method compared to existing self-supervised learning methods.

Keywords: AI/ML Image Reconstruction, Image Reconstruction, PROPELLER

Motivation: Periodically Rotated Overlapping Parallel Lines with Enhanced Reconstruction (PROPELLER) is a popular MRI acquisition scheme used for clinical and research MRI data acquisition due to its robustness to motion. However, it is known to have long scan times.

Goal(s): Reduce scan time for PROPELLER scans to make it feasible for usage in regular clinical settings.

Approach: An unrolled algorithm based deep learning reconstruction method for PROPELLER scans has been proposed, which performs reconstruction at the blade level.

Results: Proposed method has been demonstrated to perform good reconstruction on single coil data for multiple anatomies and contrasts.

Impact: This method has the potential to reduce PROPELLER scan times and make it a popular choice for acquisition in clinical settings.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence

Motivation: Typical quantitative MRI (qMRI) methods estimate parameter maps after image reconstructing, which is prone to biases and error propagation.

Goal(s): To propose an end-to-end method to directly estimate qMRI maps from undersampled k-space data using model-based reconstruction and zero-shot network regularization.

Approach: We propose a Nonlinear Conjugate Gradient (NLCG) optimizer for model-based T₂/T₁ estimation, which incorporates U-Net regularization trained in a scan-specific manner.

Results: T2 and T1 mapping results demonstrate the ability of the proposed NLCG-Net to improve estimation quality compared to subspace reconstruction at high accelerations.

Impact: We propose a model-based qMRI technique, NLCG-Net, that incorporates mono-exponential signal modeling with zero-shot scan-specific neural network regularization to enable high fidelity T1 and T2 mapping. 

Keywords: AI/ML Image Reconstruction, Image Reconstruction

Motivation: Random phase-encode undersampling of Cartesian k-space trajectories is widely implemented in magnetic resonance imaging. However, its one-dimensional randomness inherently introduces large coherent aliasing artefacts along the undersampled direction in the reconstruction, which need to be suppressed.

Goal(s): Our goal is to introduce a novel reconstruction scheme to reduce the one-dimensional undersampling-induced aliasing artefacts.

Approach: We propose an intermediate-domain network tailored for operation in image-Fourier space, which utilizes the superior non-coherent properties of decoupled one-dimensional signals to reduce aliasing artifacts.

Results: Experiments illustrate that the proposed method is particularly well-suited for regular Cartesian undersampling scenarios.

Impact: The intermediate-domain network tailored to operate in the Image-Fourier space, can efficiently reduce aliasing artefacts by utilizing the superior incoherence property of the decoupled one-dimensional signals. This could further inspire the development of MRI reconstruction technology based on machine learning.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence

Motivation: The highly reduced k-space measurements would induce noises and artifacts in the reconstructed image in parallel imaging.

Goal(s): To effectively complete the undersampled k-space points for MRI acceleration and provide high-quality images.

Approach: We developed a novel k-space completion framework based on implicit neural representation. The inherent low-rankness of k-space is incorporated into the model to capture the continuous representation in k-space. The proposed method was evaluated on the public dataset and compared with the image and k-space domain reconstruction methods.

Results: The results show that our method can effectively complete the undersampled k-space points without any priors in the image domain.

Impact: The proposed method leverages implicit neural representation in the k-space reconstruction, demonstrating the ability to complete the undersampled k-space points at high acceleration factor. This result implies our method can further reduce the measured k-space points and accelerate MRI acquisition.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, Compressed Sensing, Parallel Imaging

Motivation: To improve learning-based MRI reconstructions to achieve higher clinical accuracy and detail retention.

Goal(s): To introduce modifications in the end-to-end (E2E) variational network (VarNet) to enhance its performance for undersampled MRI reconstructions.

Approach: We performed training in feature-space instead of image-space and included an attention layer that leverages the spatial locations of the Cartesian undersampling artifacts. We combined the new network with the E2E VarNet into Feature-Image VarNet to facilitate cross-domain learning.

Results: Reconstructions were evaluated using standard metrics and clinical scoring. Feature-Image VarNet outperformed all open-source models in the fastMRI leaderboard and preserved small anatomical details that were blurred with E2E VarNet.

Impact: We propose the Feature-Image (FI) variational network (VarNet), which performs cross-domain learning between feature and image spaces. FI VarNet significantly enhances the reconstruction quality of undersampled MRI and could enable clinically acceptable reconstructions at higher acceleration factors than currently possible.

Keywords: AI/ML Image Reconstruction, Quantitative Imaging

Motivation: Physics-based deep learning has been increasingly applied to MRI image reconstruction to accelerate acquisitions.

Goal(s): Here, we investigate whether including a relaxometry model into these networks enables higher quality accelerated reconstructions, and consequently more accurate quantitative maps.

Approach: Two recurrent inference machines with different physics models were implemented: (1) reconstruction of contrast-weighted image series and (2) direct T2 map estimation, from undersampled k-space data

Results: Including relaxometry into physics-informed networks improves reconstruction and T2 map quality for acceleration factors as high as 8-fold.

Impact: Integrating relaxometry models into physics-informed deep learning-based image reconstruction methods enables high quality quantitative mapping directly from undersampled k-space data, from which contrast-weighted images can also be accurately synthesised.

Keywords: AI/ML Image Reconstruction, Cardiovascular

Motivation: Myocardial T₁ and T₂ mapping is crucial in the assessment of cardiovascular disease. 3D whole-heart joint T₁/T₂ mapping approaches have been proposed, however they require long reconstruction times.

Goal(s): By leveraging deep learning (DL)-based techniques, we aim to significantly reduce the reconstruction times for 3D whole-heart joint T₁/T₂ mapping, while maintaining high-quality results.

Approach: Recently a joint group sparsity-based DL approach was proposed for image reconstruction of undersampled multi-contrast MRI data. Here, we propose to extend this approach for non-rigid motion-corrected reconstructions for multi-contrast 3D data for joint T₁/T₂ mapping.

Results: Our approach achieves good agreement with reference techniques, while outperforming single-contrast reconstructions.

Impact: Joint group sparsity-based deep learning non-rigid motion-corrected reconstruction for multi-dimensional joint 3D T₁/T₂ whole-heart mapping achieves good agreement with reference techniques and outperforms single-contrast reconstructions. The approach significantly reduces reconstruction times, making it feasible for clinical applications.

Keywords: AI/ML Image Reconstruction, Image Reconstruction

Motivation: Self-supervised deep learning has shown good performance in reconstructing undersampled k-space. While recent developments focus on improving reconstruction performance for a given undersampling pattern, there is limited research aiming to learn and optimize k-space sampling strategies to offer a performance gain in self-supervised reconstruction.

Goal(s): To design a deep learning framework to optimize the sampling pattern in self-supervised MRI reconstruction.

Approach: An Auto Mask Module was optimized simultaneously with the self-supervised reconstruction module in an end-to-end framework.

Results: The proposed method can achieve better reconstruction results than self-supervised methods based on fixed masks.

Impact: The proposed method can produce better self-supervised reconstruction results by optimizing the k-space undersampling pattern.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence

Motivation: Deep learning methods for accelerated MRI achieve state-of-the-art results but largely ignore additional speedups possible with noncartesian sampling trajectories.

Goal(s): To create a generative diffusion model-based reconstruction algorithm for multi-coil undersampled spiral MRI.

Approach: We train a conditional diffusion model and use frequency-based guidance to ensure consistency between images and measurements.

Results: Evaluated on retrospective data, we show high quality (SSIM > 0.87) in reconstructed images with ultrafast scan times (0.02 seconds for a 2D image). We use this algorithm to identify a set of optimal variable-density spiral trajectories and show large improvements in image quality compared to conventional nufft reconstruction.

Impact: We apply diffusion models to the task of non-cartesian reconstruction. Combining efficient spiral sampling trajectories, multicoil imaging, and deep learning reconstruction enables drastically accelerated imaging. Potential applications of this technology include real-time 3D imaging.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence, Neural Fields, Undersampling reconstruction

Motivation: 3D MRI is fundamental for the assessment of cardiovascular   disease   but   suffers   from   long   scan   times. Undersampled reconstruction techniques have been proposed to accelerate the acquisition, but require long computational times for training.

Goal(s): To develop an unsupervised undersampled reconstruction approach based on implicit neural-field representations for 3D Cartesian MRI.

Approach: Dataset  was   acquired   using   image-based-navigator (iNAV). iNAV-based translational motion was corrected in k-space. Undersampled reconstruction was performed using a Neural-Fields. The method is evaluated on undersampled multi-coil data in comparison to a state-of-the-art.

Results: The feasible reconstruction results show similar image quality to the state-of-the-art reference, holding promise for future clinical evaluation.

Impact: The method proposed can be generalized to any context of reconstruction.  The use in digital devices is feasible, ensuring its possible medical use. Furthermore, this work methodology could allow the use of the net architecture given for other research contexts.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence

Motivation: The optimal selection of data consistency (DC) weight is task-dependent and a challenge in current deep learning unrolled reconstruction network which may result in compromised image quality, and thus deserves further investigations.

Goal(s): To propose a method to obtain adaptive data consistency weight which is superior to existing methods.

Approach: An image-feature-understanding data consistency (IFUDC) modulator is integrated into network to obtain adaptive DC weight based on input images.

Results: Image quality metrics (SSIM, PSNR) of proposed method are higher than those of existing method.

Impact: IFUDC is effective to modulate DC weight adaptively and helps to mitigate the difficulty in optimal DC weight selection.

Keywords: AI/ML Image Reconstruction, Machine Learning/Artificial Intelligence

Motivation: Using deep learning methods to improve image quality and computation speed of compressed sensing reconstruction.

Goal(s): Develop a model-based deep learning model with adaptive threshold selection module to improve the reconstruction quality.

Approach: Introducing a new shrinkage function with adaptive threshold selection for Model-driven deep learning networks, and emploits End-to-End strategy for multicoil reconstruction.

Results: Experiments demonstrate the efficacy of End-to-End reconstruction strategy with sensitivity reconstruction module, and show that proposed adaptive threshold selection method can effectively reduce reconstruction errors.

Impact: We develop an End-to-End deep learning reconstruction network with adaptive threshold selection module. This network canenforce the performance of state-of-art model-based deep learning method for CS reconstruction, and achieve good reconstruction quality at R=8.