Magnetic Resonance Imaging Reconstruction Research Articles

Deep Convolutional Neural Network for Compressive Sensing of Magnetic Resonance Images

Compressive Sensing for Magnetic Resonance Imaging (CS-MRI) aims to reconstruct Magnetic Resonance (MR) images from under-sampled raw data. There are two challenges to improve CS-MRI methods, i.e. designing an under-sampling algorithm to achieve optimal sampling, as well as designing fast and small deep neural networks to obtain reconstructed MR images with superior quality. To improve the reconstruction quality of MR images, we propose a novel deep convolutional neural network architecture for CS-MRI named MRCSNet. The MRCSNet consists of three sub-networks, a compressive sensing sampling sub-network, an initial reconstruction sub-network, and a refined reconstruction sub-network. Experimental results demonstrate that MRCSNet generates high-quality reconstructed MR images at various under-sampling ratios, and also meets the requirements of real-time CS-MRI applications. Compared to state-of-the-art CS-MRI approaches, MRCSNet offers a significant improvement in reconstruction accuracies, such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM). Besides, it reduces the reconstruction error evaluated by the Normalized Root-Mean-Square Error (NRMSE). The source codes are available at https://github.com/TaihuLight/MRCSNet .

Monitoring Bone Density Using Microwave Tomography of Human Legs: A Numerical Feasibility Study.

A major cause of bone mass loss worldwide is osteoporosis. X-ray is considered to be the gold-standard technique to diagnose this disease. However, there is currently a need for an alternative modality due to the ionizing radiations used in X-rays. In this vein, we conducted a numerical study herein to investigate the feasibility of using microwave tomography (MWT) to detect bone density variations that are correlated to variations in the complex relative permittivity within the reconstructed images. This study was performed using an in-house finite-element method contrast source inversion algorithm (FEM-CSI). Three anatomically-realistic human leg models based on magnetic resonance imaging reconstructions were created. Each model represents a leg with a distinct fat layer thickness; thus, the three models are for legs with thin, medium, and thick fat layers. In addition to using conventional matching media in the numerical study, the use of commercially available and cheap ultrasound gel was evaluated prior to bone image analysis. The inversion algorithm successfully localized bones in the thin and medium fat scenarios. In addition, bone volume variations were found to be inversely proportional to their relative permittivity in the reconstructed images with the root mean square error as low as 2.54. The observations found in this study suggest MWT as a promising bone imaging modality owing to its safe and non-ionizing radiations used in imaging objects with high quality.

A Generative Adversarial Network technique for high-quality super-resolution reconstruction of cardiac magnetic resonance images

PurposeIn this paper, we proposed a Denoising Super-resolution Generative Adversarial Network (DnSRGAN) method for high-quality super-resolution reconstruction of noisy cardiac magnetic resonance (CMR) images. MethodsThe proposed method is based on feed-forward denoising convolutional neural network (DnCNN) and SRGAN architecture. Firstly, we used a feed-forward denoising neural network to pre-denoise the CMR image to ensure that the input is a clean image. Secondly, we use the gradient penalty (GP) method to solve the problem of the discriminator gradient disappearing, which improves the convergence speed of the model. Finally, a new loss function is added to the original SRGAN loss function to monitor GAN gradient descent to achieve more stable and efficient model training, thereby providing higher perceptual quality for the super-resolution of CMR images. ResultsWe divided the tested cardiac images into 3 groups, each group of 25 images. Then, we calculated the Peak Signal to Noise Ratio (PSNR) /Structural Similarity (SSIM) between Ground Truth (GT) and the images generated by super-resolution, used them to evaluate our model. We compared with the current widely used method: Bicubic ESRGAN and SRGAN, our method has better reconstruction quality and higher PSNR/SSIM score. ConclusionWe used DnCNN to denoise the CMR image, and then using the improved SRGAN to perform super-resolution reconstruction of the denoised image, we can solve the problem of high noise and artifacts that cause the cardiac image to be reconstructed incorrectly during super-resolution.

Three-Dimensional Segmentation and in silico Comparison of Equine Deep Digital Flexor Tendon Pathology in Horses Undergoing Repeated MRI Examination.

The use of magnetic resonance imaging (MRI) has led to increased clinical and research applications using 3D segmentation and reconstructed volumetric data in musculoskeletal imaging. Lesions of the deep digital flexor tendon (DDFT) are a common pathology in horses undergoing MRI. Three-dimensional MRI reconstruction performed for volumetric tendon analysis in horses has not previously been documented. The aim of this proof-of-concept study was to evaluate the 3D segmentation of horses undergoing repeated MRI at several time points and to perform an analysis of the segmented DDFTs across time. MRI DICOM files were acquired from six horses undergoing repeated MRI examination of the foot for DDFT injury. Once segmented, volumetric tendon surface tessellation language (STL) files were created. Thickness and volumetric data were acquired for each tendon in addition to a tendon comparison across timepoints within each horse. Pearson correlation coefficients were calculated for comparison of MRI reports to computer analysis. There was a significant and positive correlation between MRI and medial record reports of clinical improvement or deterioration and computer analysis (r = 0.56, p = 0.01). The lower end range limit for tendon thickness varied between 0.16 and 1.74 mm. The upper end range limit for DDFT thickness varied between 4.6 and 23.6 mm. During tendon part comparison, changes in DDFT were reported between −3.0 and + 14.3 mm. Changes in DDFT size were non-uniform and demonstrated fluctuations throughout the tendon. The study was successful in establishing the volumetric appearance and thickness of the DDFT as it courses in the foot and tracking this over time. We encountered difficulties in accurate segmentation of the distal insertion of the DDFT as it blends with the distal phalanx. The data demonstrated that the DDFT can be segmented and volumetric studies based on size and shape can be performed using an in silico approach.

High frequency oscillations associate with neuroinflammation in low-grade epilepsy associated tumors

ObjectiveHigh frequency oscillations (HFOs) in intraoperative electrocorticography (ioECoG) are thought to be generated by hyperexcitable neurons. Inflammation may promote neuronal hyperexcitability. We investigated the relation between HFOs and inflammation in tumor-related epilepsy. MethodsWe identified HFOs (ripples 80–250 Hz, fast ripples 250–500 Hz) in the preresection ioECoG of 32 patients with low-grade tumors. Localization of recorded HFOs was classified based on magnetic resonance imaging reconstructions: in tumor, in resected non-tumorous area and outside the resected area. We tested if the following inflammatory markers in the tumor or peritumoral tissue were related to HFOs: activated microglia, cluster of differentiation 3 (CD3)-positive T-cells, interleukin 1-beta (IL1β), toll-like receptor 4 (TLR4) and high mobility group box 1 protein (HMGB1). ResultsTumors that generated ripples were infiltrated by more CD3-positive cells than tumors without ripples. Ripple rate outside the resected area was positively correlated with IL1β/TLR4/HMGB1 pathway activity in peritumoral area. These two areas did not directly overlap. ConclusionsRipple rates may be associated with inflammatory processes. SignificanceOur findings support that ripple generation and spread might be associated with synchronized fast firing of hyperexcitable neurons due to certain inflammatory processes. This pilot study provides arguments for further investigations in HFOs and inflammation.

Fractional Sailfish Optimizer with Deep Convolution Neural Network for Compressive Sensing Based Magnetic Resonance Image Reconstruction

AbstractMagnetic resonance image (MRI) is extensively adapted in clinical diagnosis due to its improved representation of changes with soft tissue. The current innovations in compressive sensing (CS) showed that it is viable to precisely reconstruct MRI with K-space data with reduced scanning duration and it takes more time for computation and it is complex to capture the fine details. The CS is capable to minimize the imaging time and it uses compressibility for reconstructing the images. In this paper, a novel deep-learning method is devised for reconstructing images using a set of MRI. Here, MRI acquisition and MRIsub sampling is performed for reducing the size of images to perform improved analysis. Here, a convolution layer is utilized for imitating the process of compressed sampling that can learn the sampling matrix to prevent complex designs. In addition, a convolution layer is utilized to execute the initial reconstruction. In addition, Deep Convolutional Neural Network (Deep CNN) is adapted for image reconstruction. The training of Deep CNN is performed using proposed Fractional Sailfish Optimizer (FSO) algorithm. The proposed FSO is designed by incorporating fractional concepts in Sail fish optimization for tuning the optimal weights of Deep CNN. The proposed FSO-based CSNet outperformed other methods with minimal MSE of 1.877 and maximal PSNR of 45.396 dB.

Compressively sampled magnetic resonance imaging reconstruction based on split Bregman iteration with general non-uniform threshold shrinkage

PurposeK-space under-sampling reconstruction technology is an effective means to improve the speed of magnetic resonance imaging. Among its many reconstruction algorithms, split Bregman iteration is an effective method to solve multi-constrained models. This model often contains TV variational regularization terms, the generalized threshold shrinkage operator often used to solve TV constraints subproblem. However, when the generalized threshold shrinkage operator is performing the shrinking operation, it does not consider the inconsistency of the elements in the image matrix, which will cause the loss of image details. MethodsIn response to this problem, in this paper, a non-uniform threshold shrinkage operator was proposed to solve above TV constraints subproblem, which can dynamically adjust the shrinkage threshold by the residuals of each image element. And introduce this operator when performing Split Bregman iteration to improve the performance of generalized threshold shrinkage. ResultsAfter qualitative and quantitative analysis during the experiments, it can be concluded that compared with the other three methods, the proposed method has better performance in terms of Peak Signal to Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Transferred Edge Information(TEI) and Normalized Mutual Information(NMI), and the visual perception is better. Then we also did denoising performance analysis at different noise levels, this method also showed good robustness. ConclusionsThe proposed method can improve the reconstruction performance of TV constrained subproblem in split Bregman iteration, and then improve the overall performance of reconstruction algorithm. Moreover, this method also shows good denoising performance at different noise levels.

Prior ensemble learning : Theory and application to MR image priors.

Compressed sensing (CS) reduces the measurement time of magnetic resonance (MR) imaging, where the use of regularizers or image priors are key techniques to boost reconstruction precision. The optimal prior generally depends on the subject and the hand-building of priors is hard. A methodology of combining priors to create a better one would be useful for various forms of image processing that use image priors. We propose a theory, called prior ensemble learning (PEL), which combines many weak priors (not limited to images) efficiently and approximates the posterior mean (PM) estimate, which is Bayes optimal for minimizing the mean squared error (MSE). The way of combining priors is changed from that of an exponential family to a mixture family. We applied PEL to an undersampled (10%) multicoil MR image reconstruction task. We demonstrated that PEL could combine 136 image priors (norm-based priors such as total variation (TV) and wavelets with various regularization coefficient (RC) values) from only two training samples and that it was superior to the CS-SENSE-based method in terms of the MSE of the reconstructed image. The resulting combining weights were sparse (18% of the weak priors remained), as expected. By the theory, the PM estimator was decomposed into the sparse weighted sum of each weak prior's PM estimator, and the exponential computational complexity for RCs was reduced to polynomial order w.r.t. the number of weak priors. PEL is feasible and effective for a practical MR image reconstruction task.

Tangent vector-based gradient method with [formula omitted]-regularization: Iterative half thresholding algorithm for CS-MRI

Object: This paper presents a new method using tangent vector-based l12-regularization for compressed sensing MR image reconstruction. Materials and MethodsThe proposed method with l12-regularization is tested on four datasets: (i) 1-D sparse signal (ii) numerical cardiac phantom, (iii & iv) two sets of in-vivo cardiac MRI datasets acquired using 30 receiver coil elements with Cartesian and radial trajectories on 3T scanner. The results are compared with standard CS reconstruction, which utilizes l1-regularization. The experiments were also conducted for two different types of samplings: (i) cartesian sub-sampling and (ii) 2D random Gaussian sub-sampling. ResultsThe quality of the reconstructed images is validated through Root Mean Square Error (RMSE) and Peak Signal-to-Noise Ratio (PSNR). The results show that the proposed method outperforms the standard CS reconstructions in our experiments with an improvement of 54.8% in RMSE and 14.3% in terms of PSNR. Moreover, the Gaussian random sub-sampling-based image reconstruction results are better than the Cartesian sub-sampling-based reconstruction results. ConclusionThe results show that the proposed method yields a good sparse signal approximation and superior convergence behavior, which implies a promising technique for the reconstruction of cardiac MR images as compared to the conventional CS algorithm.

Improved Image Quality for Static BLADE Magnetic Resonance Imaging Using the Total-Variation Regularized Least Absolute Deviation Solver.

In order to improve the image quality of BLADE magnetic resonance imaging (MRI) using the index tensor solvers and to evaluate MRI image quality in a clinical setting, we implemented BLADE MRI reconstructions using two tensor solvers (the least-squares solver and the L1 total-variation regularized least absolute deviation (L1TV-LAD) solver) on a graphics processing unit (GPU). The BLADE raw data were prospectively acquired and presented in random order before being assessed by two independent radiologists. Evaluation scores were examined for consistency and then by repeated measures analysis of variance (ANOVA) to identify the superior algorithm. The simulation showed the structural similarity index (SSIM) of various tensor solvers ranged between 0.995 and 0.999. Inter-reader reliability was high (Intraclass correlation coefficient (ICC) = 0.845, 95% confidence interval: 0.817, 0.87). The image score of L1TV-LAD was significantly higher than that of vendor-provided image and the least-squares method. The image score of the least-squares method was significantly lower than that of the vendor-provided image. No significance was identified in L1TV-LAD with a regularization strength of 0.4–1.0. The L1TV-LAD with a regularization strength of 0.4–0.7 was found consistently better than least-squares and vendor-provided reconstruction in BLADE MRI with a SENSitivity Encoding (SENSE) factor of 2. This warrants further development of the integrated computing system with the scanner.

Partial Fourier reconstruction of complex MR images using complex-valued convolutional neural networks.

To provide a complex-valued deep learning approach for partial Fourier (PF) reconstruction of complex MR images. Conventional PF reconstruction methods, such as projection onto convex sets (POCS), uses low-resolution image phase information from the central symmetrically sampled k-space for image reconstruction. However, this smooth phase constraint undermines the phase estimation accuracy in presence of rapid local phase variations, causing image artifacts and limiting the extent of PF reconstruction. Using both magnitude and phase characteristics in big complex image datasets, we propose a complex-valued deep learning approach with an unrolled network architecture for PF reconstruction that iteratively reconstructs PF sampled data and enforces data consistency. We evaluate our approach for reconstructing both spin-echo and gradient-echo data. The proposed method outperformed the iterative POCS PF reconstruction method. It produced better artifact suppression and recovery of both image magnitude and phase details in presence of local phase changes. No noise amplification was observed even for highly PF reconstruction. Moreover, the network trained on axial brain data could reconstruct sagittal and coronal brain and knee data. This method could be extended to 2D PF reconstruction and joint multi-slice PF reconstruction. Our proposed method can effectively reconstruct MR data even at low PF fractions, yielding high-fidelity magnitude and phase images. It presents a valuable alternative to conventional PF reconstruction, especially for phase-sensitive 2D or 3D MRI applications.

Multiscale U-net-based accelerated magnetic resonance imaging reconstruction

Multiscale U-net-based accelerated magnetic resonance imaging reconstruction

Time domain principal component analysis for rapid, real-time 2D MRI reconstruction from undersampled data.

A rapid real-time 2D accelerated method was developed for magnetic resonance imaging (MRI) using principal component analysis (PCA) in the temporal domain. This method employs a moving window of previous dynamic frames to reconstruct the current, real-time frame within this window. This technique could be particularly useful in real-time tracking applications such as in MR-guided radiotherapy, where low latency real-time reconstructions are essential. The method was tested retrospectively on 15 fully-sampled data sets of lung patient data acquired on a 3T Philips Achieva system. High frequency data are incoherently undersampled, while the central low-frequency data are always acquired to characterize the temporal fluctuations through PCA. The undersampling pattern is derived in such a way that all of k-space is acquired within a pre-determined number of frames. The missing data in the current frame are then filled in by fitting the temporal characterizations to the acquired undersampled data, using a pre-determined number of PCs. A subset of six patients was used to test the contour ability of the images. Various accelerations between 3x and 8x were tested along with the optimal number of PCs for fitting. A comparison was also performed with previous work from our group proposed by Dietz etal. as well as with a standard low resolution acquisition. In order to determine how the method would perform at lower signal to noise ratio (SNR), noise levels of 2×, 4×, and 6× were added to the 3T data. Metrics such as normalised mean square error and Dice coefficient were used to measure the reconstruction image quality and contour ability. The proposed method demonstrated good temporal robustness as consistent metrics were detected for the duration of the imaging session. It was found that the optimal number of PCs for temporal fitting was dependent on the acceleration rate. For the data tested, five PCs were found to be optimal at the acceleration rates of 3× and 4×. This number decreases to three at accelerations of 5× and 6× and further decreases to two at an acceleration rate of 8×, likely due to greater instability with fewer acquired data points. The use of too many PCs for fitting increased the chances of noisy reconstruction which affected contourability. The proposed 2D real-time MR acceleration method demonstrated greater robustness in the metrics over time when compared with previous real-time PCA methods using metrics such as normalised mean squared error, peak SNR and structural similarity up to an acceleration of 8x. Improved temporal robustness of image structure contourability and accurate definition was also demonstrated using several metrics including the Dice coefficient. Reconstruction of raw acquired data can be performed at approximately 50ms per frame using an Intel core i5 CPU. The method has the advantage of being very flexible in terms of hardware requirements as it can operate successfully on a single coil channel and does not require specialized computing power to implement in real-time.

Domain knowledge augmentation of parallel MR image reconstruction using deep learning.

A deep learning (DL) method for accelerated magnetic resonance (MR) imaging is presented that incorporates domain knowledge of parallel MR imaging to augment the DL networks for accurate and stable image reconstruction. The proposed DL method employs a novel loss function consisting of a combination of mean absolute error, structural similarity, and sobel edge loss. The DL model takes both original measurements and images reconstructed by the parallel imaging method as inputs to the network. The accuracy of the proposed method was evaluated using two anatomical regions and six MRI contrasts and was compared with state-of-the-art parallel imaging and deep learning methods. The proposed method significantly outperformed the other methods for all the six different contrasts in terms of structural similarity, peak signal to noise ratio, and normalized mean squared error. The out-of-sample performance of the proposed method was assessed for a truly "unseen" case in a volunteer scan. The method produced images without any artificial features, often occurring in the DL image reconstruction methods. A stability analysis was performed by adding perturbations to the input, which demonstrated that the proposed method is robust and stable with respect to small structural changes, and different undersampling ratios. Comprehensive validation on large datasets demonstrated that incorporation of domain knowledge sufficiently regularizes the DL based image reconstruction and produces accurate and stable image enhancement.

Results of the 2020 fastMRI Challenge for Machine Learning MR Image Reconstruction.

Accelerating MRI scans is one of the principal outstanding problems in the MRI research community. Towards this goal, we hosted the second fastMRI competition targeted towards reconstructing MR images with subsampled k-space data. We provided participants with data from 7,299 clinical brain scans (de-identified via a HIPAA-compliant procedure by NYU Langone Health), holding back the fully-sampled data from 894 of these scans for challenge evaluation purposes. In contrast to the 2019 challenge, we focused our radiologist evaluations on pathological assessment in brain images. We also debuted a new Transfer track that required participants to submit models evaluated on MRI scanners from outside the training set. We received 19 submissions from eight different groups. Results showed one team scoring best in both SSIM scores and qualitative radiologist evaluations. We also performed analysis on alternative metrics to mitigate the effects of background noise and collected feedback from the participants to inform future challenges. Lastly, we identify common failure modes across the submissions, highlighting areas of need for future research in the MRI reconstruction community.

3D dense convolutional neural network for fast and accurate single MR image super-resolution

Super-resolution (SR) MR image reconstruction has shown to be a very promising direction to improve the spatial resolution of low-resolution (LR) MR images. In this paper, we presented a novel MR image SR method based on a dense convolutional neural network (DDSR), and its enhanced version called EDDSR. There are three major innovations: first, we re-designed dense modules to extract hierarchical features directly from LR images and propagate the extracted feature maps through dense connections. Therefore, unlike other CNN-based SR MR techniques that upsample LR patches in the initial phase, our methods take the original LR images or patches as input. This effectively reduces computational complexity and speeds up SR reconstruction. Second, a final deconvolution filter in our model automatically learns filters to fuse and upscale all hierarchical feature maps to generate HR MR images. Using this, EDDSR can perform SR reconstructions at different upscale factors using a single model with one stride fixed deconvolution operation. Third, to further improve SR reconstruction accuracy, we exploited a geometric self-ensemble strategy. Experimental results on three benchmark datasets demonstrate that our methods, DDSR and EDDSR, achieved superior performance compared to state-of-the-art MR image SR methods with less computational load and memory usage.

Feasibility and Implementation of a Deep Learning MR Reconstruction for TSE Sequences in Musculoskeletal Imaging.

Magnetic Resonance Imaging (MRI) of the musculoskeletal system is one of the most common examinations in clinical routine. The application of Deep Learning (DL) reconstruction for MRI is increasingly gaining attention due to its potential to improve the image quality and reduce the acquisition time simultaneously. However, the technology has not yet been implemented in clinical routine for turbo spin echo (TSE) sequences in musculoskeletal imaging. The aim of this study was therefore to assess the technical feasibility and evaluate the image quality. Sixty examinations of knee, hip, ankle, shoulder, hand, and lumbar spine in healthy volunteers at 3 T were included in this prospective, internal-review-board-approved study. Conventional (TSES) and DL-based TSE sequences (TSEDL) were compared regarding image quality, anatomical structures, and diagnostic confidence. Overall image quality was rated to be excellent, with a significant improvement in edge sharpness and reduced noise compared to TSES (p < 0.001). No difference was found concerning the extent of artifacts, the delineation of anatomical structures, and the diagnostic confidence comparing TSES and TSEDL (p > 0.05). Therefore, DL image reconstruction for TSE sequences in MSK imaging is feasible, enabling a remarkable time saving (up to 75%), whilst maintaining excellent image quality and diagnostic confidence.

Joint total variation-based reconstruction of multiparametric magnetic resonance images for mapping tissue types.

Multispectral analysis of coregistered multiparametric magnetic resonance (MR) images provides a powerful method for tissue phenotyping and segmentation. Acquisition of a sufficiently varied set of multicontrast MR images and parameter maps to objectively define multiple normal and pathologic tissue types can require long scan times. Accelerated MRI on clinical scanners with multichannel receivers exploits techniques such as parallel imaging, while accelerated preclinical MRI scanning must rely on alternate approaches. In this work, tumor-bearing mice were imaged at 7T to acquire k-space data corresponding to a series of images with varying T1-, T2- and T2*-weighting. A joint reconstruction framework is proposed to reconstruct a series of T1-weighted images and corresponding T1 maps simultaneously from undersampled Cartesian k-space data. The ambiguity introduced by undersampling was resolved by using model-based constraints and structural information from a reference fully sampled image as the joint total variation prior. This process was repeated to reconstruct T2-weighted and T2*-weighted images and corresponding maps of T2 and T2* from undersampled Cartesian k-space data. Validation of the reconstructed images and parameter maps was carried out by computing tissue-type maps, as well as maps of the proton density fat fraction (PDFF), proton density water fraction (PDwF), fat relaxation rate ( and water relaxation rate ( from the reconstructed data, and comparing them with ground truth (GT) equivalents. Tissue-type maps computed using 18% k-space data were visually similar to GT tissue-type maps, with dice coefficients ranging from 0.43 to 0.73 for tumor, fluid adipose and muscle tissue types. The mean T1 and T2 values within each tissue type computed using only 18% k-space data were within 8%-10% of the GT values from fully sampled data. The PDFF and PDwF maps computed using 27% k-space data were within 3%-15% of GT values and showed good agreement with the expected values for the four tissue types.

Learning optical flow for fast MRI reconstruction

Reconstructing high-quality magnetic resonance images (MRIs) from undersampled raw data is of great interest from both technical and clinical point of views. To this date, however, it is still a mathematically and computationally challenging problem due to its severe ill-posedness, resulting from the highly undersampled data. Whilst a number of techniques have been presented to improve image reconstruction, they only account for spatio-temporal regularisation, which shows its limitations in several relevant scenarios including dynamic data. In this work, we propose a new mathematical model for the reconstruction of high-quality medical MRI from few measurements. Our proposed approach combines—in a multi-task and hybrid model—the traditional compressed sensing formulation for the reconstruction of dynamic MRI with motion compensation by learning an optical flow approximation. More precisely, we propose to encode the dynamics in the form of an optical flow model that is sparsely represented over a learned dictionary. This has the advantage that ground truth data is not required in the training of the optical flow term. Furthermore, we present an efficient optimisation scheme to tackle the non-convex problem based on an alternating splitting method. We demonstrate the potentials of our approach through an extensive set of numerical results using different datasets and acceleration factors. Our combined approach reaches and outperforms several state-of-the-art techniques for multi-tasking reconstruction and other classic variational reconstruction schemes. Finally, we show the ability of our technique to transfer phantom based knowledge to real datasets.

Low tensor-train rank with total variation for magnetic resonance imaging reconstruction

Low tensor-train rank with total variation for magnetic resonance imaging reconstruction