Multi-mask self-supervised learning for physics-guided neural networks in highly accelerated magnetic resonance imaging

Advancing MRI Reconstruction: A Systematic Review of Deep Learning and Compressed Sensing Integration

Magnetic resonance imaging (MRI) is a non-invasive imaging modality and provides comprehensive anatomical and functional insights into the human body. However, its long acquisition times can lead to patient discomfort, motion artifacts, and limiting real-time applications. To address these challenges, strategies such as parallel imaging have been applied, which utilize multiple receiver coils to speed up the data acquisition process. Additionally, compressed sensing (CS) is a method that facilitates image reconstruction from sparse data, significantly reducing image acquisition time by minimizing the amount of data collection needed. Recently, deep learning (DL) has emerged as a powerful tool for improving MRI reconstruction. It has been integrated with parallel imaging and CS principles to achieve faster and more accurate MRI reconstructions. This review comprehensively examines DL-based techniques for MRI reconstruction. We categorize and discuss various DL-based methods, including end-to-end approaches, unrolled optimization, and federated learning, highlighting their potential benefits. Our systematic review highlights significant contributions and underscores the potential of DL in MRI reconstruction. Additionally, we summarize key results and trends in DL-based MRI reconstruction, including quantitative metrics, the dataset, acceleration factors, and the progress of and research interest in DL techniques over time. Finally, we discuss potential future directions and the importance of DL-based MRI reconstruction in advancing medical imaging. To facilitate further research in this area, we provide a GitHub repository that includes up-to-date DL-based MRI reconstruction publications and public datasets-https://github.com/mosaf/Awesome-DL-based-CS-MRI.

Accelerating multi-coil MR image reconstruction using weak supervision

Deep-learning-based MR image reconstruction in settings where large fully sampled dataset collection is infeasible requires methods that effectively use both under-sampled and fully sampled datasets. This paper evaluates a weakly supervised, multi-coil, physics-guided approach to MR image reconstruction, leveraging both dataset types, to improve both the quality and robustness of reconstruction. A physics-guided end-to-end variational network (VarNet) is pretrained in a self-supervised manner using a 4 × under-sampled dataset following the self-supervised learning via data undersampling (SSDU) methodology. The pre-trained weights are transferred to another VarNet, which is fine-tuned using a smaller, fully sampled dataset by optimizing multi-scale structural similarity (MS-SSIM) loss in image space. The proposed methodology is compared with fully self-supervised and fully supervised training. Reconstruction quality improvements in SSIM, PSNR, and NRMSE when abundant training data is available (the high-data regime), and enhanced robustness when training data is scarce (the low-data regime) are demonstrated using weak supervision for knee and brain MR image reconstructions at 8 × and 10 × acceleration, respectively. Multi-coil physics-guided MR image reconstruction using both under-sampled and fully sampled datasets is achievable with transfer learning and fine-tuning. This methodology can provide improved reconstruction quality in the high-data regime and improved robustness in the low-data regime at high acceleration rates.

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.

Physics-Driven Deep Learning Reconstruction of Frequency-Modulated Rabi-Encoded Echoes for Faster Accessible MRI

Magnetic resonance imaging (MRI) is a powerful imaging modality with exceptional soft tissue contrast capabilities, but it is estimated to only serve 10% of the world's population reliably. This lack of access is largely due to the multi-million cost of initial investment, as well as recurring expenses. Radiofrequency (RF) imaging methods present an opportunity to reduce MRI costs by replacing expensive B0 gradients with less expensive RF ${\text{B}}_1^ +$ field gradients for spatial encoding. Frequency-modulated Rabi encoded echoes (FREE) is one such technique that has demonstrated robust phase-encoded imaging capabilities over large inhomogeneities. However, conventional reconstruction of such acquisitions lead to distortions due to nonlinear phase accrual, and imaging speed is slow when ${\text{B}}_1^ +$ phase encoding is employed in two spatial dimensions. In this work, we propose a physics-driven deep learning (PD-DL) reconstruction approach to resolve distortion artifacts of FREE acquisitions, while enabling higher acceleration rates. We map out the forward operator for FREE encoding, and devise an unrolled network that utilizes this operator. Results on sequential gradient superposition (SGS) FREE sequence indicates feasibility of up to 4-fold acceleration with a single receive-coil.

Instance-Wise MRI Reconstruction Based on Self-Supervised Implicit Neural Representation

Accelerated MRI involves a trade-off between sampling sufficiency and acquisition time. Supervised deep learning methods have shown great success in MRI reconstruction from under-sampled measurements, but they typically require a large set of fully-sampled MR images for training, which can be difficult to obtain. In this paper, we present a novel fully self-supervised method based on implicit neural representation, which requires only a single under-sampled MRI instance for training. To effectively guide the self-supervised learning process, we introduced multiple novel supervisory signals in both the image and frequency domains. Experimental results indicate that the proposed method outperforms existing self-supervised methods and even a supervised method, demonstrating its strong reliability and flexibility. Our code is publicly available at https://github.com/YSongxiao/SSLInstanceReconMRI.Clinical relevance- The proposed method can significantly enhance the image quality of under-sampled MR images without the need of ground-truth fully-sampled MR images for supervision and additional prior images for guidance.

Unsupervised deep learning with convolutional neural networks for static parallel transmit design: A retrospective study

PURPOSE

To mitigateB 1 + $$ {B}_1^{+} $$ inhomogeneity at 7T for multi-channel transmit arrays using unsupervised deep learning with convolutional neural networks (CNNs).

METHODS

Deep learning parallel transmit (pTx) pulse design has received attention, but such methods have relied on supervised training and did not use CNNs for multi-channelB 1 + $$ {B}_1^{+} $$ maps. In this work, we introduce an alternative approach that facilitates the use of CNNs with multi-channelB 1 + $$ {B}_1^{+} $$ maps while performing unsupervised training. The multi-channelB 1 + $$ {B}_1^{+} $$ maps are concatenated along the spatial dimension to enable shift-equivariant processing amenable to CNNs. Training is performed in an unsupervised manner using a physics-driven loss function that minimizes the discrepancy of the Bloch simulation with the target magnetization, which eliminates the calculation of reference transmit RF weights. The training database comprises 3824 2D sagittal, multi-channelB 1 + $$ {B}_1^{+} $$ maps of the healthy human brain from 143 subjects.B 1 + $$ {B}_1^{+} $$ data were acquired at 7T using an 8Tx/32Rx head coil. The proposed method is compared to the unregularized magnitude least-squares (MLS) solution for the target magnetization in static pTx design.

RESULTS

The proposed method outperformed the unregularized MLS solution for RMS error and coefficient-of-variation and had comparable energy consumption. Additionally, the proposed method did not show local phase singularities leading to distinct holes in the resulting magnetization unlike the unregularized MLS solution.

CONCLUSION

Proposed unsupervised deep learning with CNNs performs better than unregularized MLS in static pTx for speed and robustness.

Accelerated MRI reconstructions via variational network and feature domain learning

We introduce three architecture modifications to enhance the performance of the end-to-end (E2E) variational network (VarNet) for undersampled MRI reconstructions. We first implemented the Feature VarNet, which propagates information throughout the cascades of the network in an N-channel feature-space instead of a 2-channel feature-space. Then, we add an attention layer that utilizes the spatial locations of Cartesian undersampling artifacts to further improve performance. Lastly, we combined the Feature and E2E VarNets into the Feature-Image (FI) VarNet, to facilitate cross-domain learning and boost accuracy. Reconstructions were evaluated on the fastMRI dataset using standard metrics and clinical scoring by three neuroradiologists. Feature and FI VarNets outperformed the E2E VarNet for 4 × , 5 × and 8 × Cartesian undersampling in all studied metrics. FI VarNet secured second place in the public fastMRI leaderboard for 4 × Cartesian undersampling, outperforming all open-source models in the leaderboard. Radiologists rated FI VarNet brain reconstructions with higher quality and sharpness than the E2E VarNet reconstructions. FI VarNet excelled in preserving anatomical details, including blood vessels, whereas E2E VarNet discarded or blurred them in some cases. The proposed FI VarNet enhances the reconstruction quality of undersampled MRI and could enable clinically acceptable reconstructions at higher acceleration factors than currently possible.

ROBUST OUTER VOLUME SUBTRACTION WITH DEEP LEARNING GHOSTING DETECTION FOR HIGHLY-ACCELERATED REAL-TIME DYNAMIC MRI

Real-time dynamic MRI is important for visualizing time-varying processes in several applications, including cardiac imaging, where it enables free-breathing images of the beating heart without ECG gating. However, current real-time MRI techniques commonly face challenges in achieving the required spatio-temporal resolutions due to limited acceleration rates. In this study, we propose a deep learning (DL) technique for improving the estimation of stationary outer-volume signal from shifted time-interleaved undersampling patterns. Our approach utilizes the pseudo-periodic nature of the ghosting artifacts arising from the moving organs. Subsequently, this estimated outer-volume signal is subtracted from individual timeframes of the real-time MR time series, and each timeframe is reconstructed individually using physics-driven DL methods. Results show improved image quality at high acceleration rates, where conventional methods fail.

A CONVEX COMPRESSIBILITY-INSPIRED UNSUPERVISED LOSS FUNCTION FOR PHYSICS-DRIVEN DEEP LEARNING RECONSTRUCTION

Physics-driven deep learning (PD-DL) methods have gained popularity for improved reconstruction of fast MRI scans. Though supervised learning has been used in early works, there has been a recent interest in unsupervised learning methods for training PD-DL. In this work, we take inspiration from statistical image processing and compressed sensing (CS), and propose a novel convex loss function as an alternative learning strategy. Our loss function evaluates the compressibility of the output image while ensuring data fidelity to assess the quality of reconstruction in versatile settings, including supervised, unsupervised, and zero-shot scenarios. In particular, we leverage the reweightedl 1 norm that has been shown to approximate thel 0 norm for quality evaluation. Results show that the PD-DL networks trained with the proposed loss formulation outperform conventional methods, while maintaining similar quality to PD-DL models trained using existing supervised and unsupervised techniques.

NON-CARTESIAN SELF-SUPERVISED PHYSICS-DRIVEN DEEP LEARNING RECONSTRUCTION FOR HIGHLY-ACCELERATED MULTI-ECHO SPIRAL FMRI

Functional MRI (fMRI) is an important tool for non-invasive studies of brain function. Over the past decade, multi-echo fMRI methods that sample multiple echo times has become popular with potential to improve quantification. While these acquisitions are typically performed with Cartesian trajectories, non-Cartesian trajectories, in particular spiral acquisitions, hold promise for denser sampling of echo times. However, such acquisitions require very high acceleration rates for sufficient spatiotemporal resolutions. In this work, we propose to use a physics-driven deep learning (PD-DL) reconstruction to accelerate multi-echo spiral fMRI by 10-fold. We modify a self-supervised learning algorithm for optimized training with non-Cartesian trajectories and use it to train the PD-DL network. Results show that the proposed self-supervised PD-DL reconstruction achieves high spatio-temporal resolution with meaningful BOLD analysis.

CYCLE-CONSISTENT SELF-SUPERVISED LEARNING FOR IMPROVED HIGHLY-ACCELERATED MRI RECONSTRUCTION

Physics-driven deep learning (PD-DL) has become a powerful tool for accelerated MRI. Recent developments have also developed unsupervised learning for PD-DL, including self-supervised learning. However, at very high acceleration rates, such approaches show performance deterioration. In this study, we propose to use cyclic-consistency (CC) to improve self-supervised learning for highly accelerated MRI. In our proposed CC, simulated measurements are obtained by undersampling the network output using patterns drawn from the same distribution as the true one. The reconstructions of these simulated measurements are obtained using the same network, which are then compared to the acquired data at the true sampling locations. This CC approach is used in conjunction with a masking-based self-supervised loss. Results show that the proposed method can substantially reduce aliasing artifacts at high acceleration rates, including rate 6 and 8 fastMRI knee imaging and 20-fold HCP-style fMRI.

Fast MRI Reconstruction Using Deep Learning-based Compressed Sensing: A Systematic Review

Magnetic resonance imaging (MRI) has revolutionized medical imaging, providing a non-invasive and highly detailed look into the human body. However, the long acquisition times of MRI present challenges, causing patient discomfort, motion artifacts, and limiting real-time applications. To address these challenges, researchers are exploring various techniques to reduce acquisition time and improve the overall efficiency of MRI. One such technique is compressed sensing (CS), which reduces data acquisition by leveraging image sparsity in transformed spaces. In recent years, deep learning (DL) has been integrated with CS-MRI, leading to a new framework that has seen remarkable growth. DL-based CS-MRI approaches are proving to be highly effective in accelerating MR imaging without compromising image quality. This review comprehensively examines DL-based CS-MRI techniques, focusing on their role in increasing MR imaging speed. We provide a detailed analysis of each category of DL-based CS-MRI including end-to-end, unroll optimization, self-supervised, and federated learning. Our systematic review highlights significant contributions and underscores the exciting potential of DL in CS-MRI. Additionally, our systematic review efficiently summarizes key results and trends in DL-based CS-MRI including quantitative metrics, the dataset used, acceleration factors, and the progress of and research interest in DL techniques over time. Finally, we discuss potential future directions and the importance of DL-based CS-MRI in the advancement of medical imaging. To facilitate further research in this area, we provide a GitHub repository that includes up-to-date DL-based CS-MRI publications and publicly available datasets - https://github.com/mosaf/Awesome-DL-based-CS-MRI.

IWNeXt: an image-wavelet domain ConvNeXt-based network for self-supervised multi-contrast MRI reconstruction

Objective.Multi-contrast magnetic resonance imaging (MC MRI) can obtain more comprehensive anatomical information of the same scanning object but requires a longer acquisition time than single-contrast MRI. To accelerate MC MRI speed, recent studies only collect partial k-space data of one modality (target contrast) to reconstruct the remaining non-sampled measurements using a deep learning-based model with the assistance of another fully sampled modality (reference contrast). However, MC MRI reconstruction mainly performs the image domain reconstruction with conventional CNN-based structures by full supervision. It ignores the prior information from reference contrast images in other sparse domains and requires fully sampled target contrast data. In addition, because of the limited receptive field, conventional CNN-based networks are difficult to build a high-quality non-local dependency.Approach.In the paper, we propose an Image-Wavelet domain ConvNeXt-based network (IWNeXt) for self-supervised MC MRI reconstruction. Firstly, INeXt and WNeXt based on ConvNeXt reconstruct undersampled target contrast data in the image domain and refine the initial reconstructed result in the wavelet domain respectively. To generate more tissue details in the refinement stage, reference contrast wavelet sub-bands are used as additional supplementary information for wavelet domain reconstruction. Then we design a novel attention ConvNeXt block for feature extraction, which can capture the non-local information of the MC image. Finally, the cross-domain consistency loss is designed for self-supervised learning. Especially, the frequency domain consistency loss deduces the non-sampled data, while the image and wavelet domain consistency loss retain more high-frequency information in the final reconstruction.Main results.Numerous experiments are conducted on the HCP dataset and the M4Raw dataset with different sampling trajectories. Compared with DuDoRNet, our model improves by 1.651 dB in the peak signal-to-noise ratio.Significance.IWNeXt is a potential cross-domain method that can enhance the accuracy of MC MRI reconstruction and reduce reliance on fully sampled target contrast images.

DC-SiamNet: Deep contrastive Siamese network for self-supervised MRI reconstruction

Reconstruction methods based on deep learning have greatly shortened the data acquisition time of magnetic resonance imaging (MRI). However, these methods typically utilize massive fully sampled data for supervised training, restricting their application in certain clinical scenarios and posing challenges to the reconstruction effect when high-quality MR images are unavailable. Recently, self-supervised methods have been developed that only undersampled MRI images participate in the network training. Nevertheless, due to the lack of complete referable MR image data, self-supervised reconstruction is prone to produce incorrect structure contents, such as unnatural texture details and over-smoothed tissue sites. To solve this problem, we propose a self-supervised Deep Contrastive Siamese Network (DC-SiamNet) for fast MR imaging. First, DC-SiamNet performs the reconstruction with a Siamese unrolled structure and obtains visual representations in different iterative phases. Particularly, an attention-weighted average pooling module is employed at the bottleneck layer of the U-shape regularization unit, which can effectively aggregate valuable local information of the underlying feature map in the generated representation vector. Then, a novel hybrid loss function is designed to drive the self-supervised reconstruction and contrastive learning simultaneously by forcing the output consistency across different branches in the frequency domain, the image domain, and the latent space. The proposed method is extensively evaluated with different sampling patterns on the IXI brain dataset and the MRINet knee dataset. Experimental results show that DC-SiamNet can achieve 0.93 in structural similarity and 33.984 dB in peak signal-to-noise ratio on the IXI brain dataset under 8x acceleration. It has better reconstruction accuracy than other methods, and the performance is close to the corresponding model trained with full supervision, especially when the sampling rate is low. In addition, generalization experiments verify that our method has a strong cross-domain reconstruction ability for different contrast brain images.

A Theoretical Framework for Self-Supervised MR Image Reconstruction Using Sub-Sampling via Variable Density Noisier2Noise

In recent years, there has been attention on leveraging the statistical modeling capabilities of neural networks for reconstructing sub-sampled Magnetic Resonance Imaging (MRI) data. Most proposed methods assume the existence of a representative fully-sampled dataset and use fully-supervised training. However, for many applications, fully sampled training data is not available, and may be highly impractical to acquire. The development and understanding of self-supervised methods, which use only sub-sampled data for training, are therefore highly desirable. This work extends the Noisier2Noise framework, which was originally constructed for self-supervised denoising tasks, to variable density sub-sampled MRI data. We use the Noisier2Noise framework to analytically explain the performance of Self-Supervised Learning via Data Undersampling (SSDU), a recently proposed method that performs well in practice but until now lacked theoretical justification. Further, we propose two modifications of SSDU that arise as a consequence of the theoretical developments. Firstly, we propose partitioning the sampling set so that the subsets have the same type of distribution as the original sampling mask. Secondly, we propose a loss weighting that compensates for the sampling and partitioning densities. On the fastMRI dataset we show that these changes significantly improve SSDU's image restoration quality and robustness to the partitioning parameters.

High-fidelity Database-free Deep Learning Reconstruction for Real-time Cine Cardiac MRI

Real-time cine cardiac MRI provides an ECG-free free-breathing alternative to clinical gold-standard ECG-gated breath-hold segmented cine MRI for evaluation of heart function. Real-time cine MRI data acquisition during free breathing snapshot imaging enables imaging of patient cohorts that cannot be imaged with segmented or breath-hold acquisitions, but requires rapid imaging to achieve sufficient spatial-temporal resolutions. However, at high acceleration rates, conventional reconstruction techniques suffer from residual aliasing and temporal blurring, including advanced methods such as compressed sensing with radial trajectories. Recently, deep learning (DL) reconstruction has emerged as a powerful tool in MRI. However, its utility for free-breathing real-time cine MRI has been limited, as database-learning of spatio-temporal correlations with varying breathing and cardiac motion patterns across subjects has been challenging. Zero-shot self-supervised physics-guided deep learning (PG-DL) reconstruction has been proposed to overcome such challenges of database training by enabling subject-specific training. In this work, we adapt zero-shot PG-DL for real-time cine MRI with a spatio-temporal regularization. We compare our method to TGRAPPA, locally low-rank (LLR) regularized reconstruction and database-trained PG-DL reconstruction, both for retrospectively and prospectively accelerated datasets. Results on highly accelerated real-time Cartesian cine MRI show that the proposed method outperforms other reconstruction methods, both visibly in terms of noise and aliasing, and quantitatively.

High-fidelity Database-free Deep Learning Reconstruction for Real-time Cine Cardiac MRI

Real-time cine cardiac MRI provides an ECG-free free-breathing alternative to clinical gold-standard ECG-gated breath-hold segmented cine MRI for evaluation of heart function. Real-time cine MRI data acquisition during free breathing snapshot imaging enables imaging of patient cohorts that cannot be imaged with segmented or breath-hold acquisitions, but requires rapid imaging to achieve sufficient spatial-temporal resolutions. However, at high acceleration rates, conventional reconstruction techniques suffer from residual aliasing and temporal blurring, including advanced methods such as compressed sensing with radial trajectories. Recently, deep learning (DL) reconstruction has emerged as a powerful tool in MRI. However, its utility for free-breathing real-time cine MRI has been limited, as database-learning of spatio-temporal correlations with varying breathing and cardiac motion patterns across subjects has been challenging. Zero-shot self-supervised physics-guided deep learning (PG-DL) reconstruction has been proposed to overcome such challenges of database training by enabling subject-specific training. In this work, we adapt zero-shot PG-DL for real-time cine MRI with a spatio-temporal regularization. We compare our method to TGRAPPA, locally low-rank (LLR) regularized reconstruction and database-trained PG-DL reconstruction, both for retrospectively and prospectively accelerated datasets. Results on highly accelerated real-time Cartesian cine MRI show that the proposed method outperforms other reconstruction methods, both visibly in terms of noise and aliasing, and quantitatively.

Signal intensity informed multi-coil encoding operator for physics-guided deep learning reconstruction of highly accelerated myocardial perfusion CMR

PURPOSE

To develop a physics-guided deep learning (PG-DL) reconstruction strategy based on a signal intensity informed multi-coil (SIIM) encoding operator for highly-accelerated simultaneous multislice (SMS) myocardial perfusion cardiac MRI (CMR).

METHODS

First-pass perfusion CMR acquires highly-accelerated images with dynamically varying signal intensity/SNR following the administration of a gadolinium-based contrast agent. Thus, using PG-DL reconstruction with a conventional multi-coil encoding operator leads to analogous signal intensity variations across different time-frames at the network output, creating difficulties in generalization for varying SNR levels. We propose to use a SIIM encoding operator to capture the signal intensity/SNR variations across time-frames in a reformulated encoding operator. This leads to a more uniform/flat contrast at the output of the PG-DL network, facilitating generalizability across time-frames. PG-DL reconstruction with the proposed SIIM encoding operator is compared to PG-DL with conventional encoding operator, split slice-GRAPPA, locally low-rank (LLR) regularized reconstruction, low-rank plus sparse (L + S) reconstruction, and regularized ROCK-SPIRiT.

RESULTS

Results on highly accelerated free-breathing first pass myocardial perfusion CMR at three-fold SMS and four-fold in-plane acceleration show that the proposed method improves upon the reconstruction methods use for comparison. Substantial noise reduction is achieved compared to split slice-GRAPPA, and aliasing artifacts reduction compared to LLR regularized reconstruction, L + S reconstruction and PG-DL with conventional encoding. Furthermore, a qualitative reader study indicated that proposed method outperformed all methods.

CONCLUSION

PG-DL reconstruction with the proposed SIIM encoding operator improves generalization across different time-frames /SNRs in highly accelerated perfusion CMR.