Compressed Sensing Magnetic Resonance Imaging Research Articles

ISP-IRLNet: Joint optimization of interpretable sampler and implicit regularization learning network for accerlerated MRI

Compressed Sensing Magnetic Resonance Imaging (CS-MRI) was proposed to accelerate data acquisition and reconstruct MR images from under-sampled data in k-space. However, the traditional approaches design the sampling patterns separately from the reconstruction process, which often leads to suboptimal reconstruction performance. To address this issue, we propose a joint optimization model dubbed as ISP-IRLNet to yield optimal recovery including an interpretable sampler and a reconstructor using the implicit regularization learning network. In particular, we introduce a probabilistic distribution model of the sampling incorporated into the prior of k-space locations, which provides a robust CS-MRI sample pattern with an explicit expression. To estimate the approximating gradient in backward propagation for more stable training, we employ a differentiable binarization strategy. In addition, we unroll the ADMM optimization for solving the regularized reconstruction model to be a deep network (i.e., IRLNet) with a learnable implicit regularization sub-network to learn a regularizer or prior term implicitly. The experiments on our collected brain and public knee datasets demonstrate that our method provides an optimal sample pattern and achieves superior image reconstruction performance compared with existing methods, especially at high acceleration rates.

Comparing Strain Assessment in Compressed Sensing and Conventional Cine MRI.

The aim of this study is to assess the feasibility of compressed sensing (CS) acceleration methods compared to conventional segmented cine (Seg) cardiac magnetic resonance (CMR) for evaluating left ventricular (LV) function and strain by feature tracking (FT). In this prospective study, 45 healthy volunteers underwent CMR imaging used Seg, threefold (CS3), fourfold (CS4), and eightfold (CS8) CS acceleration. Cine images were scored for quality (1-5 scale). LV volumetric and functional parameters and global longitudinal (GLS), circumferential (GCS), and radial strains (GRS) were quantified. LV volumetric and functional parameters exhibited no differences between Seg and all CS cines (all P > 0.05). The strains were similar for Seg, CS3, and CS4 (all P > 0.05). Similarly, no significant differences were observed in GRS and GCS between Seg and CS8 (all P > 0.05), but the global longitudinal strain was significantly lower for CS8 versus Seg (P < 0.001). Image quality declined with CS acceleration, especially in long-axis views with CS8. CS cine MRI at acceleration factor 4 maintained good image quality and accurate measurements of LV function and strain, although there was a slight reduction in the quality of long-axis images and GLS with CS8. CS acceleration up to a factor of 4 enabled fast CMR evaluations, making it suitable for clinical use.

Compressed sensing magnetic resonance imaging (CS-MRI) diagnosis of rotator cuff tears

Compressed sensing magnetic resonance imaging (CS-MRI) diagnosis of rotator cuff tears

Short Communication: Localized Adversarial Artifacts for Compressed Sensing MRI

Short Communication: Localized Adversarial Artifacts for Compressed Sensing MRI

GROG-pCS: GRAPPA Operator Gridding with CS-based p-thresholding for Under-sampled Radially Encoded MRI

Major limitation of MRI is long scan time. Compressed Sensing (CS) is a contemporary technique used to accelerate MRI scan time. In CS, fully sampled MRI images are reconstructed from the partially acquired k-space data. In CS MRI, the utilization of a non-linear reconstruction algorithm is one of the key requirements for successful signal recovery. Numerous methods have been used in CS for solving the non-linear problems to get the solution image. In this paper, we proposed GRAPPA Operator gridding (GROG) with CS-based p-thresholding to reconstruct the artefact free MR images from the partially acquired radial k-space data. In this proposed scheme, initially radially acquired under-sampled k-space data is mapped onto Cartesian space using GROG gridding and then CS reconstruction is performed by using iterative p-thresholding. The proposed method is tested on four MRI data sets, (i) simulated Shepp-Logan phantom, (ii) 1.5T human brain data, (iii) 3T human brain, and (iv) 3T short-axial cardiac (SA) radial data. The reconstruction results are compared with the CS-based iterative hard-thresholding and soft-thresholding reconstructions. The quality of the solution images is evaluated by using (i) Artifact Power (AP), (ii) Root Mean Square Error (RMSE), and (iii) Peak Signal-to-Noise Ratio (PSNR).

Hierarchical Perception Adversarial Learning Framework for Compressed Sensing MRI.

The long acquisition time has limited the accessibility of magnetic resonance imaging (MRI) because it leads to patient discomfort and motion artifacts. Although several MRI techniques have been proposed to reduce the acquisition time, compressed sensing in magnetic resonance imaging (CS-MRI) enables fast acquisition without compromising SNR and resolution. However, existing CS-MRI methods suffer from the challenge of aliasing artifacts. This challenge results in the noise-like textures and missing the fine details, thus leading to unsatisfactory reconstruction performance. To tackle this challenge, we propose a hierarchical perception adversarial learning framework (HP-ALF). HP-ALF can perceive the image information in the hierarchical mechanism: image-level perception and patch-level perception. The former can reduce the visual perception difference in the entire image, and thus achieve aliasing artifact removal. The latter can reduce this difference in the regions of the image, and thus recover fine details. Specifically, HP-ALF achieves the hierarchical mechanism by utilizing multilevel perspective discrimination. This discrimination can provide the information from two perspectives (overall and regional) for adversarial learning. It also utilizes a global and local coherent discriminator to provide structure information to the generator during training. In addition, HP-ALF contains a context-aware learning block to effectively exploit the slice information between individual images for better reconstruction performance. The experiments validated on three datasets demonstrate the effectiveness of HP-ALF and its superiority to the comparative methods.

Undersampled MRI reconstruction based on spectral graph wavelet transform

Compressed sensing magnetic resonance imaging (CS-MRI) has exhibited great potential to accelerate magnetic resonance imaging if an image can be sparsely represented. How to sparsify the image significantly affects the reconstruction quality of images. In this paper, a spectral graph wavelet transform (SGWT) is introduced to sparsely represent magnetic resonance images in iterative image reconstructions. The SGWT is achieved by extending the traditional wavelets transform to the signal defined on the vertices of the weighted graph, i.e. the spectral graph domain. This SGWT uses only the connectivity information encoded in the edge weights, and does not rely on any other attributes of the vertices. Therefore, SGWT can be defined and calculated for any domain where the underlying relations between data locations can be represented by a weighted graph. Furthermore, we present a Chebyshev polynomial approximation algorithm for fast computing this SGWT transform. The l1 norm regularized CS-MRI reconstruction model is introduced and solved by the projected iterative soft-thresholding algorithm to verify its feasibility. Numerical experiment results demonstrate that our proposed method outperforms several state-of-the-art sparsify transforms in terms of suppressing artifacts and achieving lower reconstruction errors on the tested datasets.

An Efficient Lightweight Generative Adversarial Network for Compressed Sensing Magnetic Resonance Imaging Reconstruction

Compressed-sensing-based magnetic resonance imaging (CS-MRI) methods can significantly shorten scanning time while ensuring reconstructed image quality. Recently, deep learning methods, particularly generative adversarial networks (GAN), have been introduced into CS-MRI. However, these GAN-based methods suffer from their heavy learning parameters and ignore long-range dependency, which degrades the reconstructed image quality. Thus, the objective of this study is to design an efficient lightweight GAN to achieve more accurate MRI reconstruction. The proposed framework, named SepGAN, utilises depthwise separable convolution as the basic component to reduce the number of learning parameters. Two modules, the dilated depthwise separable convolution dense block and a squeeze-and-excitation lightweight self-attention module were proposed to extract the long-range dependency and improve the representational ability. The focal frequency loss was also involved in assisting the model to focus on high-frequency information. To evaluate the performance of the three proposed methods and SepGAN, two brain datasets were used in our experiment. From the results of our comparison analysis, SepGAN possesses the minimum number of parameters and multiply–accumulate operations (i.e., 7.32 M and 13.62G) and outperforms other methods in a variety of evaluation metrics, especially in Frechet inception distance, proving that reconstructed images of our method have better visual effects. For unseen pathological data, SepGAN can also perform effective reconstruction with explicit tumour textures and boundaries. The experimental results demonstrate that SepGAN can reconstruct high quality images with fewer parameters and exhibit remarkable generalisation ability.

Improving Brain Volume Measurement Workflow using Combination of Compressed Sensing MRI and Deep Learning based Super-Resolution at 1.5T Clinical Scanner

Improving Brain Volume Measurement Workflow using Combination of Compressed Sensing MRI and Deep Learning based Super-Resolution at 1.5T Clinical Scanner

Transferring Deep Gaussian Denoiser for Compressed Sensing MRI Reconstruction

Deep neural networks have achieved the most outstanding performance in compressed sensing magnetic resonance imaging (CS-MRI) reconstruction by learning the potential structures of images from a large number of training samples. However, the required data comprising hundreds of subjects are usually rare. In this article, we remedy this problem by transferring the easy-to-get deep Gaussian denoisers trained with natural images for artifact reduction in the iterative recovery process without the use of full-sampled MRI data. To this end, we first train a set of deep Gaussian networks with natural images and then incorporate them into a play-and-plug framework that is built by modifying the proximal gradient algorithm with the classical momentum strategy. Furthermore, a nonlocal denoiser is employed for efficiently removing the artifacts. It is found that the momentum strategy can make the statistical distribution of artifacts approximately be Gaussian, making the Gaussian denoisers available for CS-MRI. In the experiments, we verify the rationality of the established framework and show that our method consistently outperforms state-of-the-art methods.

Compressive Sensing MRI Reconstruction with Shearlet Sparsity and non-Convex Hybrid Total Variation

Compressive Sensing MRI Reconstruction with Shearlet Sparsity and non-Convex Hybrid Total Variation

A cascade of preconditioned conjugate gradient networks for accelerated magnetic resonance imaging

Background and objective: Recent unfolding based compressed sensing magnetic resonance imaging (CS-MRI) methods only reinterpret conventional CS-MRI optimization algorithms and, consequently, inherit the weaknesses of the alternating optimization strategy. In order to avoid the structural complexity of the alternating optimization strategy and achieve better reconstruction performance, we propose to directly optimize the ℓ1 regularized convex optimization problem using a deep learning approach. Method: In order to achieve direct optimization, a system of equations solving the ℓ1 regularized optimization problem is constructed from the optimality conditions of a novel primal-dual form proposed for the effective training of the sparsifying transform. The optimal solution is obtained by a cascade of unfolding networks of the preconditioned conjugate gradient (PCG) algorithm trained to minimize the mean element-wise absolute difference (ℓ1 loss) between the terminal output and ground truth image in an end-to-end manner. The performance of the proposed method was compared with that of U-Net, PD-Net, ISTA-Net+, and the recently proposed projection-based cascaded U-Net, using single-coil knee MR images of the fastMRI dataset. Results: In our experiment, the proposed network outperformed existing unfolding-based networks and the complex version of U-Net in several subsampling scenarios. In particular, when using the random Cartesian subsampling mask with 25 % sampling rate, the proposed model outperformed PD-Net by 0.76 dB, ISTA-Net+ by 0.43 dB, and U-Net by 1.21 dB on the positron density without suppression (PD) dataset in term of peak signal to noise ratio. In comparison with the projection-based cascade U-Net, the proposed algorithm achieved approximately the same performance when the sampling rate was 25% with only 1.62% number of network parameters at the cost of a longer reconstruction time (approximately twice). Conclusion: A cascade of unfolding networks of the PCG algorithm was proposed to directly optimize the ℓ1 regularized CS-MRI optimization problem. The proposed network achieved improved reconstruction performance compared with U-Net, PD-Net, and ISTA-Net+, and achieved approximately the same performance as the projection-based cascaded U-Net while using significantly fewer network parameters.

Comprehensive Neonatal Cardiac, Feed and Wrap, Non-contrast, Non-sedated, Free-breathing Compressed Sensing 4D Flow MRI Assessment.

Cardiac MRI is an important imaging tool in congenital cardiac disease, but its use has been limited in the neonatal population as general anesthesia has been needed for breath-holding. Technological advances in four-dimensional (4D) flow MRI have now made nonsedated free-breathing acquisition protocols a viable clinical option, but the method requires prospective validation in neonates. To test the feasibility of compressed sensing (CS) 4D flow MRI in the neonatal population and to compare with standard previously validated two-dimensional (2D) phase-contrast (PC) flow MRI. Prospective, cohort, image quality. A total of 14 healthy neonates (median [range] age: 2.5 [0-80] days; 8 male). Noncontrast 2D cine gradient echo sequence with through-plane velocity encoding (PC) sequence and compressed sensing (CS) three-dimensional (3D), time-resolved, cine phase-contrast MRI with 3D velocity-encoding (4D flow MRI) at 3 T. Aortic 2D PC, and aortic, pulmonary trunk and superior vena cava CS 4D flow MRI were acquired using the feed and wrap technique (nonsedated) and quantified using commercially available software. Aortic flow and peak velocity were compared between methods. Internal consistency of 4D flow MRI was determined by comparing mean forward flow of the main pulmonary artery (MPA) vs. the sum of left and right pulmonary artery flows (LPA and RPA) and by comparing mean ascending aorta forward flow (AAo) vs. the sum of superior vena cava (SVC) and descending aorta flows (DAo). Flow and peak-velocity comparisons were assessed using paired t-tests, with P < 0.05 considered significant, and Bland-Altman analysis. Interobserver and intraobserver agreement and internal consistency were analyzed by intraclass correlation co-efficient (ICC). There was no statistically significant difference between ascending aortic forward flow between 2D PC and CS 4D Flow MRI (P=0.26) with a bias of 0.11 mL (-0.59 to 0.82 mL) nor peak velocity (P =0.11), with a bias of -5cm/sec and (-26 to 16 cm/sec). There was excellent interobserver and intraobserver agreement for each vessel (interobserver ICC: AAo 1.00; DAo 0.94, SVC 0.90, MPA 0.99, RPA 0.98, LPA 0.96; intraobserver ICC: AAo 1.00; DAo 0.99, SVC 0.98, MPA 1.00, RPA 1.00, LPA 0.99). Internal consistency measures showed excellent agreement for both mean forward flow of main pulmonary artery vs. the sum of left and right pulmonary arteries (ICC: 0.95) and mean ascending aorta forward flow vs. the sum of superior vena cava and descending aorta flows (ICC: 1.00). Sedation-free neonatal feed and wrap MRI is well tolerated and feasible. CS 4D flow MRI quantification is similar to validated 2D PC free-breathing imaging with excellent interobserver and intraobserver agreement. 1 TECHNICAL EFFICACY: Stage 2.

PUERT: Probabilistic Under-Sampling and Explicable Reconstruction Network for CS-MRI

Compressed Sensing MRI (CS-MRI) aims at reconstructing de-aliased images from sub-Nyquist sampling <i>k</i>-space data to accelerate MR Imaging, thus presenting two basic issues, <i>i.e.</i>, where to sample and how to reconstruct. To deal with both problems simultaneously, we propose a novel end-to-end Probabilistic Under-sampling and Explicable Reconstruction neTwork, dubbed PUERT, to jointly optimize the sampling pattern and the reconstruction network. Instead of learning a deterministic mask, the proposed sampling subnet explores an optimal probabilistic sub-sampling pattern, which describes independent Bernoulli random variables at each possible sampling point, thus retaining robustness and stochastics for a more reliable CS reconstruction. A dynamic gradient estimation strategy is further introduced to gradually approximate the binarization function in backward propagation, which efficiently preserves the gradient information and further improves the reconstruction quality. Moreover, in our reconstruction subnet, we adopt a model-based network design scheme with high efficiency and interpretability, which is shown to assist in further exploitation for the sampling subnet. Extensive experiments on two widely used MRI datasets demonstrate that our proposed PUERT not only achieves state-of-the-art results in terms of both quantitative metrics and visual quality but also yields a sub-sampling pattern and a reconstruction model that are both customized to training data.

High-Throughput Deep Unfolding Network for Compressive Sensing MRI

Deep unfolding network (DUN) has become the mainstream for compressive sensing MRI (CS-MRI) due to its good interpretability and high performance. Different optimization algorithms are usually unfolded into deep networks with different architectures, in which one iteration corresponds to one stage of DUN. However, there are few works discussing the following two questions: Which optimization algorithm is better after being unfolded into a DUN? What are the bottlenecks in existing DUNs? This paper attempts to answer these questions and give a feasible solution. For the first question, our mathematical and empirical analysis verifies the similarity of DUNs unfolded by alternating minimization (AM), alternating iterative shrinkage-thresholding algorithm (ISTA) and alternating direction method of multipliers (ADMM). For the second question, we point out that one major bottleneck of existing DUNs is that the input and output of each stage are just images of one channel, which greatly limits the transmission of network information. To break the information bottleneck, this paper proposes a novel, simple yet powerful high-throughput deep unfolding network (HiTDUN), which is not constrained by any optimization algorithm and can transmit multi-channel information between adjacent network stages. The developed multi-channel fusion strategy can also be easily incorporated into existing DUNs to further boost their performance. Extensive CS-MRI experiments on three benchmark datasets demonstrate that the proposed HiTDUN outperforms existing state-of-the-art DUNs by large margins while maintaining fast computational speed.<xref ref-type="fn" rid="fn1">¹</xref><fn id="fn1"><label>¹</label> For reproducible research, the source codes and training models of our HiTDUN. [Online]. Available: <uri>https://github.com/jianzhangcs/HiTDUN</uri>. </fn>

MRI Reconstruction with Separate Magnitude and Phase Priors Based on Dual-Tree Complex Wavelet Transform.

The methods of compressed sensing magnetic resonance imaging (CS-MRI) can be divided into two categories roughly based on the number of target variables. One group devotes to estimating the complex-valued MRI image. And the other calculates the magnitude and phase parts of the complex-valued MRI image, respectively, by enforcing separate penalties on them. We propose a new CS-based method based on dual-tree complex wavelet (DT CWT) sparsity, which is under the frame of the second class of CS-MRI. Owing to the separate regularization frame, this method reduces the impact of the phase jumps (that means the jumps or discontinuities of phase values) on magnitude reconstruction. Moreover, by virtue of the excellent features of DT CWT, such as nonoscillating envelope of coefficients and multidirectional selectivity, the proposed method is capable of capturing more details in the magnitude and phase images. The experimental results show that the proposed method recovers the image contour and edges information well and can eliminate the artifacts in magnitude results caused by phase jumps.

DBGAN: A dual-branch generative adversarial network for undersampled MRI reconstruction

Compressed sensing magnetic resonance imaging (CS-MRI) greatly accelerates the acquisition process and yield considerable reconstructed images. Deep learning was introduced into CS-MRI to further speed up the reconstruction process and improve the image quality. Recently, generative adversarial network (GAN) using two-stage cascaded U-Net structure as generator has been proven to be effective in MRI reconstruction. However, previous cascaded structure was limited to few feature information propagation channels thus may lead to information missing. In this paper, we proposed a GAN-based model, DBGAN, for MRI reconstruction from undersampled k-space data. The model uses cross-stage skip connection (CSSC) between two end-to-end cascaded U-Net in our generator to widen the channels of feature propagation. To avoid discrepancy between training and inference, we replaced classical batch normalization (BN) with instance normalization (IN) . A stage loss is involved in the loss function to boost the training performance. In addition, a bilinear interpolation decoder branch is introduced in the generator to supplement the missing information of the deconvolution decoder. Tested under five variant patterns with four undersampling rates on different modality of MRI data, the quantitative results show that DBGAN model achieves mean improvements of 3.65 dB in peak signal-to-noise ratio (PSNR) and 0.016 in normalized mean square error (NMSE) compared with state-of-the-art GAN-based methods on T1-Weighted brain dataset from MICCAI 2013 grand challenge. The qualitative visual results show that our method can reconstruct considerable images on brain and knee MRI data from different modality. Furthermore, DBGAN is light and fast – the model parameters are fewer than half of state-of-the-art GAN-based methods and each 256 × 256 image is reconstructed in 60 milliseconds, which is suitable for real-time processing.

IDPCNN: Iterative denoising and projecting CNN for MRI reconstruction

Compressed sensing magnetic resonance imaging (CS-MRI) makes it possible to shorten data acquisition time substantially. The traditional iteration-based CS-MRI method is flexible in modeling but is usually time-consuming. Recently, the deep neural network method becomes popular in CS-MRI due to its high efficiency. However, the drawback of the deep learning method is inflexibility. It depends overly on the training images and scanning method of the k-space data. In this paper, we propose an iterative method for MRI reconstruction, called IDPCNN, combining the merits of both the traditional method and the deep learning methods, realizing quick, flexible, and accurate reconstruction. The proposed method incorporates two stages: denoising and projection. The denoising step employs a state-of-the-art denoiser to smooth the image. The projection step explores the prior information from the frequency domain and adds details to the spatial domain iteratively. The reconstruction quality is superior to the best MRI reconstruction methods under different sampling masks and rates. The stability, speed, and good reconstruction quality mean that our IDPCNN has the potential for widespread clinical applications.

An Investigation of 2D Spine Magnetic Resonance Imaging (MRI) with Compressed Sensing (CS).

To investigate the feasibility of compressed sensing MRI (CS-MRI) in the application of 2D spinal imaging and compare its performance with conventional MR imaging (non-CS-MRI). The CS imaging protocol was optimized on 5 volunteers. Non-CS-MRI and CS-MRI of 2D sagittal T1 weighted imaging (WI), Sag T2WI, and axial T2WI were performed for 71 patients (22 cervical, 8 thoracic, 41 lumbar MRI). Paired t tests were conducted to compare the total scan time. Three radiologists assessed image quality and lesion diagnosis independently. A Kendall W test was performed to assess interobserver agreement of the image quality scores and lesion diagnosis between readers. A nonparametric test (Wilcoxon test) was performed to compare the image quality. For lesion diagnosis, the interobserver and interstudy agreements were evaluated by kappa analysis. Paired t tests were conducted for SNR and CNR comparison. The mean scan time for spine CS-MRI (4min 28.7s ± 34.6s) was significantly shorter than that with non-CS-MRI (7min 21.3s ± 38.7s, t = - 47.464, P < 0.0001). CS-MRI achieved higher SNR and CNR than Non-CS-MRI in image quality assessment. Interobserver agreements of lesion diagnosis were excellent between non-CS-MRI and CS-MRI (kappa value from 0.913 to 1.000, P < 0.001). Interstudy agreements of lesion assessments were also excellent (kappa value = 1.000, with P < 0.001). CS-MRI spine imaging can significantly reduce the scan time, while maintaining comparable imaging quality to non-CS-MRI.

Adjustable shrinkage-thresholding projection algorithm for compressed sensing magnetic resonance imaging

Compressed sensing (CS) aims to reconstruct a high quality images with as little sample data as possible. Magnetic resonance imaging (MRI) plays an important role in medical imaging tools but has a slower data acquisition process. Applying CS to MRI offers significant scan time reductions. In this paper, we proposed a fast and efficient algorithm for compressed sensing magnetic resonance imaging (CS-MRI) reconstruction, denoted as adjustable shrinkage-thresholding projection algorithm (ASTP). It is designed to use adjustable shrinkage rules for lp-norm based CS-MRI model. This algorithm is established by using an iterative projection and acceleration scheme. In each iteration, the proposed adjustable shrinkage-thresholding rules are employed to ensure global convergence to accurate solution. Furthermore, the parameter p can be selected flexibly according to different practical application situations, and the orthogonal projection operation is used to reduce the dimension of the solution space to accelerate the convergence speed and improve the reconstruction quality. Numerical experiments show that proposed ASTP algorithm provides a higher accuracy, convergence speed and ability to suppress noise compared with some certain state-of-the-art algorithms.