Keywords: Sparse & Low-Rank Models, Liver, Low-Field MRI, Quantitative Imaging

Motivation: Multi-contrast acquisitions are the basis for accurate water-fat separation. For fat quantification in the liver, insufficient SNR and long acquisition times are main confounding factors.

Goal(s): Provide enhanced image quality of individual contrast images to allow water-fat separation using conventional algorithms for accelerated acquisitions.

Approach: Joint reconstruction of multiple contrasts using a deep learning-based reconstruction that performs regularization in a locally transformed contrast domain.

Results: The proposed method yielded contrasts with PSNR = 34.85 dB and SSIM = 0.94 , showcasing its superiority over the conventional reconstruction technique (PSNR = 31.28, SSIM = 0.86) when applied to a challenging low-field MRI scenario.

Impact: Combining iterative DL-based reconstruction with LLR regularization not only allows to accelerate multi-contrast acquisitions but also yields images with high SNR for accurate fat fraction quantification. The approach has the potential to translate established liver fat quantification to low-field MRI.

Keywords: Sparse & Low-Rank Models, Image Reconstruction

Motivation: ASL-based non-contrast enhanced (non-CE) time-resolved 4D MRA is a promising approach in the diagnosis of cerebrovascular disease, however, it suffers from low SNR and insufficient spatial and temporal resolution.

Goal(s): Our goal was to enhance the quality of 4D non-CE MRA and diminish artifacts.

Approach:  This study proposed a novel reconstruct method combining the angiography sparsity and subspace modeling on data acquired by golden-angle stack-of-stars radial pulse sequence.

Results: The performance of the proposed method was compared with NUFFT, conventional GRASP and the state-of-the-art GRASP-pro reconstructions. The results suggest that the proposed method improves the image quality of the 4D ASL-based non-CE MRA.

Impact: The proposed reconstruction method can produce 4D non-CE MRA with high spatial-temporal resolution.

Keywords: Sparse & Low-Rank Models, Diffusion Tensor Imaging

Motivation: Diffusion tensor imaging (DTI) is inherently resolution limited by MRI gradient performance and human tolerance of gradient slew rates but could benefit greatly from finer detail for tissue mapping.

Goal(s): To increase the resolution of DTI images by using the joint information between different encoding directions to improve efficiency of upsampling.

Approach: Using a compressed sensing subspace-based reconstruction algorithm on zero-padded k-space to estimate a higher resolution image with finer detail than current interpolation strategies.

Results: Smoother diffusion encoded images and reduced spatial blurring in calculated metrics compared to standard cubic interpolation was achieved.

Impact: A new method for upsampling diffusion images using subspace-based compressed sensing reconstruction is introduced that includes fine detail and reduces noise. Potential for improving on standard cubic interpolation is demonstrated, which will benefit DTI analysis including tractography.

Keywords: Sparse & Low-Rank Models, Multi-Contrast, low-rank, Brain imaging, Quantitative MRI, MR fingerprinting

Motivation: 3D MR fingerprinting (MRF) and MPnRAGE acquire a large number (400-1000+) of non-steady state images with different encodings to estimate quantitative relaxometry parameters. The large number of volumetric images presents serious computational and memory issues for many advanced image reconstruction techniques, especially those utilizing deep learning.

Goal(s): This work develops a memory efficient, pretrained deep factor model (DFM) for high quality, high temporal images. 

Approach: We apply a progressive training and pre-training strategy to accelerate the convergence for a self-supervised DFM.

Results: DFM recovers 384 3D brain images (1mm isotropic resolution) from a 2.3 minutes MPnRAGE scan within 30 minutes of reconstruction time.

Impact: The proposed pretrained deep factor model enables fast MRF reconstruction from accelerated acquisition in a 3D+time setting. It has the potential to significantly shorten the acquisition time for quantitative MRI, while providing high quality weighted MRI results.

Keywords: Sparse & Low-Rank Models, Image Reconstruction, EPTI

Motivation: Subspace reconstruction is widely used in MRI reconstruction, but the selection and impact of bases need further analysis.

Goal(s): We want to evaluate the influence of bases obtained by different methods on subspace reconstruction.

Approach: We generated different bases from Bloch simulation and calibration scan on healthy subjects and brain tumor subjects and evaluate the subspace reconstruction results with GESE-EPTI data.

Results: Subspace bases from calibration scan can optimize the reconstructed results without increasing scan time. The bases from brain tumor subjects and healthy subjects are evaluated and indicate consistent results. Besides, linear transformations can optimize results without the need for reconstruction.

Impact: We evaluate the subspace bases from Bloch simulation and calibration scan in MRI subspace reconstruction and demonstrate bases from calibration scan can optimize the reconstructed results. Subspace reconstruction results were consistent by bases from brain tumor and healthy subjects.

Keywords: Sparse & Low-Rank Models, DSC & DCE Perfusion, Pancreas, Abdomen, Quantitative Imaging, Image Reconstruction, Free-breathing

Motivation: Efficient image models are needed to enable low-dose, free-breathing quantitative dynamic contrast-enhanced (DCE) imaging in the abdomen.

Goal(s): Integrate non-rigid motion compensation (MoCo) into the MR Multitasking framework and evaluate its impact on low-dose, free-breathing abdominal DCE.

Approach: Non-rigid MoCo was incorporated into MR Multitasking by directly applying motion fields to eigenimages. This was tested on n=5 healthy volunteers who received 0.02 mmol/kg Gd, only 20% of the standard dose.

Results: Non-rigid MoCo of eigenimages was compatible with MR Multitasking. MoCo more efficiently modeled respiratory motion and minimized intra-bin motion, demonstrating potential for improved DCE quantification.

Impact: Non-rigid motion compensation reduces intra-bin respiratory motion in low-dose free-breathing, whole-abdomen quantitative dynamic contrast-enhanced (DCE) MR Multitasking. Low-dose quantitative DCE may benefit longitudinal monitoring of neoadjuvant treatment in patients with borderline resectable/locally advanced pancreatic ductal adenocarcinoma.

Keywords: Sparse & Low-Rank Models, Hyperpolarized MR (Non-Gas), Kidney, MR Fingerprinting, Metabolism

Motivation: T1 of hyperpolarized Carbon-13 (HP-¹³C) molecules results in limited data acquisition period. In conventional imaging approaches, this results in sacrifices in imaging resolution, which leads to limited sensitive and interpretability of results.  

Goal(s): Combine efficient MR fingerprinting based acquisition with spatiotemporal low-rank constraint for accelerated high resolution metabolic imaging.

Approach: bSSFP-type MRF acquisition was reconstructed iteratively with low-rank temporal constraint derived from HP-¹³C signal model. Method was assessed in digital phantom, retrospective, and prospective preclinical rat kidney.

Results: Strong undersampling capacity was observed in simulation and retrospective studies.  Preclinical experiment with 20-fold smaller voxel volume showed reasonable results.

Impact: Improved resolution is a critical prerequisite for clinical utility of HP-¹³C measurements. The methods shown here demonstrate potential for robust metabolic measurements at order of magnitude higher resolution, and is adaptable for wide range of organ systems and metabolic processes.

Keywords: Image Reconstruction, New Trajectories & Spatial Encoding Methods, Field-of-view

Motivation: MRI acceleration increases the utility of the machine and increases the robustness to motion.

Goal(s): In this work, we present a completely novel method to accelerate MRI.

Approach: We propose to have a technologist supply a contour surrounding the patient, a non-rectangular field-of-view.  We have created a method to reduce the number of samples required while maintaining high image quality for the contoured region.

Results: We present results of the ankle, knee, and brain where we are able to reconstruct an image of high quality with fewer samples that could be collected with a faster scan.

Impact: A non-rectangular field-of-view can better separates the patient's anatomy from the surrounding air than a rectangular field-of-view would.  We present a non-iterative reconstruction algorithm that takes advantage of the non-rectangular field-of-view and reconstructs a high-quality image from fewer samples.

Keywords: Image Reconstruction, Low-Field MRI, Denoising

Motivation: Low-field MRI is limited by the low signal-to-noise ratio (SNR). Multiple scan averages increase SNR but also increase the acquisition time. In applications that acquire multiple contrasts, such as quantitative imaging, acquisition time can be further prolonged.

Goal(s): To develop a multi-coil multi-contrast k-space denoising technique that can also be compatible with parallel imaging-accelerated datasets.

Approach: A low-rank block-Hankel matrix was constructed from the multi-dimensional k-space data, followed by optimal singular value shrinkage to suppress Gaussian noise.

Results: In a pilot cohort, the proposed method improved SNR by 1.6-fold and reduced standard deviations in quantitative maps in the liver.

Impact: The proposed k-space denoising technique effectively suppresses noise in multi-coil multi-contrast k-space data from low-field MRI and is compatible with parallel imaging-accelerated datasets. It can improve image quality and/or shorten the acquisition time for multi-contrast low-field MRI.

Keywords: Image Reconstruction, Image Reconstruction, Short T2 sequences, Inhomogeneous B0

Motivation: : ZTE has proven to be a powerful MRI sequence for ultrashort T₂ tissues, but it fails to produce useful images in the presence of strong field inhomogeneities.

Goal(s): To develop a method to correct artifacts induced by strong B₀ inhomogeneities in ZTE sequences, based on on-the-fly B₀ maps.

Approach: A B₀ map, obtained by phase difference between two fast SPRITE sequences, is fed into an encoding matrix for posterior image reconstruction by Kaczmarz’s Algebraic Reconstruction Techniques.

Results: Geometric distortions and hyperintense regions resulting from B₀ strong quadratic components are largely reverted with this approach.

Impact: This method can be exploited for e.g. dental imaging with ZTE in affordable low-field MRI systems, and can be generalized to other non-Cartesian sequences. Furthermore, it may prove useful for imaging with extreme magnet geometries as in e.g. single-sided MRI.

Keywords: Image Reconstruction, Cardiovascular

Motivation: Whole-heart CMR with high isotropic spatial resolution involves long and unpredictable scan times.

Goal(s): To propose and validate a super-resolution motion-corrected reconstruction framework to enable accelerated high-resolution whole-heart CMR from lower-resolution acquisitions.

Approach: Low resolution was treated as a k-space down-sampling problem, enabling the adaptation of an end-to-end motion-corrected iterative deep-learning network reconstruction, previously demonstrated for undersampled whole-heart CMRA.

Results: High-resolution whole-heart images (1.5×1.5×1.5 mm³) were obtained from prospective low-resolution data (1.5×6×6 mm³) using the proposed Super-MoCo-MoDL framework, with comparable image quality to a high-resolution acquisition. Scan times decreased from ~3.2 to ~1.2 minutes and reconstruction times were clinically feasible, at ~30 seconds.

Impact: The proposed Super-MoCo-MoDL framework enables data-consistent 3D whole-heart image reconstruction at high isotropic resolution from lower-resolution anisotropic scans. It has the potential to either accelerate whole-heart CMR, increase the feasibility of high-resolution clinical scanning, or a combination of the two.

Keywords: Image Reconstruction, Image Reconstruction

Motivation: Self-supervised neural network reconstruction improves multi-shot diffusion MRI (dMRI), yet suffers from prohibitively long computation times.
 

Goal(s): To develop a zero-shot self-supervised learning method for fast multi-shot dMRI reconstruction.

Approach: We propose a physics-guided neural network that operates in both k- and image-spaces to combine information from different EPI shots. We show that reconstruction quality can be improved with a novel sampling mask strategy, and that faster training is possible with a new training strategy. Finally, we extend our results to SMS acquisitions.

Results: Our results show that the proposed method provides improved and fast reconstructions compared to 2-shot LORAKS and 2-shot ZS-SSL.

Impact: The proposed physics-guided self-supervised learning method provides fast and high-quality reconstruction of multi-shot diffusion MRI volumes, while also eliminating the need for external training datasets. 

Keywords: Image Reconstruction, Heart

Motivation: Cardiac MRI (CMR) is widely used for assessment of cardiac diseases, but long acquisition time can cause patient discomfort and motion artifacts. Existing methods face challenges in reconstructing detailed information from highly undersampled spatiotemporal CMR acquisitions.

Goal(s): We propose a self-consistency guided multi-prior deep learning framework termed $$$k$$$-$$$t$$$ CLAIR to address this challenge.

Approach: This method exploits spatiotemporal correlations in data and incorporates calibration information to learn complementary priors across the $$$x$$$-$$$t$$$, $$$x$$$-$$$f$$$, and $$$k$$$-$$$t$$$ domains.

Results: Evaluation performed on publicly available cardiac cine and T1/T2 mapping datasets demonstrated that the proposed method can effectively reconstruct detailed information from highly undersampled CMR data.

Impact: The proposed method achieves high-quality reconstruction of highly undersampled CMR datasets including both cine imaging and T1/T2 mapping. This method has potential to improve CMR in clinical use.

Keywords: Image Reconstruction, Image Reconstruction, Artifact Reduction

Motivation: Artifacts caused by concomitant fields are significant at lower B0 field strengths, larger distances from isocenter, and when using longer readouts. This is a significant limitation for large-FOV dynamic imaging at low- and mid- field strengths.

Goal(s): To demonstrate dynamic MRI with concomitant field correction.

Approach: We combine state-of-the-art dynamic MRI using undersampling and constrained reconstruction, with a concomitant field mitigation approach (MaxGIRF) that uses a higher-order encoding matrix.

Results: We demonstrate significant artifact reduction in large-FOV dynamic MRI of the lung in coronal orientation at 0.55 Tesla.

Impact: This work resolves one major constraint that currently limits large-FOV and off-isocenter dynamic MRI at low field strengths.  This could provide better imaging of the lung, abdomen, and obese subjects, and better guidance of interventions that utilize a table shift.

Keywords: Image Reconstruction, Image Reconstruction

Motivation: Multiband imaging in EPI Diffusion sequences can suffer from Nyquist ghosting artifacts and poor slice separation. This affects evaluation of ADC, FA, and kurtosis maps in high performance gradient systems.

Goal(s): Reduce ghosting and improve SNR in multiband images so that ADC, FA, and kurtosis maps deviate minimally from single-band imaging.

Approach: EPI data is split into odd and even echoes and independently reconstructed with ARC algorithm. Virtual channel combination with phase correction along with a Deep Learning algorithm provides SNR enhancement.

Results: There was minimal error in the ADC, FA, and kurtosis maps with the proposed approach compared to single-band images.

Impact: Our reconstruction algorithm helps multiband imaging achieve minimal deviation in ADC, FA, orthogonal and parallel kurtoses as in single-band imaging but in a shorter acquisition time.

Keywords: Image Reconstruction, Image Reconstruction, FFT, NUFFT, GFFT, GIFT, Re-gridding, Arbitrary K-space Sampling, Arbitrary K-space trajectory

Motivation: K-space re-gridding or sampling density compensation is required for image reconstruction with arbitrary K-space trajectory, e.g., in FFT, NUFFT, GFFT, etc.

Goal(s): We propose a generalized inverse Fourier transform (GIFT) approach to direct image reconstruction. The reconstruction is continuous in image space.

Approach: We generalize continuous Fourier transform to any coordinate systems, arbitrary K-space sampling/trajectory, and arbitrary K-point size and shape.

Results: Images were calculated from a spiral k-space trajectory in the 2D Cartesian coordinate system, which we use as examples to demonstrate GIFT's reconstruction flexibility for different resolutions, also within any small, focused region of interest (ROI).

Impact: The generalized image reconstruction algorithm apply to both Cartesian and Polar, can be readily used for non-uniform K-space with arbitrary trajectory. Images can be reconstructed with arbitrary resolution also within any small ROIs.