Fast l₁-SPIRiT compressed sensing parallel imaging MRI: scalable parallel implementation and clinically feasible runtime

Knowledge-driven deep learning for fast MR imaging: Undersampled MR image reconstruction from supervised to un-supervised learning

Deep learning (DL) has emerged as a leading approach in accelerating MRI. It employs deep neural networks to extract knowledge from available datasets and then applies the trained networks to reconstruct accurate images from limited measurements. Unlike natural image restoration problems, MRI involves physics-based imaging processes, unique data properties, and diverse imaging tasks. This domain knowledge needs to be integrated with data-driven approaches. Our review will introduce the significant challenges faced by such knowledge-driven DL approaches in the context of fast MRI along with several notable solutions, which include learning neural networks and addressing different imaging application scenarios. The traits and trends of these techniques have also been given which have shifted from supervised learning to semi-supervised learning, and finally, to unsupervised learning methods. In addition, MR vendors' choices of DL reconstruction have been provided along with some discussions on open questions and future directions, which are critical for the reliable imaging systems.

Multi-Device Parallel MRI Reconstruction: Efficient Partitioning for Undersampled 5D Cardiac CINE

Cardiac CINE, a form of dynamic cardiac MRI, is indispensable in the diagnosis and treatment of heart conditions, offering detailed visualization essential for the early detection of cardiac diseases. As the demand for higher-resolution images increases, so does the volume of data requiring processing, presenting significant computational challenges that can impede the efficiency of diagnostic imaging. Our research presents an approach that takes advantage of the computational power of multiple Graphics Processing Units (GPUs) to address these challenges. GPUs are devices capable of performing large volumes of computations in a short period, and have significantly improved the cardiac MRI reconstruction process, allowing images to be produced faster. The innovation of our work resides in utilizing a multi-device system capable of processing the substantial data volumes demanded by high-resolution, five-dimensional cardiac MRI. This system surpasses the memory capacity limitations of single GPUs by partitioning large datasets into smaller, manageable segments for parallel processing, thereby preserving image integrity and accelerating reconstruction times. Utilizing OpenCL technology, our system offers adaptability and cross-platform functionality, ensuring wider applicability. The proposed multi-device approach offers an advancement in medical imaging, accelerating the reconstruction process and facilitating faster and more effective cardiac health assessment.

Coil sketching for computationally efficient MR iterative reconstruction

PURPOSE

Parallel imaging and compressed sensing reconstructions of large MRI datasets often have a prohibitive computational cost that bottlenecks clinical deployment, especially for three-dimensional (3D) non-Cartesian acquisitions. One common approach is to reduce the number of coil channels actively used during reconstruction as in coil compression. While effective for Cartesian imaging, coil compression inherently loses signal energy, producing shading artifacts that compromise image quality for 3D non-Cartesian imaging. We propose coil sketching, a general and versatile method for computationally-efficient iterative MR image reconstruction.

THEORY AND METHODS

We based our method on randomized sketching algorithms, a type of large-scale optimization algorithms well established in the fields of machine learning and big data analysis. We adapt the sketching theory to the MRI reconstruction problem via a structured sketching matrix that, similar to coil compression, considers high-energy virtual coils obtained from principal component analysis. But, unlike coil compression, it also considers random linear combinations of the remaining low-energy coils, effectively leveraging information from all coils.

RESULTS

First, we performed ablation experiments to validate the sketching matrix design on both Cartesian and non-Cartesian datasets. The resulting design yielded both improved computatioanal efficiency and preserved signal-to-noise ratio (SNR) as measured by the inverse g-factor. Then, we verified the efficacy of our approach on high-dimensional non-Cartesian 3D cones datasets, where coil sketching yielded up to three-fold faster reconstructions with equivalent image quality.

CONCLUSION

Coil sketching is a general and versatile reconstruction framework for computationally fast and memory-efficient reconstruction.

Parallel-stream fusion of scan-specific and scan-general priors for learning deep MRI reconstruction in low-data regimes

Magnetic resonance imaging (MRI) is an essential diagnostic tool that suffers from prolonged scan times. Reconstruction methods can alleviate this limitation by recovering clinically usable images from accelerated acquisitions. In particular, learning-based methods promise performance leaps by employing deep neural networks as data-driven priors. A powerful approach uses scan-specific (SS) priors that leverage information regarding the underlying physical signal model for reconstruction. SS priors are learned on each individual test scan without the need for a training dataset, albeit they suffer from computationally burdening inference with nonlinear networks. An alternative approach uses scan-general (SG) priors that instead leverage information regarding the latent features of MRI images for reconstruction. SG priors are frozen at test time for efficiency, albeit they require learning from a large training dataset. Here, we introduce a novel parallel-stream fusion model (PSFNet) that synergistically fuses SS and SG priors for performant MRI reconstruction in low-data regimes, while maintaining competitive inference times to SG methods. PSFNet implements its SG prior based on a nonlinear network, yet it forms its SS prior based on a linear network to maintain efficiency. A pervasive framework for combining multiple priors in MRI reconstruction is algorithmic unrolling that uses serially alternated projections, causing error propagation under low-data regimes. To alleviate error propagation, PSFNet combines its SS and SG priors via a novel parallel-stream architecture with learnable fusion parameters. Demonstrations are performed on multi-coil brain MRI for varying amounts of training data. PSFNet outperforms SG methods in low-data regimes, and surpasses SS methods with few tens of training samples. On average across tasks, PSFNet achieves 3.1 dB higher PSNR, 2.8% higher SSIM, and 0.3 × lower RMSE than baselines. Furthermore, in both supervised and unsupervised setups, PSFNet requires an order of magnitude lower samples compared to SG methods, and enables an order of magnitude faster inference compared to SS methods. Thus, the proposed model improves deep MRI reconstruction with elevated learning and computational efficiency.

Six-minute, in vivo MRI quantification of proximal femur trabecular bone 3D microstructure

BACKGROUND

Assessment of proximal femur trabecular bone microstructure in vivo by magnetic resonance imaging has recently been validated for acquiring information independent of bone mineral density in osteoporotic patients. However, the requisite signal-to-noise ratio (SNR) and resolution for interrogation of the trabecular microstructure at this anatomical location prolongs the scan duration and renders the imaging protocol clinically infeasible. Parallel imaging and compressed sensing (PICS) techniques can reduce the scan duration of the imaging protocol without substantially compromising image quality. The present work investigates the limits of acceleration for a commonly used PICS technique, ℓ1-ESPIRiT, for the purpose of quantifying measures of trabecular bone microarchitecture. Based on a desired error tolerance, a six-minute, prospectively accelerated variant of the imaging protocol was developed and assessed for intersession reproducibility and agreement with the longer reference scan.

PURPOSE

To investigate the limits of acceleration for MRI-based trabecular bone quantification by parallel imaging and compressed sensing reconstruction, and to develop a prototypical imaging protocol for assessing the proximal femur microstructure in a clinically practical scan time.

METHODS

Healthy participants (n = 11) were scanned by a 3D balanced steady-state free precession (bSSFP) sequence satisfying the Nyquist criterion with a scan duration of about 18 min. The raw data were retrospectively undersampled and reconstructed to mimic various acceleration factors ranging from 2 to 6. Trabecular volumes-of-interest in four major femoral regions (greater trochanter, intertrochanteric region, femoral neck, and femoral head) were analyzed and six relevant measures of trabecular bone microarchitecture (bone volume fraction, surface-to-curve ratio, erosion index, elastic modulus, trabecular thickness, plates-to-rods ratio) were obtained for images of all accelerations. To assess agreement, median percent error and intraclass correlation coefficients (ICCs) were computed using the fully-sampled data as reference. Based on this analysis, a prospectively 3-fold accelerated sequence with a duration of about 6 min was developed and the analysis was repeated.

RESULTS

A prospective acceleration factor of 3 demonstrated comparable performance in reproducibility and absolute agreement to the fully-sampled scan. The median CoV over all image-derived metrics was generally <6 % and ICCs >0.70. Also, measurements from prospectively 3-fold accelerated scans demonstrated in general median percent errors of <7 % and ICCs >0.70.

CONCLUSION

The present work proposes a method to make in vivo quantitative assessment of proximal femur trabecular microstructure with a clinically practical scan duration of about 6 min.

Reconstruction for 7T high-resolution whole-brain diffusion MRI using two-stage N/2 ghost correction and L1-SPIRiT without single-band reference

PURPOSE

To combine a new two-stage N/2 ghost correction and an adapted L1-SPIRiT method for reconstruction of 7T highly accelerated whole-brain diffusion MRI (dMRI) using only autocalibration scans (ACS) without the need of additional single-band reference (SBref) scans.

METHODS

The proposed ghost correction consisted of a 3-line reference approach in stage 1 and the reference-free entropy method in stage 2. The adapted L1-SPIRiT method was formulated within the 3D k-space framework. Its efficacy was examined by acquiring two dMRI data sets at 1.05-mm isotropic resolutions with a total acceleration of 6 or 9 (i.e., 2-fold or 3-fold slice and 3-fold in-plane acceleration). Diffusion analysis was performed to derive DTI metrics and estimate fiber orientation distribution functions (fODFs). The results were compared with those of 3D k-space GRAPPA using only ACS, all in reference to 3D k-space GRAPPA using both ACS and SBref (serving as a reference).

RESULTS

The proposed ghost correction eliminated artifacts more robustly than conventional approaches. Our adapted L1-SPIRiT method outperformed 3D k-space GRAPPA when using only ACS, improving image quality to what was achievable with 3D k-space GRAPPA using both ACS and SBref scans. The improvement in image quality further resulted in an improvement in estimation performances for DTI and fODFs.

CONCLUSION

The combination of our new ghost correction and adapted L1-SPIRiT method can reliably reconstruct 7T highly accelerated whole-brain dMRI without the need of SBref scans, increasing acquisition efficiency and reducing motion sensitivity.

LARO: Learned acquisition and reconstruction optimization to accelerate quantitative susceptibility mapping

Quantitative susceptibility mapping (QSM) involves acquisition and reconstruction of a series of images at multi-echo time points to estimate tissue field, which prolongs scan time and requires specific reconstruction technique. In this paper, we present our new framework, called Learned Acquisition and Reconstruction Optimization (LARO), which aims to accelerate the multi-echo gradient echo (mGRE) pulse sequence for QSM. Our approach involves optimizing a Cartesian multi-echo k-space sampling pattern with a deep reconstruction network. Next, this optimized sampling pattern was implemented in an mGRE sequence using Cartesian fan-beam k-space segmenting and ordering for prospective scans. Furthermore, we propose to insert a recurrent temporal feature fusion module into the reconstruction network to capture signal redundancies along echo time. Our ablation studies show that both the optimized sampling pattern and proposed reconstruction strategy help improve the quality of the multi-echo image reconstructions. Generalization experiments show that LARO is robust on the test data with new pathologies and different sequence parameters. Our code is available at https://github.com/Jinwei1209/LARO-QSM.git.

DSMENet: Detail and Structure Mutually Enhancing Network for under-sampled MRI reconstruction

Reconstructing zero-filled MR images (ZF) from partial k-space by convolutional neural networks (CNN) is an important way to accelerate MRI. However, due to the lack of attention to different components in ZF, it is challenging to learn the mapping from ZF to targets effectively. To ameliorate this issue, we propose a Detail and Structure Mutually Enhancing Network (DSMENet), which benefits from the complementary of the Structure Reconstruction UNet (SRUN) and the Detail Feature Refinement Module (DFRM). The SRUN learns structure-dominated information at multiple scales. And the DRFM enriches detail-dominated information from coarse to fine. The bidirectional alternate connections then exchange information between them. Moreover, the Detail Representation Construction Module (DRCM) extracts valuable initial detail representation for DFRM. And the Detail Guided Fusion Module (DGFM) facilitates the deep fusion of these complementary information. With the help of them, various components in ZF can be applied with discriminative attentions and mutually enhanced. In addition, the performance can be further improved by the Deep Enhanced Restoration (DER), a strategy based on recursion and constrain. Extensive experiments on fastMRI and CC-359 datasets demonstrate that DSMENet has robustness in terms of various body parts, under-sampling rates, and masks. Furthermore, DSMENet can achieve promising performance on qualitative and quantitative results, especially the competitive NMSE of 0.0268, PSNE of 33.7, and SSIM of 0.7808 on fastMRI 4 × single-coil knee leaderboard.

Multi-channel GAN-based calibration-free diffusion-weighted liver imaging with simultaneous coil sensitivity estimation and reconstruction

Introduction

Diffusion-weighted imaging (DWI) with parallel reconstruction may suffer from a mismatch between the coil calibration scan and imaging scan due to motions, especially for abdominal imaging.

Methods

This study aimed to construct an iterative multichannel generative adversarial network (iMCGAN)-based framework for simultaneous sensitivity map estimation and calibration-free image reconstruction. The study included 106 healthy volunteers and 10 patients with tumors.

Results

The performance of iMCGAN was evaluated in healthy participants and patients and compared with the SAKE, ALOHA-net, and DeepcomplexMRI reconstructions. The peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), root mean squared error (RMSE), and histograms of apparent diffusion coefficient (ADC) maps were calculated for assessing image qualities. The proposed iMCGAN outperformed the other methods in terms of the PSNR (iMCGAN: 41.82 ± 2.14; SAKE: 17.38 ± 1.78; ALOHA-net: 20.43 ± 2.11 and DeepcomplexMRI: 39.78 ± 2.78) for b = 800 DWI with an acceleration factor of 4. Besides, the ghosting artifacts in the SENSE due to the mismatch between the DW image and the sensitivity maps were avoided using the iMCGAN model.

Discussion

The current model iteratively refined the sensitivity maps and the reconstructed images without additional acquisitions. Thus, the quality of the reconstructed image was improved, and the aliasing artifact was alleviated when motions occurred during the imaging procedure.

Latest Advances in Image Acceleration: All Dimensions are Fair Game

Magnetic resonance imaging (MRI) is a versatile modality that can generate high-resolution images with a variety of tissue contrasts. However, MRI is a slow technique and requires long acquisition times, which increase with higher temporal and spatial resolution and/or when multiple contrasts and large volumetric coverage is required. In order to speedup MR data acquisition, several approaches have been introduced in the literature. Most of these techniques acquire less data than required and exploit intrinsic redundancies in the MR images to recover the information that was not sampled. This article presents a review of MR acquisition and reconstruction methods that have exploited redundancies in the temporal, spatial, and contrast/parametric dimensions to accelerate image data acquisition, focusing on cardiac and abdominal MR imaging applications. The review describes how each of these dimensions has been separately exploited for speeding up MR acquisition to then discuss more advanced techniques where multiple dimensions are exploited together for further reducing scan times. Finally, future directions for multidimensional image acceleration and remaining technical challenges are discussed. EVIDENCE LEVEL: 5 TECHNICAL EFFICACY: 1.

Wave-Encoded Model-Based Deep Learning for Highly Accelerated Imaging with Joint Reconstruction

A recently introduced model-based deep learning (MoDL) technique successfully incorporates convolutional neural network (CNN)-based regularizers into physics-based parallel imaging reconstruction using a small number of network parameters. Wave-controlled aliasing in parallel imaging (CAIPI) is an emerging parallel imaging method that accelerates imaging acquisition by employing sinusoidal gradients in the phase- and slice/partition-encoding directions during the readout to take better advantage of 3D coil sensitivity profiles. We propose wave-encoded MoDL (wave-MoDL) combining the wave-encoding strategy with unrolled network constraints for highly accelerated 3D imaging while enforcing data consistency. We extend wave-MoDL to reconstruct multicontrast data with CAIPI sampling patterns to leverage similarity between multiple images to improve the reconstruction quality. We further exploit this to enable rapid quantitative imaging using an interleaved look-locker acquisition sequence with T₂ preparation pulse (3D-QALAS). Wave-MoDL enables a 40 s MPRAGE acquisition at 1 mm resolution at 16-fold acceleration. For quantitative imaging, wave-MoDL permits a 1:50 min acquisition for T₁, T₂, and proton density mapping at 1 mm resolution at 12-fold acceleration, from which contrast-weighted images can be synthesized as well. In conclusion, wave-MoDL allows rapid MR acquisition and high-fidelity image reconstruction and may facilitate clinical and neuroscientific applications by incorporating unrolled neural networks into wave-CAIPI reconstruction.

Improving high frequency image features of deep learning reconstructions via k-space refinement with null-space kernel

PURPOSE

Deep learning (DL) based reconstruction using unrolled neural networks has shown great potential in accelerating MRI. However, one of the major drawbacks is the loss of high-frequency details and textures in the output. The purpose of the study is to propose a novel refinement method that uses null-space kernel to refine k-space and improve blurred image details and textures.

METHODS

The proposed method constrains the output of the DL to comply to the linear neighborhood relationship calibrated in the auto-calibration lines. To demonstrate efficacy, we tested our refinement method on the DL reconstruction under a variety of conditions (i.e., dataset, unrolled neural networks, and under-sampling scheme). Specifically, the method was tested on three large-scale public datasets (knee and brain) from fastMRI's multi-coil track.

RESULTS

The proposed scheme visually reduces the structural error in the k-space domain, enhance the homogeneity of the k-space intensity. Consequently, reconstructed image shows sharper images with enhanced details and textures. The proposed method is also successful in improving high-frequency image details (SSIM, GMSD) without sacrificing overall image error (PSNR).

CONCLUSION

Our findings imply that refining DL output using the proposed method may generally improve DL reconstruction as tested with various large-scale dataset and networks.

Dual-Domain Reconstruction Network with V-Net and K-Net for Fast MRI

PURPOSE

To introduce a dual-domain reconstruction network with V-Net and K-Net for accurate MR image reconstruction from undersampled k-space data.

METHODS

Most state-of-the-art reconstruction methods apply U-Net or cascaded U-Nets in the image domain and/or k-space domain. Nevertheless, these methods have the following problems: (1) directly applying U-Net in the k-space domain is not optimal for extracting features; (2) classical image-domain-oriented U-Net is heavyweighted and hence inefficient when cascaded many times to yield good reconstruction accuracy; (3) classical image-domain-oriented U-Net does not make full use of information of the encoder network for extracting features in the decoder network; and (4) existing methods are ineffective in simultaneously extracting and fusing features in the image domain and its dual k-space domain. To tackle these problems, we present 3 different methods: (1) V-Net, an image-domain encoder-decoder subnetwork that is more lightweight for cascading and effective in fully utilizing features in the encoder for decoding; (2) K-Net, a k-space domain subnetwork that is more suitable for extracting hierarchical features in the k-space domain, and (3) KV-Net, a dual-domain reconstruction network in which V-Nets and K-Nets are effectively combined and cascaded.

RESULTS

Extensive experimental results on the fastMRI dataset demonstrate that the proposed KV-Net can reconstruct high-quality images and outperform state-of-the-art approaches with fewer parameters.

CONCLUSIONS

To reconstruct images effectively and efficiently from incomplete k-space data, we have presented a dual-domain KV-Net to combine K-Nets and V-Nets. The KV-Net achieves better results with 9% and 5% parameters than comparable methods (XPD-Net and i-RIM).

Compressed sensing based dynamic MR image reconstruction by using 3D-total generalized variation and tensor decomposition: k-t TGV-TD

PURPOSE

Compressed Sensing Magnetic Resonance Imaging (CS-MRI) is a promising technique to accelerate dynamic cardiac MR imaging (DCMRI). For DCMRI, the CS-MRI usually exploits image signal sparsity and low-rank property to reconstruct dynamic images from the undersampled k-space data. In this paper, a novel CS algorithm is investigated to improve dynamic cardiac MR image reconstruction quality under the condition of minimizing the k-space recording.

METHODS

The sparse representation of 3D cardiac magnetic resonance data is implemented by synergistically integrating 3D total generalized variation (3D-TGV) algorithm and high order singular value decomposition (HOSVD) based Tensor Decomposition, termed k-t TGV-TD method. In the proposed method, the low rank structure of the 3D dynamic cardiac MR data is performed with the HOSVD method, and the localized image sparsity is achieved by the 3D-TGV method. Moreover, the Fast Composite Splitting Algorithm (FCSA) method, combining the variable splitting with operator splitting techniques, is employed to solve the low-rank and sparse problem. Two different cardiac MR datasets (cardiac perfusion and cine MR datasets) are used to evaluate the performance of the proposed method.

RESULTS

Compared with the state-of-art methods, such as k-t SLR, 3D-TGV, HOSVD based tensor decomposition and low-rank plus sparse method, the proposed k-t TGV-TD method can offer improved reconstruction accuracy in terms of higher peak SNR (PSNR) and structural similarity index (SSIM). The proposed k-t TGV-TD method can achieve significantly better and stable reconstruction results than state-of-the-art methods in terms of both PSNR and SSIM, especially for cardiac perfusion MR dataset.

CONCLUSIONS

This work proved that the k-t TGV-TD method was an effective sparse representation way for DCMRI, which was capable of significantly improving the reconstruction accuracy with different acceleration factors.

A Slice-Low-Rank Plus Sparse (slice-L + S) Reconstruction Method for k-t Undersampled Multiband First-Pass Myocardial Perfusion MRI

Purpose

The synergistic use of k‐t undersampling and multiband (MB) imaging has the potential to provide extended slice coverage and high spatial resolution for first‐pass perfusion MRI. The low‐rank plus sparse (L + S) model has shown excellent performance for accelerating single‐band (SB) perfusion MRI.

Methods

A MB data consistency method employing ESPIRiT maps and through‐plane coil information was developed. This data consistency method was combined with the temporal L + S constraint to form the slice‐L + S method. Slice‐L + S was compared to SB L + S and the sequential operations of split slice‐GRAPPA and SB L + S (seq‐SG‐L + S) using synthetic data formed from multislice SB images. Prospectively k‐t undersampled MB data were also acquired and reconstructed using seq‐SG‐L + S and slice‐L + S.

Results

Using synthetic data with total acceleration rates of 6–12, slice‐L + S outperformed SB L + S and seq‐SG‐L + S (N = 7 subjects) with respect to normalized RMSE and the structural similarity index (P < 0.05 for both). For the specific case with MB factor = 3 and rate 3 undersampling, or for SB imaging with rate 9 undersampling (N = 7 subjects), the normalized RMSE values were 0.037 ± 0.007, 0.042 ± 0.005, and 0.031 ± 0.004; and the structural similarity index values were 0.88 ± 0.03, 0.85 ± 0.03, and 0.89 ± 0.02 for SB L + S, seq‐SG‐L + S, and slice‐L + S, respectively (P < 0.05 for both). For prospectively undersampled MB data, slice‐L + S provided better image quality than seq‐SG‐L + S for rate 6 (N = 7) and rate 9 acceleration (N = 7) as scored by blinded experts.

Conclusion

Slice‐L + S outperformed SB‐L + S and seq‐SG‐L + S and provides 9 slice coverage of the left ventricle with a spatial resolution of 1.5 mm × 1.5 mm with good image quality.

Temporospatial characterization of ventricular wall motion with real-time cardiac magnetic resonance imaging in health and disease

Cardiac magnetic resonance imaging (MRI) has been largely dependent on retrospective cine for data acquisition. Real-time imaging, although inferior in image quality to retrospective cine, is more informative about motion dynamics. We herein developed a real-time cardiac MRI approach to temporospatial characterization of left ventricle (LV) and right ventricle (RV) wall motion. This approach provided two temporospatial indices, temporal periodicity and spatial coherence, for quantitative assessment of ventricular function. In a cardiac MRI study, we prospectively investigated temporospatial characterization in reference to standard volumetric measurements with retrospective cine. The temporospatial indices were found to be effective for evaluating the difference of ventricular performance between the healthy volunteers and the heart failure (HF) patients (LV temporal periodicity 0.24 ± 0.037 vs. 0.14 ± 0.021; RV temporal periodicity 0.18 ± 0.030 vs. 0.10 ± 0.014; LV spatial coherence 0.52 ± 0.039 vs. 0.38 ± 0.040; RV spatial coherence 0.50 ± 0.036 vs. 0.35 ± 0.035; all in arbitrary unit). The HF patients and healthy volunteers were well differentiated in the scatter plots of spatial coherence and temporal periodicity while they were mixed in those of end-systolic volume (ESV) and ejection fraction (EF) from volumetric measurements. This study demonstrated the potential of real-time cardiac MRI for intricate analysis of ventricular function beyond retrospective cine.

SDnDTI: Self-supervised deep learning-based denoising for diffusion tensor MRI

Diffusion tensor magnetic resonance imaging (DTI) is a widely adopted neuroimaging method for the in vivo mapping of brain tissue microstructure and white matter tracts. Nonetheless, the noise in the diffusion-weighted images (DWIs) decreases the accuracy and precision of DTI derived microstructural parameters and leads to prolonged acquisition time for achieving improved signal-to-noise ratio (SNR). Deep learning-based image denoising using convolutional neural networks (CNNs) has superior performance but often requires additional high-SNR data for supervising the training of CNNs, which reduces the feasibility of supervised learning-based denoising in practice. In this work, we develop a self-supervised deep learning-based method entitled "SDnDTI" for denoising DTI data, which does not require additional high-SNR data for training. Specifically, SDnDTI divides multi-directional DTI data into many subsets of six DWI volumes and transforms DWIs from each subset to along the same diffusion-encoding directions through the diffusion tensor model, generating multiple repetitions of DWIs with identical image contrasts but different noise observations. SDnDTI removes noise by first denoising each repetition of DWIs using a deep 3-dimensional CNN with the average of all repetitions with higher SNR as the training target, following the same approach as normal supervised learning based denoising methods, and then averaging CNN-denoised images for achieving higher SNR. The denoising efficacy of SDnDTI is demonstrated in terms of the similarity of output images and resultant DTI metrics compared to the ground truth generated using substantially more DWI volumes on two datasets with different spatial resolutions, b-values and numbers of input DWI volumes provided by the Human Connectome Project (HCP) and the Lifespan HCP in Aging. The SDnDTI results preserve image sharpness and textural details and substantially improve upon those from the raw data. The results of SDnDTI are comparable to those from supervised learning-based denoising and outperform those from state-of-the-art conventional denoising algorithms including BM4D, AONLM and MPPCA. By leveraging domain knowledge of diffusion MRI physics, SDnDTI makes it easier to use CNN-based denoising methods in practice and has the potential to benefit a wider range of research and clinical applications that require accelerated DTI acquisition and high-quality DTI data for mapping of tissue microstructure, fiber tracts and structural connectivity in the living human brain.

Automated Parameter Selection for Accelerated MRI Reconstruction via Low-Rank Modeling of Local k-Space Neighborhoods

PURPOSE

Image quality in accelerated MRI rests on careful selection of various reconstruction parameters. A common yet tedious and error-prone practice is to hand-tune each parameter to attain visually appealing reconstructions. Here, we propose a parameter tuning strategy to automate hybrid parallel imaging (PI) - compressed sensing (CS) reconstructions via low-rank modeling of local k-space neighborhoods (LORAKS) supplemented with sparsity regularization in wavelet and total variation (TV) domains.

METHODS

For low-rank regularization, we leverage a soft-thresholding operation based on singular values for matrix rank selection in LORAKS. For sparsity regularization, we employ Stein's unbiased risk estimate criterion to select the wavelet regularization parameter and local standard deviation of reconstructions to select the TV regularization parameter. Comprehensive demonstrations are presented on a numerical brain phantom and in vivo brain and knee acquisitions. Quantitative assessments are performed via PSNR, SSIM and NMSE metrics.

RESULTS

The proposed hybrid PI-CS method improves reconstruction quality compared to PI-only techniques, and it achieves on par image quality to reconstructions with brute-force optimization of reconstruction parameters. These results are prominent across several different datasets and the range of examined acceleration rates.

CONCLUSION

A data-driven parameter tuning strategy to automate hybrid PI-CS reconstructions is presented. The proposed method achieves reliable reconstructions of accelerated multi-coil MRI datasets without the need for exhaustive hand-tuning of reconstruction parameters.

Alternating Learning Approach for Variational Networks and Undersampling Pattern in Parallel MRI Applications

This work proposes an alternating learning approach to learn the sampling pattern (SP) and the parameters of variational networks (VN) in accelerated parallel magnetic resonance imaging (MRI). We investigate four variations of the learning approach, that alternates between improving the SP, using bias-accelerated subset selection, and improving parameters of the VN, using ADAM. The variations include the use of monotone or non-monotone alternating steps and systematic reduction of learning rates. The algorithms learn an effective pair to be used in future scans, including an SP that captures fewer k-space samples in which the generated undersampling artifacts are removed by the VN reconstruction. The quality of the VNs and SPs obtained by the proposed approaches is compared against different methods, including other kinds of joint learning methods and state-of-art reconstructions, on two different datasets at various acceleration factors (AF). We observed improvements visually and in three different figures of merit commonly used in deep learning (RMSE, SSIM, and HFEN) on AFs from 2 to 20 with brain and knee joint datasets when compared to the other approaches. The improvements ranged from 1% to 62% over the next best approach tested with VNs. The proposed approach has shown stable performance, obtaining similar learned SPs under different initial training conditions. We observe that the improvement is not only due to the learned sampling density, it is also due to the learned position of samples in k-space. The proposed approach was able to learn effective pairs of SPs and reconstruction VNs, improving 3D Cartesian accelerated parallel MRI applications.

Reduced Acquisition Time per Bed Position for PET/MRI Using 68Ga-RM2 or 68Ga-PSMA-11 in Patients With Prostate Cancer: A Retrospective Analysis

BACKGROUND. Growing clinical adoption of PET/MRI for prostate cancer (PC) evaluation has increased interest in reducing PET/MRI scanning times. Reducing acquisition time per bed position below current times of at least 5 minutes would allow shorter examination lengths. OBJECTIVE. The purpose of this study was to evaluate the effect of different reduced PET acquisition times in patients with PC who underwent ⁶⁸Ga-PSMA-11 or ⁶⁸Ga-RM2 PET/MRI using highly sensitive silicon photomultiplier-based PET detectors. METHODS. This study involved retrospective review of men with PC who underwent PET/MRI as part of one of two prospective trials. Fifty men (mean [± SD] age, 69.9 ± 6.8 years) who underwent ⁶⁸Ga-RM2 PET/MRI and 50 men (mean age, 66.6 ± 5.7 years) who underwent ⁶⁸Ga-PSMA-11 PET/MRI were included. PET/MRI used a time-of-flight-enabled system with silicon photomultiplier-based detectors. The acquisition time was 4 minutes per bed position. PET data were reconstructed using acquisition times of 30 seconds, 1 minute, 2 minutes, 3 minutes, and 4 minutes. Three readers independently assessed image quality for each reconstruction using a 5-point Likert scale (with 1 denoting nondiagnostic and 5 indicating excellent quality). One reader measured SUV_max for up to six lesions per patient. Two readers independently assessed lesion conspicuity using a a 3-point Likert scale (with 1 indicating that lesions were not visualized and 3 denoting that they were definitely visualized). RESULTS. Mean image quality across readers at 30 seconds, 1 minutes, 2 minutes, 3 minutes, and 4 minutes was, for ⁶⁸Ga-RM2 PET/MRI, from 1.0 ± 0.2 to 1.7 ± 0.7, 2.0 ± 0.3 to 2.6 ± 0.8, 3.1 ± 0.5 to 3.9 ± 0.8, 4.6 ± 0.6 to 4.7 ± 0.6, and 4.8 ± 0.4 to 4.8 ± 0.5, respectively, and for ⁶⁸Ga-PSMA-11 PET/MRI it was from 1.2 ± 0.4 to 1.8 ± 0.6, 2.2 ± 0.4 to 2.8 ± 0.7, 3.6 ± 0.6 to 4.1± 0.8, 4.8 ± 0.4 to 4.9 ± 0.4, and 4.9 ± 0.3 to 5.0 ± 0.2, respectively. The mean lesion SUV_max for ⁶⁸Ga-RM2 PET/MRI was 11.1 ± 12.4, 10.2 ± 11.7, 9.6 ± 11.3, 9.5 ± 11.6, and 9.4 ± 11.6, respectively, and for ⁶⁸Ga-PSMA-11 PET/MRI it was 14.7 ± 8.2, 12.9 ± 7.4, 12.1 ± 7.8, 11.7 ± 7.9, and 11.6 ± 7.9, respectively. Mean lesion conspicuity (reader 1/reader 2) was, for ⁶⁸Ga-RM2 PET/MRI, 2.4 ± 0.5/2.7 ± 0.5, 2.9 ± 0.3/2.9 ± 0.3, 3.0 ± 0.0/3.0 ± 0.0, 3.0 ± 0.0/3.0 ± 0.0, and 3.0 ± 0.0/3.0 ± 0.0, respectively, and for ⁶⁸Ga-PSMA-11 PET/MRI it was 2.6 ± 0.5/2.8 ± 0.4, 3.0 ± 0.2/2.9 ± 0.3, 3.0 ± 0.1/3.0 ± 0.2, 3.0 ± 0.0/3.0 ± 0.0, and 3.0 ± 0.0/3.0 ± 0.0, respectively. CONCLUSION. Our data support routine 3-minute acquisitions, which provided results very similar to those for 4-minute acquisitions. Two-minute acquisitions, although they lowered quality somewhat, provided acceptable performance and warrant consideration. CLINICAL IMPACT. When PC is evaluated using modern PET/MRI equipment, time per bed position may be reduced compared with historically used times. TRIAL REGISTRATION. ClinicalTrials.gov NCT02624518 and NCT02678351.

Joint reconstruction framework of compressed sensing and nonlinear parallel imaging for dynamic cardiac magnetic resonance imaging

Compressed Sensing (CS) and parallel imaging are two promising techniques that accelerate the MRI acquisition process. Combining these two techniques is of great interest due to the complementary information used in each. In this study, we proposed a novel reconstruction framework that effectively combined compressed sensing and nonlinear parallel imaging technique for dynamic cardiac imaging. Specifically, the proposed method decouples the reconstruction process into two sequential steps: In the first step, a series of aliased dynamic images were reconstructed from the highly undersampled k-space data using compressed sensing; In the second step, nonlinear parallel imaging technique, i.e. nonlinear GRAPPA, was utilized to reconstruct the original dynamic images from the reconstructed k-space data obtained from the first step. In addition, we also proposed a tailored k-space down-sampling scheme that satisfies both the incoherent undersampling requirement for CS and the structured undersampling requirement for nonlinear parallel imaging. The proposed method was validated using four in vivo experiments of dynamic cardiac cine MRI with retrospective undersampling. Experimental results showed that the proposed method is superior at reducing aliasing artifacts and preserving the spatial details and temporal variations, compared with the competing k-t FOCUSS and k-t FOCUSS with sensitivity encoding methods, with the same numbers of measurements.

Deep learning based multiplexed sensitivity-encoding (DL-MUSE) for high-resolution multi-shot DWI

PURPOSE

A phase correction method for high-resolution multi-shot (MSH) diffusion weighted imaging (DWI) is proposed. The efficacy and generalization capability of the method were validated on both healthy volunteers and patients.

THEORY AND METHODS

Conventionally, inter-shot phase variations for MSH echo-planar imaging (EPI) DWI are corrected by model-based algorithms. However, many acquisition imperfections are hard to measure accurately for conventional model-based methods, making the phase estimation and artifacts suppression unreliable. We propose a deep learning multiplexed sensitivity-encoding (DL-MUSE) framework to improve the phase estimations based on convolutional neural network (CNN) reconstruction. Aliasing-free single-shot (SSH) DW images, which have been used routinely in clinical settings, were used for training before the aliasing correction of MSH-DWI images. A dual-channel U-net comprising multiple convolutional layers was used for the phase estimation of MSH-EPI. The network was trained on a dataset containing 30 healthy volunteers and tested on another dataset of 52 healthy subjects and 15 patients with lesions or tumors with different shot numbers (4, 6 and 8). To further validate the generalization capability of our network, we acquired a dataset with different numbers of shots, TEs, partial Fourier factors, resolutions, ETLs, FOVs, coil numbers, and image orientations from two sites. We also compared the reconstruction performance of our proposed method with that of the conventional MUSE and SSH-EPI qualitatively and quantitatively.

RESULTS

Our results show that DL-MUSE is capable of correcting inter-shot phase errors with high and robust performance. Compared to conventional model-based MUSE, our method, by applying deep learning-based phase corrections, showed reduced distortion, noise level, and signal loss in high b-value DWIs. The improvements of image quality become more evident as the shot number increases from 4 to 8, especially in those central regions of the images, where g-factor artifacts are severe. Furthermore, the proposed method could provide the information about the orientation of the white matter with better consistency and achieve finer fibers delineation compared to the SSH-EPI method. Besides, the experiments on volunteers and patients from two different sites demonstrated the generalizability of our proposed method preliminarily.

CONCLUSION

A deep learning-based reconstruction algorithm for MSH-EPI images, which helps improve image quality greatly, was proposed. Results from healthy volunteers and tumor patients demonstrated the feasibility and generalization performances of our method for high-resolution MSH-EPI DWI, which can be used for routine clinical applications as well as neuroimaging research.

Aliasing-free reduced field-of-view parallel imaging

PURPOSE

To reconstruct aliasing-free full field-of-view (FOV) images for reduced FOV (rFOV) parallel imaging (PI) with Cartesian and Wave sampling, which suffers from aliasing artifacts using existing PI methods.

THEORY AND METHODS

The sensitivity encoding method (SENSE) was extended to the Soft-SENSE models supporting multiple-set coil sensitivity maps (CSM) and point spread functions (PSF) for Cartesian and Wave sampled rFOV PI, respectively. The multiple-set CSM and PSF were created from full FOV CSM and PSF according to the image folding process induced by rFOV sampling. The Soft-SENSE reconstructions could be solved by the same algorithms for the conventional full FOV SENSE reconstruction.

RESULTS

Soft-SENSE using multiple-set full FOV CSM and PSF successfully reconstruct aliasing-free full FOV image from rFOV PI data with Cartesian and Wave sampling. The proposed rFOV PI enables flexible control of the aliasing and achieves comparable geometry factors as the standard full FOV PI with the same net acceleration factor. Reduced FOV PI improves the computational efficiency of iterative compressed sensing (CS) and PI reconstruction, especially for high-resolution volumetric imaging, thanks to the reduced fast Fourier transform (FFT) size. Moreover, rFOV PI reconstruction provides a potential alternative to the phase oversampling for the FOV aliasing problem.

CONCLUSION

The proposed Soft-SENSE using full FOV CSM and PSF could reconstruct aliasing-free full FOV image for rFOV PI, and make it a viable solution enabling more flexible PI acceleration and effectively improving the computational efficiency of iterative CSPI reconstruction.

Learning a preconditioner to accelerate compressed sensing reconstructions in MRI

Purpose

To learn a preconditioner that accelerates parallel imaging (PI) and compressed sensing (CS) reconstructions.

Methods

A convolutional neural network (CNN) with residual connections was used to train a preconditioning operator. Training and validation data were simulated using 50% brain images and 50% white Gaussian noise images. Each multichannel training example contains a simulated sampling mask, complex coil sensitivity maps, and two regularization parameter maps. The trained model was integrated in the preconditioned conjugate gradient (PCG) method as part of the split Bregman CS method. The acceleration performance was compared with that of a circulant PI‐CS preconditioner for varying undersampling factors, number of coil elements and anatomies.

Results

The learned preconditioner reduces the number of PCG iterations by a factor of 4, yielding a similar acceleration as an efficient circulant preconditioner. The method generalizes well to different sampling schemes, coil configurations and anatomies.

Conclusion

It is possible to learn adaptable preconditioners for PI and CS reconstructions that meet the performance of state‐of‐the‐art preconditioners. Further acceleration could be achieved by optimizing the network architecture and the training set. Such a preconditioner could also be integrated in fully learned reconstruction methods to accelerate the training process of unrolled networks.

Learning Data Consistency and its Application to Dynamic MR Imaging

Magnetic resonance (MR) image reconstruction from undersampled k-space data can be formulated as a minimization problem involving data consistency and image prior. Existing deep learning (DL)-based methods for MR reconstruction employ deep networks to exploit the prior information and integrate the prior knowledge into the reconstruction under the explicit constraint of data consistency, without considering the real distribution of the noise. In this work, we propose a new DL-based approach termed Learned DC that implicitly learns the data consistency with deep networks, corresponding to the actual probability distribution of system noise. The data consistency term and the prior knowledge are both embedded in the weights of the networks, which provides an utterly implicit manner of learning reconstruction model. We evaluated the proposed approach with highly undersampled dynamic data, including the dynamic cardiac cine data with up to 24-fold acceleration and dynamic rectum data with the acceleration factor equal to the number of phases. Experimental results demonstrate the superior performance of the Learned DC both quantitatively and qualitatively than the state-of-the-art.

Applying Deep Learning to Accelerated Clinical Brain Magnetic Resonance Imaging for Multiple Sclerosis

Background: Magnetic resonance (MR) scans are routine clinical procedures for monitoring people with multiple sclerosis (PwMS). Patient discomfort, timely scheduling, and financial burden motivate the need to accelerate MR scan time. We examined the clinical application of a deep learning (DL) model in restoring the image quality of accelerated routine clinical brain MR scans for PwMS.

Methods: We acquired fast 3D T1w BRAVO and fast 3D T2w FLAIR MRI sequences (half the phase encodes and half the number of slices) in parallel to conventional parameters. Using a subset of the scans, we trained a DL model to generate images from fast scans with quality similar to the conventional scans and then applied the model to the remaining scans. We calculated clinically relevant T1w volumetrics (normalized whole brain, thalamic, gray matter, and white matter volume) for all scans and T2 lesion volume in a sub-analysis. We performed paired t-tests comparing conventional, fast, and fast with DL for these volumetrics, and fit repeated measures mixed-effects models to test for differences in correlations between volumetrics and clinically relevant patient-reported outcomes (PRO).

Results: We found statistically significant but small differences between conventional and fast scans with DL for all T1w volumetrics. There was no difference in the extent to which the key T1w volumetrics correlated with clinically relevant PROs of MS symptom burden and neurological disability.

Conclusion: A deep learning model that improves the image quality of the accelerated routine clinical brain MR scans has the potential to inform clinically relevant outcomes in MS.

Utilizing the Wavelet Transform's Structure in Compressed Sensing

Compressed sensing has empowered quality image reconstruction with fewer data samples than previously thought possible. These techniques rely on a sparsifying linear transformation. The Daubechies wavelet transform is commonly used for this purpose. In this work, we take advantage of the structure of this wavelet transform and identify an affine transformation that increases the sparsity of the result. After inclusion of this affine transformation, we modify the resulting optimization problem to comply with the form of the Basis Pursuit Denoising problem. Finally, we show theoretically that this yields a lower bound on the error of the reconstruction and present results where solving this modified problem yields images of higher quality for the same sampling patterns using both magnetic resonance and optical images.

Achieving high-resolution 1H-MRSI of the human brain with compressed-sensing and low-rank reconstruction at 7 Tesla

Low sensitivity MR techniques such as magnetic resonance spectroscopic imaging (MRSI) greatly benefit from the gain in signal-to-noise provided by ultra-high field MR. High-resolution and whole-slab brain MRSI remains however very challenging due to lengthy acquisition, low signal, lipid contamination and field inhomogeneity. In this study, we propose an acquisition-reconstruction scheme that combines 1H free-induction-decay (FID)-MRSI sequence, short TR acquisition, compressed sensing acceleration and low-rank modeling with total-generalized-variation constraint to achieve metabolite imaging in two and three dimensions at 7 Tesla. The resulting images and volumes reveal highly detailed distributions that are specific to each metabolite and follow the underlying brain anatomy. The MRSI method was validated in a high-resolution phantom containing fine metabolite structures, and in five healthy volunteers. This new application of compressed sensing acceleration paves the way for high-resolution MRSI in clinical setting with acquisition times of 5 min for 2D MRSI at 2.5 mm and of 20 min for 3D MRSI at 3.3 mm isotropic.

Fast data-driven learning of parallel MRI sampling patterns for large scale problems

In this study, a fast data-driven optimization approach, named bias-accelerated subset selection (BASS), is proposed for learning efficacious sampling patterns (SPs) with the purpose of reducing scan time in large-dimensional parallel MRI. BASS is applicable when Cartesian fully-sampled k-space measurements of specific anatomy are available for training and the reconstruction method for undersampled measurements is specified; such information is used to define the efficacy of any SP for recovering the values at the non-sampled k-space points. BASS produces a sequence of SPs with the aim of finding one of a specified size with (near) optimal efficacy. BASS was tested with five reconstruction methods for parallel MRI based on low-rankness and sparsity that allow a free choice of the SP. Three datasets were used for testing, two of high-resolution brain images ([Formula: see text]-weighted images and, respectively, [Formula: see text]-weighted images) and another of knee images for quantitative mapping of the cartilage. The proposed approach has low computational cost and fast convergence; in the tested cases it obtained SPs up to 50 times faster than the currently best greedy approach. Reconstruction quality increased by up to 45% over that provided by variable density and Poisson disk SPs, for the same scan time. Optionally, the scan time can be nearly halved without loss of reconstruction quality. Quantitative MRI and prospective accelerated MRI results show improvements. Compared with greedy approaches, BASS rapidly learns effective SPs for various reconstruction methods, using larger SPs and larger datasets; enabling better selection of sampling-reconstruction pairs for specific MRI problems.

A deep cascade of ensemble of dual domain networks with gradient-based T1 assistance and perceptual refinement for fast MRI reconstruction

Deep learning networks have shown promising results in fast magnetic resonance imaging (MRI) reconstruction. In our work, we develop deep networks to further improve the quantitative and the perceptual quality of reconstruction. To begin with, we propose reconsynergynet (RSN), a network that combines the complementary benefits of independently operating on both the image and the Fourier domain. For a single-coil acquisition, we introduce deep cascade RSN (DC-RSN), a cascade of RSN blocks interleaved with data fidelity (DF) units. Secondly, we improve the structure recovery of DC-RSN for T2 weighted Imaging (T2WI) through assistance of T1 weighted imaging (T1WI), a sequence with short acquisition time. T1 assistance is provided to DC-RSN through a gradient of log feature (GOLF) fusion. Furthermore, we propose perceptual refinement network (PRN) to refine the reconstructions for better visual information fidelity (VIF), a metric highly correlated to radiologist's opinion on the image quality. Lastly, for multi-coil acquisition, we propose variable splitting RSN (VS-RSN), a deep cascade of blocks, each block containing RSN, multi-coil DF unit, and a weighted average module. We extensively validate our models DC-RSN and VS-RSN for single-coil and multi-coil acquisitions and report the state-of-the-art performance. We obtain a SSIM of 0.768, 0.923, and 0.878 for knee single-coil-4x, multi-coil-4x, and multi-coil-8x in fastMRI, respectively. We also conduct experiments to demonstrate the efficacy of GOLF based T1 assistance and PRN.

Improved cortical surface reconstruction using sub-millimeter resolution MPRAGE by image denoising

Automatic cerebral cortical surface reconstruction is a useful tool for cortical anatomy quantification, analysis and visualization. Recently, the Human Connectome Project and several studies have shown the advantages of using T₁-weighted magnetic resonance (MR) images with sub-millimeter isotropic spatial resolution instead of the standard 1-mm isotropic resolution for improved accuracy of cortical surface positioning and thickness estimation. Nonetheless, sub-millimeter resolution images are noisy by nature and require averaging multiple repetitions to increase the signal-to-noise ratio for precisely delineating the cortical boundary. The prolonged acquisition time and potential motion artifacts pose significant barriers to the wide adoption of cortical surface reconstruction at sub-millimeter resolution for a broad range of neuroscientific and clinical applications. We address this challenge by evaluating the cortical surface reconstruction resulting from denoised single-repetition sub-millimeter T₁-weighted images. We systematically characterized the effects of image denoising on empirical data acquired at 0.6 mm isotropic resolution using three classical denoising methods, including denoising convolutional neural network (DnCNN), block-matching and 4-dimensional filtering (BM4D) and adaptive optimized non-local means (AONLM). The denoised single-repetition images were found to be highly similar to 6-repetition averaged images, with a low whole-brain averaged mean absolute difference of ~0.016, high whole-brain averaged peak signal-to-noise ratio of ~33.5 dB and structural similarity index of ~0.92, and minimal gray matter–white matter contrast loss (2% to 9%). The whole-brain mean absolute discrepancies in gray matter–white matter surface placement, gray matter–cerebrospinal fluid surface placement and cortical thickness estimation were lower than 165 μm, 155 μm and 145 μm—sufficiently accurate for most applications. These discrepancies were approximately one third to half of those from 1-mm isotropic resolution data. The denoising performance was equivalent to averaging ~2.5 repetitions of the data in terms of image similarity, and 1.6–2.2 repetitions in terms of the cortical surface placement accuracy. The scan-rescan variability of the cortical surface positioning and thickness estimation was lower than 170 μm. Our unique dataset and systematic characterization support the use of denoising methods for improved cortical surface reconstruction at sub-millimeter resolution.

Fully Balanced SSFP Without an Endorectal Coil for Postimplant QA of MRI-Assisted Radiosurgery (MARS) of Prostate Cancer: A Prospective Study

PURPOSE

To investigate fully balanced steady-state free precession (bSSFP) with optimized acquisition protocols for magnetic resonance imaging (MRI)-based postimplant quality assessment of low-dose-rate (LDR) prostate brachytherapy without an endorectal coil (ERC).

METHODS AND MATERIALS

Seventeen patients at a major academic cancer center who underwent MRI-assisted radiosurgery (MARS) LDR prostate cancer brachytherapy were imaged with moderate, high, or very high spatial resolution fully bSSFP MRIs without using an ERC. Between 1 and 3 signal averages (NEX) were acquired with acceleration factors (R) between 1 and 2, with the goal of keeping scan times between 4 and 6 minutes. Acquisitions with R >1 were reconstructed with parallel imaging and compressed sensing (PICS) algorithms. Radioactive seeds were identified by 3 medical dosimetrists. Additionally, some of the MRI techniques were implemented and tested at a community hospital; 3 patients underwent MARS LDR prostate brachytherapy and were imaged without an ERC.

RESULTS

Increasing the in-plane spatial resolution mitigated partial volume artifacts and improved overall seed and seed marker visualization at the expense of reduced signal-to-noise ratio (SNR). The reduced SNR as a result of imaging at higher spatial resolution and without an ERC was partially compensated for by the multi-NEX acquisitions enabled by PICS. Resultant image quality was superior to the current clinical standard. All 3 dosimetrists achieved near-perfect precision and recall for seed identification in the 17 patients. The 3 postimplant MRIs acquired at the community hospital were sufficient to identify 208 out of 211 seeds implanted without reference to computed tomography (CT).

CONCLUSIONS

Acquiring postimplant prostate brachytherapy MRI without an ERC has several advantages including better patient tolerance, lower costs, higher clinical throughput, and widespread access to precision LDR prostate brachytherapy. This prospective study confirms that the use of an ERC can be circumvented with fully bSSFP and advanced MRI scan techniques in a major academic cancer center and community hospital, potentially enabling postimplant assessment of MARS LDR prostate brachytherapy without CT.

A Deep Framework Assembling Principled Modules for CS-MRI: Unrolling Perspective, Convergence Behaviors, and Practical Modeling

Compressed Sensing Magnetic Resonance Imaging (CS-MRI) significantly accelerates MR acquisition at a sampling rate much lower than the Nyquist criterion. A major challenge for CS-MRI lies in solving the severely ill-posed inverse problem to reconstruct aliasing-free MR images from the sparse k -space data. Conventional methods typically optimize an energy function, producing restoration of high quality, but their iterative numerical solvers unavoidably bring extremely large time consumption. Recent deep techniques provide fast restoration by either learning direct prediction to final reconstruction or plugging learned modules into the energy optimizer. Nevertheless, these data-driven predictors cannot guarantee the reconstruction following principled constraints underlying the domain knowledge so that the reliability of their reconstruction process is questionable. In this paper, we propose a deep framework assembling principled modules for CS-MRI that fuses learning strategy with the iterative solver of a conventional reconstruction energy. This framework embeds an optimal condition checking mechanism, fostering efficient and reliable reconstruction. We also apply the framework to three practical tasks, i.e., complex-valued data reconstruction, parallel imaging and reconstruction with Rician noise. Extensive experiments on both benchmark and manufacturer-testing images demonstrate that the proposed method reliably converges to the optimal solution more efficiently and accurately than the state-of-the-art in various scenarios.

SARA-GAN: Self-Attention and Relative Average Discriminator Based Generative Adversarial Networks for Fast Compressed Sensing MRI Reconstruction

Research on undersampled magnetic resonance image (MRI) reconstruction can increase the speed of MRI imaging and reduce patient suffering. In this paper, an undersampled MRI reconstruction method based on Generative Adversarial Networks with the Self-Attention mechanism and the Relative Average discriminator (SARA-GAN) is proposed. In our SARA-GAN, the relative average discriminator theory is applied to make full use of the prior knowledge, in which half of the input data of the discriminator is true and half is fake. At the same time, a self-attention mechanism is incorporated into the high-layer of the generator to build long-range dependence of the image, which can overcome the problem of limited convolution kernel size. Besides, spectral normalization is employed to stabilize the training process. Compared with three widely used GAN-based MRI reconstruction methods, i.e., DAGAN, DAWGAN, and DAWGAN-GP, the proposed method can obtain a higher peak signal-to-noise ratio (PSNR) and structural similarity index measure(SSIM), and the details of the reconstructed image are more abundant and more realistic for further clinical scrutinization and diagnostic tasks.

Highly accelerated submillimeter resolution 3D GRASE with controlled T 2 blurring in T 2 -weighted functional MRI at 7 Tesla: A feasibility study

PURPOSE

To achieve highly accelerated submillimeter resolution T 2 -weighted functional MRI at 7T by developing a three-dimensional gradient and spin echo imaging (GRASE) with inner-volume selection and variable flip angles (VFA).

METHODS

GRASE imaging has disadvantages in that (a) k-space modulation causes T 2 blurring by limiting the number of slices and (b) a VFA scheme results in partial success with substantial SNR loss. In this work, accelerated GRASE with controlled T 2 blurring is developed to improve a point spread function (PSF) and temporal signal-to-noise ratio (tSNR) with a large number of slices. To this end, the VFA scheme is designed by minimizing a trade-off between SNR and blurring for functional sensitivity, and a new GRASE-optimized random encoding, which takes into account the complex signal decays of T 2 and T 2 ∗ weightings, is proposed by achieving incoherent aliasing for constrained reconstruction. Numerical and experimental studies were performed to validate the effectiveness of the proposed method over regular and VFA GRASE (R- and V-GRASE).

RESULTS

The proposed method, while achieving 0.8 mm isotropic resolution, functional MRI compared to R- and V-GRASE improves the spatial extent of the excited volume up to 36 slices with 52%-68% full width at half maximum (FWHM) reduction in PSF but approximately 2- to 3-fold mean tSNR improvement, thus resulting in higher BOLD activations.

CONCLUSIONS

We successfully demonstrated the feasibility of the proposed method in T 2 -weighted functional MRI. The proposed method is especially promising for cortical layer-specific functional MRI.

Real-time exercise stress cardiac MRI with Fourier-series reconstruction from golden-angle radial data

Magnetic resonance imaging (MRI) can measure cardiac response to exercise stress for evaluating and managing heart patients in the practice of clinical cardiology. However, exercise stress cardiac MRI have been clinically limited by the ability of available MRI techniques to quantitatively measure fast and unstable cardiac dynamics during exercise. The presented work is to develop a new real-time MRI technique for improved quantitative performance of exercise stress cardiac MRI. This technique seeks to represent real-time cardiac images as a sparse Fourier-series along the time. With golden-angle radial acquisition, parallel imaging and compressed sensing can be integrated into a linear system of equations for resolving Fourier coefficients that are in turn used to generate real-time cardiac images from the Fourier-series representation. Fourier-series reconstruction from golden-angle radial data can effectively address data insufficiency due to MRI speed limitation, providing a real-time approach to exercise stress cardiac MRI. To demonstrate the feasibility, an exercise stress cardiac MRI experiment was run to investigate biventricular response to in-scanner biking exercise in a cohort of sixteen healthy volunteers. It was found that Fourier-series reconstruction from golden-angle radial data effectively detected exercise-induced increase in stroke volume and ejection fraction in a healthy heart. The presented work will improve the applications of exercise stress cardiac MRI in the practice of clinical cardiology.

Transfer learning in deep neural network based under-sampled MR image reconstruction

In Magnetic Resonance Imaging (MRI), the success of deep learning-based under-sampled MR image reconstruction depends on: (i) size of the training dataset, (ii) generalization capabilities of the trained neural network. Whenever there is a mismatch between the training and testing data, there is a need to retrain the neural network from scratch with thousands of MR images obtained using the same protocol. This may not be possible in MRI as it is costly and time consuming to acquire data. In this research, a transfer learning approach i.e. end-to-end fine tuning is proposed for U-Net to address the data scarcity and generalization problems of deep learning-based MR image reconstruction. First the generalization capabilities of a pre-trained U-Net (initially trained on the human brain images of 1.5 T scanner) are assessed for: (a) MR images acquired from MRI scanners of different magnetic field strengths, (b) MR images of different anatomies and (c) MR images under-sampled by different acceleration factors. Later, end-to-end fine tuning of the pre-trained U-Net is proposed for the reconstruction of the above-mentioned MR images (i.e. (a), (b) and (c)). The results show successful reconstructions obtained from the proposed method as reflected by the Structural SIMilarity index, Root Mean Square Error, Peak Signal-to-Noise Ratio and central line profile of the reconstructed images.

Towards accelerated quantitative sodium MRI at 7 T in the skeletal muscle: Comparison of anisotropic acquisition- and compressed sensing techniques

PURPOSE

To compare three anisotropic acquisition schemes and three compressed sensing (CS) approaches for accelerated tissue sodium concentration (TSC) quantification using ²³Na MRI at 7 T.

MATERIALS AND METHODS

Three anisotropic 3D-radial acquisition sequences were evaluated using simulations, phantom- and in vivo TSC measurements: An anisotropic density-adapted 3D-radial sequence (3DPR-C), a 3D acquisition-weighted density-adapted stack-of-stars sampling scheme (SOS) and a SOS approach with golden-ratio rotation (SOS-GR). Eight healthy volunteers were examined at a 7 Tesla MRI system. TSC measurements of the calf were conducted with a nominal spatial resolution of Δx = (3.0 × 3.0 × 15.0) mm³ and a field of view of (156.0 × 156.0 × 240.0) mm³ for multiple undersampling factors (USF). Three CS reconstructions were evaluated: Total variation CS (TV-CS), 3D dictionary-learning compressed sensing (3D-DLCS) and TV-CS with a block matching prior (TV-BL-CS). Results of the simulations and measurements were compared to a simulated ground truth (GT) or a fully sampled reference measurement (FS), respectively. The deviation of the mean TSC evaluated in multiple ROI (mE_GT/FS) and the normalized root-mean-squared error (NRMSE) for simulations were evaluated for CS and NUFFT reconstructions.

RESULTS

In simulations, the SOS-GR yielded the lowest NRMSE and mE_GT (< 4%) with NUFFT for an acquisition time (TA) of less than 2 min. CS further improved the results. In simulations and measurements, the best TSC quantification results were obtained with 3D-DLCS and SOS-GR (lowest NRMSE, mE_GT < 2.6% in simulations, mE_GT < 10.7% for phantom measurements and mE_FS < 6% in vivo) with an USF = 4.1 (TA < 2 min). TV-CS showed no or only slight improvements to NUFFT. The results of TV-BL-CS were similar to 3D-DLCS.

DISCUSSION

The TA for TSC measurements could be reduced to less than 2 min by using adapted sequences such as SOS-GR and CS reconstruction approaches such as 3D-DLCS or TV-BL-CS, while the quantitative accuracy stays comparable to a fully sampled NUFFT reconstruction (approx. 8 min TA). In future, the lower TA could improve clinical applicability of TSC measurements.

Diagnostic quality assessment of IR-prepared 3D magnetic resonance neuroimaging accelerated using compressed sensing and k-space sampling order optimization

PURPOSE

To evaluate the clinical diagnostic efficacy of accelerated 3D magnetic resonance (MR) neuroimaging by radiological assessment for image quality and artefacts.

STUDY TYPE

Prospective healthy volunteer study.

SUBJECTS

Eight healthy subjects.

FIELD STRENGTH/SEQUENCE

Inversion Recovery (IR) prepared 3D Gradient Echo (GRE) sequence on a 1.5 T GE Signa HDx scanner.

ASSESSMENT

Independent radiological diagnostic quality assessments of accelerated 3D MR brain datasets were carried out by four experienced neuro-radiologists who were blinded to the acceleration factor and to the subject. The radiological grading was based on a previously reported radiological scoring key that was used for image quality assessment of human brains.

STATISTICAL TESTS

Bland-Altman analysis.

RESULTS

Optimization of the k-space sampling order was important for preserving contrast in accelerated scans. Despite having lower scores than fully sampled datasets, the majority of the compressed sensing (CS) accelerated brain datasets with k-space sampling order optimization (19/24 datasets by Radiologist 1, 24/24 datasets by Radiologist 2 and 16/24 datasets by Radiologist 3) were graded to be fully diagnostic indicating that there was adequate confidence for performing gross structural assessment of the brain.

CONCLUSION

Optimization of k-space acquisition order improves the clinical utility of CS accelerated 3D neuroimaging. This method may be appropriate for routine radiological assessment of the brain.

Rapid Imaging: Recent Advances in Abdominal MRI for Reducing Acquisition Time and Its Clinical Applications

Magnetic resonance imaging (MRI) plays an important role in abdominal imaging. The high contrast resolution offered by MRI provides better lesion detection and its capacity to provide multiparametric images facilitates lesion characterization more effectively than computed tomography. However, the relatively long acquisition time of MRI often detrimentally affects the image quality and limits its accessibility. Recent developments have addressed these drawbacks. Specifically, multiphasic acquisition of contrast-enhanced MRI, free-breathing dynamic MRI using compressed sensing technique, simultaneous multi-slice acquisition for diffusion-weighted imaging, and breath-hold three-dimensional magnetic resonance cholangiopancreatography are recent notable advances in this field. This review explores the aforementioned state-of-the-art techniques by focusing on their clinical applications and potential benefits, as well as their likely future direction.

Deep complex convolutional network for fast reconstruction of 3D late gadolinium enhancement cardiac MRI

Several deep-learning models have been proposed to shorten MRI scan time. Prior deep-learning models that utilize real-valued kernels have limited capability to learn rich representations of complex MRI data. In this work, we utilize a complex-valued convolutional network (ℂNet) for fast reconstruction of highly under-sampled MRI data and evaluate its ability to rapidly reconstruct 3D late gadolinium enhancement (LGE) data. ℂNet preserves the complex nature and optimal combination of real and imaginary components of MRI data throughout the reconstruction process by utilizing complex-valued convolution, novel radial batch normalization, and complex activation function layers in a U-Net architecture. A prospectively under-sampled 3D LGE cardiac MRI dataset of 219 patients (17 003 images) at acceleration rates R = 3 through R = 5 was used to evaluate ℂNet. The dataset was further retrospectively under-sampled to a maximum of R = 8 to simulate higher acceleration rates. We created three reconstructions of the 3D LGE dataset using (1) ℂNet, (2) a compressed-sensing-based low-dimensional-structure self-learning and thresholding algorithm (LOST), and (3) a real-valued U-Net (realNet) with the same number of parameters as ℂNet. LOST-reconstructed data were considered the reference for training and evaluation of all models. The reconstructed images were quantitatively evaluated using mean-squared error (MSE) and the structural similarity index measure (SSIM), and subjectively evaluated by three independent readers. Quantitatively, ℂNet-reconstructed images had significantly improved MSE and SSIM values compared with realNet (MSE, 0.077 versus 0.091; SSIM, 0.876 versus 0.733, respectively; p < 0.01). Subjective quality assessment showed that ℂNet-reconstructed image quality was similar to that of compressed sensing and significantly better than that of realNet. ℂNet reconstruction was also more than 300 times faster than compressed sensing. Retrospective under-sampled images demonstrate the potential of ℂNet at higher acceleration rates. ℂNet enables fast reconstruction of highly accelerated 3D MRI with superior performance to real-valued networks, and achieves faster reconstruction than compressed sensing.

A half-century of innovation in technology-preparing MRI for the 21st century

MRI developed during the last half-century from a very basic concept to an indispensable non-ionising medical imaging technique that has found broad application in diagnostics, therapy control and far beyond. Due to its excellent soft-tissue contrast and the huge variety of accessible tissue- and physiological-parameters, MRI is often preferred to other existing modalities. In the course of its development, MRI underwent many substantial transformations. From the beginning, starting as a proof of concept, much effort was expended to develop the appropriate basic scanning technology and methodology, and to establish the many clinical contrasts (e.g., T₁, T₂, flow, diffusion, water/fat, etc.) that MRI is famous for today. Beyond that, additional prominent innovations to the field have been parallel imaging and compressed sensing, leading to significant scanning time reductions, and the move towards higher static magnetic field strengths, which led to increased sensitivity and improved image quality. Improvements in workflow and the use of artificial intelligence are among many current trends seen in this field, paving the way for a broad use of MRI. The 125th anniversary of the BJR is a good point to reflect on all these changes and developments and to offer some slightly speculative ideas as to what the future may bring.

Accelerated coronary MRI with sRAKI: A database-free self-consistent neural network k-space reconstruction for arbitrary undersampling

Purpose

To accelerate coronary MRI acquisitions with arbitrary undersampling patterns by using a novel reconstruction algorithm that applies coil self-consistency using subject-specific neural networks.

Methods

Self-consistent robust artificial-neural-networks for k-space interpolation (sRAKI) performs iterative parallel imaging reconstruction by enforcing self-consistency among coils. The approach bears similarity to SPIRiT, but extends the linear convolutions in SPIRiT to nonlinear interpolation using convolutional neural networks (CNNs). These CNNs are trained individually for each scan using the scan-specific autocalibrating signal (ACS) data. Reconstruction is performed by imposing the learned self-consistency and data-consistency, which enables sRAKI to support random undersampling patterns. Fully-sampled targeted right coronary artery MRI was acquired in six healthy subjects. The data were retrospectively undersampled, and reconstructed using SPIRiT, l₁-SPIRiT and sRAKI for acceleration rates of 2 to 5. Additionally, prospectively undersampled whole-heart coronary MRI was acquired to further evaluate reconstruction performance.

Results

sRAKI reduces noise amplification and blurring artifacts compared with SPIRiT and l₁-SPIRiT, especially at high acceleration rates in targeted coronary MRI. Quantitative analysis shows that sRAKI outperforms these techniques in terms of normalized mean-squared-error (~44% and ~21% over SPIRiT and l1-SPIRiT at rate 5) and vessel sharpness (~10% and ~20% over SPIRiT and l₁-SPIRiT at rate 5). Whole-heart data shows the sharpest coronary arteries when resolved using sRAKI, with 11% and 15% improvement in vessel sharpness over SPIRiT and l₁-SPIRiT, respectively.

Conclusion

sRAKI is a database-free neural network-based reconstruction technique that may further accelerate coronary MRI with arbitrary undersampling patterns, while improving noise resilience over linear parallel imaging and image sharpness over l₁ regularization techniques.

Compressed Sensing: From Research to Clinical Practice with Deep Neural Networks

Compressed sensing (CS) reconstruction methods leverage sparse structure in underlying signals to recover high-resolution images from highly undersampled measurements. When applied to magnetic resonance imaging (MRI), CS has the potential to dramatically shorten MRI scan times, increase diagnostic value, and improve overall patient experience. However, CS has several shortcomings which limit its clinical translation such as: 1) artifacts arising from inaccurate sparse modelling assumptions, 2) extensive parameter tuning required for each clinical application, and 3) clinically infeasible reconstruction times. Recently, CS has been extended to incorporate deep neural networks as a way of learning complex image priors from historical exam data. Commonly referred to as unrolled neural networks, these techniques have proven to be a compelling and practical approach to address the challenges of sparse CS. In this tutorial, we will review the classical compressed sensing formulation and outline steps needed to transform this formulation into a deep learning-based reconstruction framework. Supplementary open source code in Python will be used to demonstrate this approach with open databases. Further, we will discuss considerations in applying unrolled neural networks in the clinical setting.