Compressed Sensing MRI Reconstruction Using a Generative Adversarial Network With a Cyclic Loss

Correlation of objective image quality metrics with radiologists' diagnostic confidence depends on the clinical task performed

Purpose

Objective image quality metrics (IQMs) are widely used as outcome measures to assess acquisition and reconstruction strategies for diagnostic images. For nonpathological magnetic resonance (MR) images, these IQMs correlate to varying degrees with expert radiologists' confidence scores of overall perceived diagnostic image quality. However, it is unclear whether IQMs also correlate with task-specific diagnostic image quality or expert radiologists' confidence in performing a specific diagnostic task, which calls into question their use as surrogates for radiologist opinion.

Approach

0.5 T MR images from 16 stroke patients and two healthy volunteers were retrospectively undersampled ( R = 1 to 7 × ) and reconstructed via compressed sensing. Three neuroradiologists reported the presence/absence of acute ischemic stroke (AIS) and assigned a Fazekas score describing the extent of chronic ischemic lesion burden. Neuroradiologists ranked their confidence in performing each task using a 1 to 5 Likert scale. Confidence scores were correlated with noise quality measure, the visual information fidelity criterion, the feature similarity index, root mean square error, and structural similarity (SSIM) via nonlinear regression modeling.

Results

Although acceleration alters image quality, neuroradiologists remain able to report pathology. All of the IQMs tested correlated to some degree with diagnostic confidence for assessing chronic ischemic lesion burden, but none correlated with diagnostic confidence in diagnosing the presence/absence of AIS due to consistent radiologist performance regardless of image degradation.

Conclusions

Accelerated images were helpful for understanding the ability of IQMs to assess task-specific diagnostic image quality in the context of chronic ischemic lesion burden, although not in the case of AIS diagnosis. These findings suggest that commonly used IQMs, such as the SSIM index, do not necessarily indicate an image's utility when performing certain diagnostic tasks.

DGEDDGAN: A dual-domain generator and edge-enhanced dual discriminator generative adversarial network for MRI reconstruction

Magnetic resonance imaging (MRI) as a critical clinical tool in medical imaging, requires a long scan time for producing high-quality MRI images. To accelerate the speed of MRI while reconstructing high-quality images with sharper edges and fewer aliases, a novel dual-domain generator and edge-enhancement dual discriminator generative adversarial network structure named DGEDDGAN for MRI reconstruction is proposed, in which one discriminator is responsible for holistic image reconstruction, whereas the other is adopted to enhance the edge preservation. A dual-domain U-Net structure that cascades the frequency domain and image domain is designed for the generator. The densely connected residual block is used to replace the traditional U-Net convolution block to improve the feature reuse capability while overcoming the gradient vanishing problem. The coordinate attention mechanism in each skip connection is employed to effectively reduce the loss of spatial information and enforce the feature selection capability. Extensive experiments on two publicly available datasets i.e., IXI dataset and CC-359, demonstrate that the proposed method can reconstruct the high-quality MRI images with more edge details and fewer artifacts, outperforming several state-of-the-art methods under various sampling rates and masks. The time of single-image reconstruction is below 13 ms, which meets the demand of faster processing.

Towards High-Quality MRI Reconstruction With Anisotropic Diffusion-Assisted Generative Adversarial Networks and Its Multi-Modal Images Extension

Recently, fast Magnetic Resonance Imaging reconstruction technology has emerged as a promising way to improve the clinical diagnostic experience by significantly reducing scan times. While existing studies have used Generative Adversarial Networks to achieve impressive results in reconstructing MR images, they still suffer from challenges such as blurred zones/boundaries and abnormal spots caused by inevitable noise in the reconstruction process. To this end, we propose a novel deep framework termed Anisotropic Diffusion-Assisted Generative Adversarial Networks, which aims to maximally preserve valid high-frequency information and structural details while minimizing noises in reconstructed images by optimizing a joint loss function in a unified framework. In doing so, it enables more authentic and accurate MR image generation. To specifically handle unforeseeable noises, an Anisotropic Diffused Reconstruction Module is developed and added aside the backbone network as a denoise assistant, which improves the final image quality by minimizing reconstruction losses between targets and iteratively denoised generative outputs with no extra computational complexity during the testing phase. To make the most of valuable MRI data, we extend its application to support multi-modal learning to boost reconstructed image quality by aggregating more valid information from images of diverse modalities. Extensive experiments on public datasets show that the proposed framework can achieve superior performance in polishing up the quality of reconstructed MR images. For example, the proposed method obtains average PSNR and mSSIM values of 35.785 dB and 0.9765 on the MRNet dataset, which are at least about 2.9 dB and 0.07 higher than those from the baselines.

A lightweight adaptive spatial channel attention efficient net B3 based generative adversarial network approach for MR image reconstruction from under sampled data

Magnetic Resonance Imaging (MRI) stands out as a notable non-invasive method for medical imaging assessments, widely employed in early medical diagnoses due to its exceptional resolution in portraying soft tissue structures. However, the MRI method faces challenges with its inherently slow acquisition process, stemming from the sequential sampling in k-space and limitations in traversal speed due to physiological and hardware constraints. Compressed Sensing in MRI (CS-MRI) accelerates image acquisition by utilizing greatly under-sampled k-space information. Despite its advantages, conventional CS-MRI encounters issues such as sluggish iterations and artefacts at higher acceleration factors. Recent advancements integrate deep learning models into CS-MRI, inspired by successes in various computer vision domains. It has drawn significant attention from the MRI community because of its great potential for image reconstruction from undersampled k-space data in fast MRI. This paper proposes a lightweight Adaptive Spatial-Channel Attention EfficientNet B3-based Generative Adversarial Network (ASCA-EffNet GAN) for fast, high-quality MR image reconstruction from greatly under-sampled k-space information in CS-MRI. The proposed GAN employs a U-net generator with ASCA-based EfficientNet B3 for encoder blocks and a ResNet decoder. The discriminator is a binary classifier with ASCA-based EfficientNet B3, a fully connected layer and a sigmoid layer. The EfficientNet B3 utilizes a compound scaling strategy that achieves a balance amongst model depth, width, and resolution, resulting in optimal performance with a reduced number of parameters. Furthermore, the adaptive attention mechanisms in the proposed ASCA-EffNet GAN effectively capture spatial and channel-wise features, contributing to detailed anatomical structure reconstruction. Experimental evaluations on the dataset demonstrate ASCA-EffNet GAN's superior performance across various metrics, surpassing conventional reconstruction methods. Hence, ASCA-EffNet GAN showcases remarkable reconstruction capabilities even under high under-sampling rates, making it suitable for clinical applications.

DONet: Dual-Octave Network for Fast MR Image Reconstruction

Magnetic resonance (MR) image acquisition is an inherently prolonged process, whose acceleration has long been the subject of research. This is commonly achieved by obtaining multiple undersampled images, simultaneously, through parallel imaging. In this article, we propose the dual-octave network (DONet), which is capable of learning multiscale spatial-frequency features from both the real and imaginary components of MR data, for parallel fast MR image reconstruction. More specifically, our DONet consists of a series of dual-octave convolutions (Dual-OctConvs), which are connected in a dense manner for better reuse of features. In each Dual-OctConv, the input feature maps and convolutional kernels are first split into two components (i.e., real and imaginary) and then divided into four groups according to their spatial frequencies. Then, our Dual-OctConv conducts intragroup information updating and intergroup information exchange to aggregate the contextual information across different groups. Our framework provides three appealing benefits: 1) it encourages information interaction and fusion between the real and imaginary components at various spatial frequencies to achieve richer representational capacity; 2) the dense connections between the real and imaginary groups in each Dual-OctConv make the propagation of features more efficient by feature reuse; and 3) DONet enlarges the receptive field by learning multiple spatial-frequency features of both the real and imaginary components. Extensive experiments on two popular datasets (i.e., clinical knee and fastMRI), under different undersampling patterns and acceleration factors, demonstrate the superiority of our model in accelerated parallel MR image reconstruction.

AI for image quality and patient safety in CT and MRI

Substantial endeavors have been recently dedicated to developing artificial intelligence (AI) solutions, especially deep learning-based, tailored to enhance radiological procedures, in particular algorithms designed to minimize radiation exposure and enhance image clarity. Thus, not only better diagnostic accuracy but also reduced potential harm to patients was pursued, thereby exemplifying the intersection of technological innovation and the highest standards of patient care. We provide herein an overview of recent AI developments in computed tomography and magnetic resonance imaging. Major AI results in CT regard: optimization of patient positioning, scan range selection (avoiding "overscanning"), and choice of technical parameters; reduction of the amount of injected contrast agent and injection flow rate (also avoiding extravasation); faster and better image reconstruction reducing noise level and artifacts. Major AI results in MRI regard: reconstruction of undersampled images; artifact removal, including those derived from unintentional patient's (or fetal) movement or from heart motion; up to 80-90% reduction of GBCA dose. Challenges include limited generalizability, lack of external validation, insufficient explainability of models, and opacity of decision-making. Developing explainable AI algorithms that provide transparent and interpretable outputs is essential to enable seamless AI integration into CT and MRI practice. RELEVANCE STATEMENT: This review highlights how AI-driven advancements in CT and MRI improve image quality and enhance patient safety by leveraging AI solutions for dose reduction, contrast optimization, noise reduction, and efficient image reconstruction, paving the way for safer, faster, and more accurate diagnostic imaging practices. KEY POINTS: Advancements in AI are revolutionizing the way radiological images are acquired, reconstructed, and interpreted. AI algorithms can assist in optimizing radiation doses, reducing scan times, and enhancing image quality. AI techniques are paving the way for a future of more efficient, accurate, and safe medical imaging examinations.

Fourier Convolution Block with global receptive field for MRI reconstruction

Reconstructing images from under-sampled Magnetic Resonance Imaging (MRI) signals significantly reduces scan time and improves clinical practice. However, Convolutional Neural Network (CNN)-based methods, while demonstrating great performance in MRI reconstruction, may face limitations due to their restricted receptive field (RF), hindering the capture of global features. This is particularly crucial for reconstruction, as aliasing artifacts are distributed globally. Recent advancements in Vision Transformers have further emphasized the significance of a large RF. In this study, we proposed a novel global Fourier Convolution Block (FCB) with whole image RF and low computational complexity by transforming the regular spatial domain convolutions into frequency domain. Visualizations of the effective RF and trained kernels demonstrated that FCB improves the RF of reconstruction models in practice. The proposed FCB was evaluated on four popular CNN architectures using brain and knee MRI datasets. Models with FCB achieved superior PSNR and SSIM than baseline models and exhibited more details and texture recovery. The code is publicly available at https://github.com/Haozhoong/FCB.

AutoSamp: Autoencoding k-Space Sampling via Variational Information Maximization for 3D MRI

Accelerated MRI protocols routinely involve a predefined sampling pattern that undersamples the k-space. Finding an optimal pattern can enhance the reconstruction quality, however this optimization is a challenging task. To address this challenge, we introduce a novel deep learning framework, AutoSamp, based on variational information maximization that enables joint optimization of sampling pattern and reconstruction of MRI scans. We represent the encoder as a non-uniform Fast Fourier Transform that allows continuous optimization of k-space sample locations on a non-Cartesian plane, and the decoder as a deep reconstruction network. Experiments on public 3D acquired MRI datasets show improved reconstruction quality of the proposed AutoSamp method over the prevailing variable density and variable density Poisson disc sampling for both compressed sensing and deep learning reconstructions. We demonstrate that our data-driven sampling optimization method achieves 4.4dB, 2.0dB, 0.75dB, 0.7dB PSNR improvements over reconstruction with Poisson Disc masks for acceleration factors of R =5, 10, 15, 25, respectively. Prospectively accelerated acquisitions with 3D FSE sequences using our optimized sampling patterns exhibit improved image quality and sharpness. Furthermore, we analyze the characteristics of the learned sampling patterns with respect to changes in acceleration factor, measurement noise, underlying anatomy, and coil sensitivities. We show that all these factors contribute to the optimization result by affecting the sampling density, k-space coverage and point spread functions of the learned sampling patterns.

Blip-up blip-down circular EPI (BUDA-cEPI) for distortion-free dMRI with rapid unrolled deep learning reconstruction

PURPOSE

BUDA-cEPI has been shown to achieve high-quality, high-resolution diffusion magnetic resonance imaging (dMRI) with fast acquisition time, particularly when used in conjunction with S-LORAKS reconstruction. However, this comes at a cost of more complex reconstruction that is computationally prohibitive. In this work we develop rapid reconstruction pipeline for BUDA-cEPI to pave the way for its deployment in routine clinical and neuroscientific applications. The proposed reconstruction includes the development of ML-based unrolled reconstruction as well as rapid ML-based B0 and eddy current estimations that are needed. The architecture of the unroll network was designed so that it can mimic S-LORAKS regularization well, with the addition of virtual coil channels.

METHODS

BUDA-cEPI RUN-UP - a model-based framework that incorporates off-resonance and eddy current effects was unrolled through an artificial neural network with only six gradient updates. The unrolled network alternates between data consistency (i.e., forward BUDA-cEPI and its adjoint) and regularization steps where U-Net plays a role as the regularizer. To handle the partial Fourier effect, the virtual coil concept was also introduced into the reconstruction to effectively take advantage of the smooth phase prior and trained to predict the ground-truth images obtained by BUDA-cEPI with S-LORAKS.

RESULTS

The introduction of the Virtual Coil concept into the unrolled network was shown to be key to achieving high-quality reconstruction for BUDA-cEPI. With the inclusion of an additional non-diffusion image (b-value = 0 s/mm2), a slight improvement was observed, with the normalized root mean square error further reduced by approximately 5 %. The reconstruction times for S-LORAKS and the proposed unrolled networks were approximately 225 and 3 s per slice, respectively.

CONCLUSION

BUDA-cEPI RUN-UP was shown to reduce the reconstruction time by ∼88× when compared to the state-of-the-art technique, while preserving imaging details as demonstrated through DTI application.

PixCUE: Joint Uncertainty Estimation and Image Reconstruction in MRI using Deep Pixel Classification

Deep learning (DL) models are effective in leveraging latent representations from MR data, emerging as state-of-the-art solutions for accelerated MRI reconstruction. However, challenges arise due to the inherent uncertainties associated with undersampling in k-space, coupled with the over- or under-parameterized and opaque nature of DL models. Addressing uncertainty has thus become a critical issue in DL MRI reconstruction. Monte Carlo (MC) inference techniques are commonly employed to estimate uncertainty, involving multiple reconstructions of the same scan to compute variance as a measure of uncertainty. Nevertheless, these methods entail significant computational expenses, requiring multiple inferences through the DL model. In this context, we propose a novel approach to uncertainty estimation during MRI reconstruction using a pixel classification framework. Our method, PixCUE (Pixel Classification Uncertainty Estimation), generates both the reconstructed image and an uncertainty map in a single forward pass through the DL model. We validate the efficacy of this approach by demonstrating that PixCUE-generated uncertainty maps exhibit a strong correlation with reconstruction errors across various MR imaging sequences and under diverse adversarial conditions. We present an empirical relationship between uncertainty estimations using PixCUE and established reconstruction metrics such as NMSE, PSNR, and SSIM. Furthermore, we establish a correlation between the estimated uncertainties from PixCUE and the conventional MC method. Our findings affirm that PixCUE reliably estimates uncertainty in MRI reconstruction with minimal additional computational cost.

Decoding Glioblastoma Heterogeneity: Neuroimaging Meets Machine Learning

Recent advancements in neuroimaging and machine learning have significantly improved our ability to diagnose and categorize isocitrate dehydrogenase (IDH)-wildtype glioblastoma, a disease characterized by notable tumoral heterogeneity, which is crucial for effective treatment. Neuroimaging techniques, such as diffusion tensor imaging and magnetic resonance radiomics, provide noninvasive insights into tumor infiltration patterns and metabolic profiles, aiding in accurate diagnosis and prognostication. Machine learning algorithms further enhance glioblastoma characterization by identifying distinct imaging patterns and features, facilitating precise diagnoses and treatment planning. Integration of these technologies allows for the development of image-based biomarkers, potentially reducing the need for invasive biopsy procedures and enabling personalized therapy targeting specific pro-tumoral signaling pathways and resistance mechanisms. Although significant progress has been made, ongoing innovation is essential to address remaining challenges and further improve these methodologies. Future directions should focus on refining machine learning models, integrating emerging imaging techniques, and elucidating the complex interplay between imaging features and underlying molecular processes. This review highlights the pivotal role of neuroimaging and machine learning in glioblastoma research, offering invaluable noninvasive tools for diagnosis, prognosis prediction, and treatment planning, ultimately improving patient outcomes. These advances in the field promise to usher in a new era in the understanding and classification of IDH-wildtype glioblastoma.

A review of deep learning-based reconstruction methods for accelerated MRI using spatiotemporal and multi-contrast redundancies

Accelerated magnetic resonance imaging (MRI) has played an essential role in reducing data acquisition time for MRI. Acceleration can be achieved by acquiring fewer data points in k-space, which results in various artifacts in the image domain. Conventional reconstruction methods have resolved the artifacts by utilizing multi-coil information, but with limited robustness. Recently, numerous deep learning-based reconstruction methods have been developed, enabling outstanding reconstruction performances with higher acceleration. Advances in hardware and developments of specialized network architectures have produced such achievements. Besides, MRI signals contain various redundant information including multi-coil redundancy, multi-contrast redundancy, and spatiotemporal redundancy. Utilization of the redundant information combined with deep learning approaches allow not only higher acceleration, but also well-preserved details in the reconstructed images. Consequently, this review paper introduces the basic concepts of deep learning and conventional accelerated MRI reconstruction methods, followed by review of recent deep learning-based reconstruction methods that exploit various redundancies. Lastly, the paper concludes by discussing the challenges, limitations, and potential directions of future developments.

An unrolled neural network for accelerated dynamic MRI based on second-order half-quadratic splitting model

The reconstruction of dynamic magnetic resonance images from incomplete k-space data has sparked significant research interest due to its potential to reduce scan time. However, traditional iterative optimization algorithms fail to faithfully reconstruct images at higher acceleration factors and incur long reconstruction time. Furthermore, end-to-end deep learning-based reconstruction algorithms suffer from large model parameters and lack robustness in the reconstruction results. Recently, unrolled deep learning models, have shown immense potential in algorithm stability and applicability flexibility. In this paper, we propose an unrolled deep learning network based on a second-order Half-Quadratic Splitting(HQS) algorithm, where the forward propagation process of this framework strictly follows the computational flow of the HQS algorithm. In particular, we propose a degradation-sense module by associating random sampling patterns with intermediate variables to guide the iterative process. We introduce the Information Fusion Transformer(IFT) to extract both local and non-local prior information from image sequences, thereby removing aliasing artifacts resulting from random undersampling. Finally, we impose low-rank constraints within the HQS algorithm to further enhance the reconstruction results. The experiments demonstrate that each component module of our proposed model contributes to the improvement of the reconstruction task. Our proposed method achieves comparably satisfying performance to the state-of-the-art methods and it exhibits excellent generalization capabilities across different sampling masks. At the low acceleration factor, there is a 0.7% enhancement in the PSNR. Furthermore, when the acceleration factor reached 8 and 12, the PSNR achieves an improvement of 3.4% and 5.8% respectively.

Myo-regressor Deep Informed Neural NetwOrk (Myo-DINO) for fast MR parameters mapping in neuromuscular disorders

Magnetic Resonance (MR) parameters mapping in muscle Magnetic Resonance Imaging (mMRI) is predominantly performed using pattern recognition-based algorithms, which are characterised by high computational costs and scalability issues in the context of multi-parametric mapping. Deep Learning (DL) has been demonstrated to be a robust and efficient method for rapid MR parameters mapping. However, its application in mMRI domain to investigate Neuromuscular Disorders (NMDs) has not yet been explored. In addition, data-driven DL models suffered in interpretation and explainability of the learning process. We developed a Physics Informed Neural Network called Myo-Regressor Deep Informed Neural NetwOrk (Myo-DINO) for efficient and explainable Fat Fraction (FF), water-T2 (wT2) and B1 mapping from a cohort of NMDs.A total of 2165 slices (232 subjects) from Multi-Echo Spin Echo (MESE) images were selected as the input dataset for which FF, wT2,B1 ground truth maps were computed using the MyoQMRI toolbox. This toolbox exploits the Extended Phase Graph (EPG) theory with a two-component model (water and fat signal) and slice profile to simulate the signal evolution in the MESE framework. A customized U-Net architecture was implemented as the Myo-DINO architecture. The squared L2 norm loss was complemented by two distinct physics models to define two 'Physics-Informed' loss functions: Cycling Loss 1 embedded a mono-exponential model to describe the relaxation of water protons, while Cycling Loss 2 incorporated the EPG theory with slice profile to model the magnetization dephasing under the effect of gradients and RF pulses. The Myo-DINO was trained with the hyperparameter value of the 'Physics-Informed' component held constant, i.e. λmodel = 1, while different hyperparameter values (λcnn) were applied to the squared L2 norm component in both the cycling loss. In particular, hard (λcnn=10), normal (λcnn=1) and self-supervised (λcnn=0) constraints were applied to gradually decrease the impact of the squared L2 norm component on the 'Physics Informed' term during the Myo-DINO training process. Myo-DINO achieved higher performance with Cycling Loss 2 for FF, wT2 and B1 prediction. In particular, high reconstruction similarity and quality (Structural Similarity Index > 0.92, Peak Signal to Noise ratio > 30.0 db) and small reconstruction error (Normalized Root Mean Squared Error < 0.038) to the reference maps were shown with self-supervised weighting of the Cycling Loss 2. In addition muscle-wise FF, wT2 and B1 predicted values showed good agreement with the reference values. The Myo-DINO has been demonstrated to be a robust and efficient workflow for MR parameters mapping in the context of mMRI. This provides preliminary evidence that it can be an effective alternative to the reference post-processing algorithm. In addition, our results demonstrate that Cycling Loss 2, which incorporates the Extended Phase Graph (EPG) model, provides the most robust and relevant physical constraints for Myo-DINO in this multi-parameter regression task. The use of Cycling Loss 2 with self-supervised constraint improved the explainability of the learning process because the network acquired domain knowledge solely in accordance with the assumptions of the EPG model.

Comprehensive assessment of imaging quality of artificial intelligence-assisted compressed sensing-based MR images in routine clinical settings

BACKGROUND

Conventional MR acceleration techniques, such as compressed sensing, parallel imaging, and half Fourier often face limitations, including noise amplification, reduced signal-to-noise ratio (SNR) and increased susceptibility to artifacts, which can compromise image quality, especially in high-speed acquisitions. Artificial intelligence (AI)-assisted compressed sensing (ACS) has emerged as a novel approach that combines the conventional techniques with advanced AI algorithms. The objective of this study was to examine the imaging quality of the ACS approach by qualitative and quantitative analysis for brain, spine, kidney, liver, and knee MR imaging, as well as compare the performance of this method with conventional (non-ACS) MR imaging.

METHODS

This study included 50 subjects. Three radiologists independently assessed the quality of MR images based on artefacts, image sharpness, overall image quality and diagnostic efficacy. SNR, contrast-to-noise ratio (CNR), edge content (EC), enhancement measure (EME), scanning time were used for quantitative evaluation. The Cohen's kappa correlation coefficient (k) was employed to measure radiologists' inter-observer agreement, and the Mann Whitney U-test used for comparison between non-ACS and ACS.

RESULTS

The qualitative analysis of three radiologists demonstrated that ACS images showed superior clinical information than non-ACS images with a mean k of ~ 0.70. The images acquired with ACS approach showed statistically higher values (p < 0.05) for SNR, CNR, EC, and EME compared to the non-ACS images. Furthermore, the study's findings indicated that ACS-enabled images reduced scan time by more than 50% while maintaining high imaging quality.

CONCLUSION

Integrating ACS technology into routine clinical settings has the potential to speed up image acquisition, improve image quality, and enhance diagnostic procedures and patient throughput.

Numerical and Clinical Evaluation of the Robustness of Open-source Networks for Parallel MR Imaging Reconstruction

PURPOSE

Deep neural networks (DNNs) for MRI reconstruction often require large datasets for training. Still, in clinical settings, the domains of datasets are diverse, and how robust DNNs are to domain differences between training and testing datasets has been an open question. Here, we numerically and clinically evaluate the generalization of the reconstruction networks across various domains under clinically practical conditions and provide practical guidance on what points to consider when selecting models for clinical application.

METHODS

We compare the reconstruction performance between four network models: U-Net, the deep cascade of convolutional neural networks (DC-CNNs), Hybrid Cascade, and variational network (VarNet). We used the public multicoil dataset fastMRI for training and testing and performed a single-domain test, where the domains of the dataset used for training and testing were the same, and cross-domain tests, where the source and target domains were different. We conducted a single-domain test (Experiment 1) and cross-domain tests (Experiments 2-4), focusing on six factors (the number of images, sampling pattern, acceleration factor, noise level, contrast, and anatomical structure) both numerically and clinically.

RESULTS

U-Net had lower performance than the three model-based networks and was less robust to domain shifts between training and testing datasets. VarNet had the highest performance and robustness among the three model-based networks, followed by Hybrid Cascade and DC-CNN. Especially, VarNet showed high performance even with a limited number of training images (200 images/10 cases). U-Net was more robust to domain shifts concerning noise level than the other model-based networks. Hybrid Cascade showed slightly better performance and robustness than DC-CNN, except for robustness to noise-level domain shifts. The results of the clinical evaluations generally agreed with the results of the quantitative metrics.

CONCLUSION

In this study, we numerically and clinically evaluated the robustness of the publicly available networks using the multicoil data. Therefore, this study provided practical guidance for clinical applications.

Model-based federated learning for accurate MR image reconstruction from undersampled k-space data

Deep learning-based methods have achieved encouraging performances in the field of Magnetic Resonance (MR) image reconstruction. Nevertheless, building powerful and robust deep learning models requires collecting large and diverse datasets from multiple centers. This raises concerns about ethics and data privacy. Recently, federated learning has emerged as a promising solution, enabling the utilization of multi-center data without the need for data transfer between institutions. Despite its potential, existing federated learning methods face challenges due to the high heterogeneity of data from different centers. Aggregation methods based on simple averaging, which are commonly used to combine the client's information, have shown limited reconstruction and generalization capabilities. In this paper, we propose a Model-based Federated learning framework (ModFed) to address these challenges. ModFed has three major contributions: (1) Different from existing data-driven federated learning methods, ModFed designs attention-assisted model-based neural networks that can alleviate the need for large amounts of data on each client; (2) To address the data heterogeneity issue, ModFed proposes an adaptive dynamic aggregation scheme, which can improve the generalization capability and robustness of the trained neural network models; (3) ModFed incorporates a spatial Laplacian attention mechanism and a personalized client-side loss regularization to capture the detailed information for accurate image reconstruction. The effectiveness of the proposed ModFed is evaluated on three in-vivo datasets. Experimental results show that when compared to six existing state-of-the-art federated learning approaches, ModFed achieves better MR image reconstruction performance with increased generalization capability. Codes will be made available at https://github.com/ternencewu123/ModFed.

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.

MPIGAN: An end-to-end deep based generative framework for high-resolution magnetic particle imaging reconstruction

BACKGROUND

Magnetic particle imaging (MPI) is a recently developed, non-invasive in vivo imaging technique to map the spatial distribution of superparamagnetic iron oxide nanoparticles (SPIONs) in animal tissues with high sensitivity and speed. It is a challenge to reconstruct images directly from the received signals of MPI device due to the complex physical behavior of the nanoparticles. System matrix and X-space are two commonly used MPI reconstruction methods, where the former is extremely time-consuming and the latter usually produces blurry images.

PURPOSE

Currently, we proposed an end-to-end machine learning framework to reconstruct high-resolution MPI images from 1-D voltage signals directly and efficiently.

METHODS

The proposed framework, which we termed "MPIGAN", was trained on a large MPI simulation dataset containing 291 597 pairs of high-resolution 2-D phantom images and each image's corresponding voltage signals, so that it was able to accurately capture the nonlinear relationship between the spatial distribution of SPIONs and the received voltage signal, and realized high-resolution MPI image reconstruction.

RESULTS

Experiment results showed that, MPIGAN exhibited remarkable abilities in high-resolution MPI image reconstruction. MPIGAN outperformed the traditional methods of system matrix and X-space in recovering the fine-scale structure of magnetic nanoparticles' spatial distribution and achieving enhanced reconstruction performance in both visual effects and quantitative assessments. Moreover, even when the received signals were severely contaminated with noise, MPIGAN could still generate high-quality MPI images.

CONCLUSION

Our study provides a promising AI solution for end-to-end, efficient, and high-resolution magnetic particle imaging reconstruction.

Mutli-modal straight flow matching for accelerated MR imaging

Diffusion models have garnered great interest lately in Magnetic Resonance (MR) image reconstruction. A key component of generating high-quality samples from noise is iterative denoising for thousands of steps. However, the complexity of inference steps has limited its applications. To solve the challenge in obtaining high-quality reconstructed images with fewer inference steps and computational complexity, we introduce a novel straight flow matching, based on a neural ordinary differential equation (ODE) generative model. Our model creates a linear path between undersampled images and reconstructed images, which can be accurately simulated with a few Euler steps. Furthermore, we propose a multi-modal straight flow matching model, which uses relatively easily available modalities as supplementary information to guide the reconstruction of target modalities. We introduce the low frequency fusion layer and the high frequency fusion layer into our multi-modal model, which has been proved to produce promising results in fusion tasks. The proposed multi-modal straight flow matching (MMSflow) achieves state-of-the-art performances in task of reconstruction in fastMRI and Brats-2020 and improves the sampling rate by an order of magnitude than other methods based on stochastic differential equations (SDE).

Towards reliable healthcare Imaging: conditional contrastive generative adversarial network for handling class imbalancing in MR Images

Background

Medical imaging datasets frequently encounter a data imbalance issue, where the majority of pixels correspond to healthy regions, and the minority belong to affected regions. This uneven distribution of pixels exacerbates the challenges associated with computer-aided diagnosis. The networks trained with imbalanced data tends to exhibit bias toward majority classes, often demonstrate high precision but low sensitivity.

Method

We have designed a new network based on adversarial learning namely conditional contrastive generative adversarial network (CCGAN) to tackle the problem of class imbalancing in a highly imbalancing MRI dataset. The proposed model has three new components: (1) class-specific attention, (2) region rebalancing module (RRM) and supervised contrastive-based learning network (SCoLN). The class-specific attention focuses on more discriminative areas of the input representation, capturing more relevant features. The RRM promotes a more balanced distribution of features across various regions of the input representation, ensuring a more equitable segmentation process. The generator of the CCGAN learns pixel-level segmentation by receiving feedback from the SCoLN based on the true negative and true positive maps. This process ensures that final semantic segmentation not only addresses imbalanced data issues but also enhances classification accuracy.

Results

The proposed model has shown state-of-art-performance on five highly imbalance medical image segmentation datasets. Therefore, the suggested model holds significant potential for application in medical diagnosis, in cases characterized by highly imbalanced data distributions. The CCGAN achieved the highest scores in terms of dice similarity coefficient (DSC) on various datasets: 0.965 ± 0.012 for BUS2017, 0.896 ± 0.091 for DDTI, 0.786 ± 0.046 for LiTS MICCAI 2017, 0.712 ± 1.5 for the ATLAS dataset, and 0.877 ± 1.2 for the BRATS 2015 dataset. DeepLab-V3 follows closely, securing the second-best position with DSC scores of 0.948 ± 0.010 for BUS2017, 0.895 ± 0.014 for DDTI, 0.763 ± 0.044 for LiTS MICCAI 2017, 0.696 ± 1.1 for the ATLAS dataset, and 0.846 ± 1.4 for the BRATS 2015 dataset.

Deep learning for accelerated and robust MRI reconstruction

Deep learning (DL) has recently emerged as a pivotal technology for enhancing magnetic resonance imaging (MRI), a critical tool in diagnostic radiology. This review paper provides a comprehensive overview of recent advances in DL for MRI reconstruction, and focuses on various DL approaches and architectures designed to improve image quality, accelerate scans, and address data-related challenges. It explores end-to-end neural networks, pre-trained and generative models, and self-supervised methods, and highlights their contributions to overcoming traditional MRI limitations. It also discusses the role of DL in optimizing acquisition protocols, enhancing robustness against distribution shifts, and tackling biases. Drawing on the extensive literature and practical insights, it outlines current successes, limitations, and future directions for leveraging DL in MRI reconstruction, while emphasizing the potential of DL to significantly impact clinical imaging practices.

Adversarial Learning for MRI Reconstruction and Classification of Cognitively Impaired Individuals

Game theory-inspired deep learning using a generative adversarial network provides an environment to competitively interact and accomplish a goal. In the context of medical imaging, most work has focused on achieving single tasks such as improving image resolution, segmenting images, and correcting motion artifacts. We developed a dual-objective adversarial learning framework that simultaneously 1) reconstructs higher quality brain magnetic resonance images (MRIs) that 2) retain disease-specific imaging features critical for predicting progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD). We obtained 3-Tesla, T1-weighted brain MRIs of participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI, N=342) and the National Alzheimer's Coordinating Center (NACC, N = 190) datasets. We simulated MRIs with missing data by removing 50% of sagittal slices from the original scans (i.e., diced scans). The generator was trained to reconstruct brain MRIs using the diced scans as input. We introduced a classifier into the GAN architecture to discriminate between stable (i.e., sMCI) and progressive MCI (i.e., pMCI) based on the generated images to facilitate encoding of disease-related information during reconstruction. The framework was trained using ADNI data and externally validated on NACC data. In the NACC cohort, generated images had better image quality than the diced scans (Structural similarity (SSIM) index: 0.553 ± 0.116 versus 0.348 ± 0.108). Furthermore, a classifier utilizing the generated images distinguished pMCI from sMCI more accurately than with the diced scans (F1-score: 0.634 ± 0.019 versus 0.573 ± 0.028). Competitive deep learning has potential to facilitate disease-oriented image reconstruction in those at risk of developing Alzheimer's disease.

Fourier Domain Robust Denoising Decomposition and Adaptive Patch MRI Reconstruction

The sparsity of the Fourier transform domain has been applied to magnetic resonance imaging (MRI) reconstruction in k -space. Although unsupervised adaptive patch optimization methods have shown promise compared to data-driven-based supervised methods, the following challenges exist in MRI reconstruction: 1) in previous k -space MRI reconstruction tasks, MRI with noise interference in the acquisition process is rarely considered. 2) Differences in transform domains should be resolved to achieve the high-quality reconstruction of low undersampled MRI data. 3) Robust patch dictionary learning problems are usually nonconvex and NP-hard, and alternate minimization methods are often computationally expensive. In this article, we propose a method for Fourier domain robust denoising decomposition and adaptive patch MRI reconstruction (DDAPR). DDAPR is a two-step optimization method for MRI reconstruction in the presence of noise and low undersampled data. It includes the low-rank and sparse denoising reconstruction model (LSDRM) and the robust dictionary learning reconstruction model (RDLRM). In the first step, we propose LSDRM for different domains. For the optimization solution, the proximal gradient method is used to optimize LSDRM by singular value decomposition and soft threshold algorithms. In the second step, we propose RDLRM, which is an effective adaptive patch method by introducing a low-rank and sparse penalty adaptive patch dictionary and using a sparse rank-one matrix to approximate the undersampled data. Then, the block coordinate descent (BCD) method is used to optimize the variables. The BCD optimization process involves valid closed-form solutions. Extensive numerical experiments show that the proposed method has a better performance than previous methods in image reconstruction based on compressed sensing or deep learning.

DiffGAN: An adversarial diffusion model with local transformer for MRI reconstruction

Magnetic resonance imaging (MRI) is non-invasive and crucial for clinical diagnosis, but it has long acquisition time and aliasing artifacts. Accelerated imaging techniques can effectively reduce the scanning time of MRI, thereby decreasing the anxiety and discomfort of patients. Vision Transformer (ViT) based methods have greatly improved MRI image reconstruction, but their computational complexity and memory requirements for the self-attention mechanism grow quadratically with image resolution, which limits their use for high resolution images. In addition, the current generative adversarial networks in MRI reconstruction are difficult to train stably. To address these problems, we propose a Local Vision Transformer (LVT) based adversarial Diffusion model (Diff-GAN) for accelerating MRI reconstruction. We employ a generative adversarial network (GAN) as the reverse diffusion model to enable large diffusion steps. In the forward diffusion module, we use a diffusion process to generate Gaussian mixture distribution noise, which mitigates the gradient vanishing issue in GAN training. This network leverages the LVT module with the local self-attention, which can capture high-quality local features and detailed information. We evaluate our method on four datasets: IXI, MICCAI 2013, MRNet and FastMRI, and demonstrate that Diff-GAN can outperform several state-of-the-art GAN-based methods for MRI reconstruction.

Cerebrospinal fluid flow artifact reduction with deep learning to optimize the evaluation of spinal canal stenosis on spine MRI

PURPOSE

The aim of study was to employ the Cycle Generative Adversarial Network (CycleGAN) deep learning model to diminish the cerebrospinal fluid (CSF) flow artifacts in cervical spine MRI. We also evaluate the agreement in quantifying spinal canal stenosis.

METHODS

For training model, we collected 9633 axial MR image pairs from 399 subjects. Then, additional 104 image pairs from 19 subjects were gathered for the test set. The deep learning model was developed using CycleGAN to reduce CSF flow artifacts, where T2 TSE images served as input, and T2 FFE images, known for fewer CSF flow artifacts. Post training, CycleGAN-generated images were subjected to both quantitative and qualitative evaluations for CSF artifacts. For assessing the agreement of spinal canal stenosis, four raters utilized an additional 104 pairs of original and CycleGAN-generated images, with inter-rater agreement evaluated using a weighted kappa value.

RESULTS

CSF flow artifacts were reduced in the CycleGAN-generated images compared to the T2 TSE and FFE images in both quantitative and qualitative analysis. All raters concordantly displayed satisfactory estimation results when assessing spinal canal stenosis using the CycleGAN-generated images with T2 TSE images (kappa = 0.61-0.75) compared to the original FFE with T2 TSE images (kappa = 0.48-0.71).

CONCLUSIONS

CycleGAN demonstrated the capability to produce images with diminished CSF flow artifacts. When paired with T2 TSE images, the CycleGAN-generated images allowed for more consistent assessment of spinal canal stenosis and exhibited agreement levels that were comparable to the combination of T2 TSE and FFE images.

The Use of Generative Adversarial Network and Graph Convolution Network for Neuroimaging-Based Diagnostic Classification

Functional connectivity (FC) obtained from resting-state functional magnetic resonance imaging has been integrated with machine learning algorithms to deliver consistent and reliable brain disease classification outcomes. However, in classical learning procedures, custom-built specialized feature selection techniques are typically used to filter out uninformative features from FC patterns to generalize efficiently on the datasets. The ability of convolutional neural networks (CNN) and other deep learning models to extract informative features from data with grid structure (such as images) has led to the surge in popularity of these techniques. However, the designs of many existing CNN models still fail to exploit the relationships between entities of graph-structure data (such as networks). Therefore, graph convolution network (GCN) has been suggested as a means for uncovering the intricate structure of brain network data, which has the potential to substantially improve classification accuracy. Furthermore, overfitting in classifiers can be largely attributed to the limited number of available training samples. Recently, the generative adversarial network (GAN) has been widely used in the medical field for its generative aspect that can generate synthesis images to cope with the problems of data scarcity and patient privacy. In our previous work, GCN and GAN have been designed to investigate FC patterns to perform diagnosis tasks, and their effectiveness has been tested on the ABIDE-I dataset. In this paper, the models will be further applied to FC data derived from more public datasets (ADHD, ABIDE-II, and ADNI) and our in-house dataset (PTSD) to justify their generalization on all types of data. The results of a number of experiments show the powerful characteristic of GAN to mimic FC data to achieve high performance in disease prediction. When employing GAN for data augmentation, the diagnostic accuracy across ADHD-200, ABIDE-II, and ADNI datasets surpasses that of other machine learning models, including results achieved with BrainNetCNN. Specifically, in ADHD, the accuracy increased from 67.74% to 73.96% with GAN, in ABIDE-II from 70.36% to 77.40%, and in ADNI, reaching 52.84% and 88.56% for multiclass and binary classification, respectively. GCN also obtains decent results, with the best accuracy in ADHD datasets at 71.38% for multinomial and 75% for binary classification, respectively, and the second-best accuracy in the ABIDE-II dataset (72.28% and 75.16%, respectively). Both GAN and GCN achieved the highest accuracy for the PTSD dataset, reaching 97.76%. However, there are still some limitations that can be improved. Both methods have many opportunities for the prediction and diagnosis of diseases.

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.

Improved reconstruction for highly accelerated propeller diffusion 1.5 T clinical MRI

PURPOSE

Propeller fast-spin-echo diffusion magnetic resonance imaging (FSE-dMRI) is essential for the diagnosis of Cholesteatoma. However, at clinical 1.5 T MRI, its signal-to-noise ratio (SNR) remains relatively low. To gain sufficient SNR, signal averaging (number of excitations, NEX) is usually used with the cost of prolonged scan time. In this work, we leveraged the benefits of Locally Low Rank (LLR) constrained reconstruction to enhance the SNR. Furthermore, we enhanced both the speed and SNR by employing Convolutional Neural Networks (CNNs) for the accelerated PROPELLER FSE-dMRI on a 1.5 T clinical scanner.

METHODS

Residual U-Net (RU-Net) was found to be efficient for propeller FSE-dMRI data. It was trained to predict 2-NEX images obtained by Locally Low Rank (LLR) constrained reconstruction and used 1-NEX images obtained via simplified reconstruction as the inputs. The brain scans from healthy volunteers and patients with cholesteatoma were performed for model training and testing. The performance of trained networks was evaluated with normalized root-mean-square-error (NRMSE), structural similarity index measure (SSIM), and peak SNR (PSNR).

RESULTS

For 4 × under-sampled with 7 blades data, online reconstruction appears to provide suboptimal images-some small details are missing due to high noise interferences. Offline LLR enables suppression of noises and discovering some small structures. RU-Net demonstrated further improvement compared to LLR by increasing 18.87% of PSNR, 2.11% of SSIM, and reducing 53.84% of NRMSE. Moreover, RU-Net is about 1500 × faster than LLR (0.03 vs. 47.59 s/slice).

CONCLUSION

The LLR remarkably enhances the SNR compared to online reconstruction. Moreover, RU-Net improves propeller FSE-dMRI as reflected in PSNR, SSIM, and NRMSE. It requires only 1-NEX data, which allows a 2 × scan time reduction. In addition, its speed is approximately 1500 times faster than that of LLR-constrained reconstruction.

Multi-Contrast Complementary Learning for Accelerated MR Imaging

Thanks to its powerful ability to depict high-resolution anatomical information, magnetic resonance imaging (MRI) has become an essential non-invasive scanning technique in clinical practice. However, excessive acquisition time often leads to the degradation of image quality and psychological discomfort among subjects, hindering its further popularization. Besides reconstructing images from the undersampled protocol itself, multi-contrast MRI protocols bring promising solutions by leveraging additional morphological priors for the target modality. Nevertheless, previous multi-contrast techniques mainly adopt a simple fusion mechanism that inevitably ignores valuable knowledge. In this work, we propose a novel multi-contrast complementary information aggregation network named MCCA, aiming to exploit available complementary representations fully to reconstruct the undersampled modality. Specifically, a multi-scale feature fusion mechanism has been introduced to incorporate complementary-transferable knowledge into the target modality. Moreover, a hybrid convolution transformer block was developed to extract global-local context dependencies simultaneously, which combines the advantages of CNNs while maintaining the merits of Transformers. Compared to existing MRI reconstruction methods, the proposed method has demonstrated its superiority through extensive experiments on different datasets under different acceleration factors and undersampling patterns.

NPB-REC: A non-parametric Bayesian deep-learning approach for undersampled MRI reconstruction with uncertainty estimation

The ability to reconstruct high-quality images from undersampled MRI data is vital in improving MRI temporal resolution and reducing acquisition times. Deep learning methods have been proposed for this task, but the lack of verified methods to quantify the uncertainty in the reconstructed images hampered clinical applicability. We introduce "NPB-REC", a non-parametric fully Bayesian framework, for MRI reconstruction from undersampled data with uncertainty estimation. We use Stochastic Gradient Langevin Dynamics during training to characterize the posterior distribution of the network parameters. This enables us to both improve the quality of the reconstructed images and quantify the uncertainty in the reconstructed images. We demonstrate the efficacy of our approach on a multi-coil MRI dataset from the fastMRI challenge and compare it to the baseline End-to-End Variational Network (E2E-VarNet). Our approach outperforms the baseline in terms of reconstruction accuracy by means of PSNR and SSIM (34.55, 0.908 vs. 33.08, 0.897, p<0.01, acceleration rate R=8) and provides uncertainty measures that correlate better with the reconstruction error (Pearson correlation, R=0.94 vs. R=0.91). Additionally, our approach exhibits better generalization capabilities against anatomical distribution shifts (PSNR and SSIM of 32.38, 0.849 vs. 31.63, 0.836, p<0.01, training on brain data, inference on knee data, acceleration rate R=8). NPB-REC has the potential to facilitate the safe utilization of deep learning-based methods for MRI reconstruction from undersampled data. Code and trained models are available at https://github.com/samahkh/NPB-REC.

Integrated prediction and control of mobility changes of young talents in the field of science and technology based on convolutional neural network

As the scientific and technological levels continue to rise, the dynamics of young talent within these fields are increasingly significant. Currently, there is a lack of comprehensive models for predicting the movement of young professionals in science and technology. To address this gap, this study introduces an integrated approach to forecasting and managing the flow of these talents, leveraging the power of convolutional neural networks (CNNs). The performance test of the proposed method shows that the prediction accuracy of this method is 76.98%, which is superior to the two comparison methods. In addition, the results showed that the average error of the model was 0.0285 lower than that of the model based on the recurrent prediction error (RPE) algorithm learning algorithm, and the average time was 41.6 s lower than that of the model based on the backpropagation (BP) learning algorithm. In predicting the flow of young talent, the study uses flow characteristics including personal characteristics, occupational characteristics, organizational characteristics and network characteristics. Through the above results, the study found that convolutional neural network can effectively use these features to predict the flow of young talents, and its model is superior to other commonly used models in processing speed and accuracy. The above results indicate that the model can provide organizations and government agencies with useful information about the flow trend of young talents, and help them to formulate better talent management strategies.

MA-RECON: Mask-aware deep-neural-network for robust fast MRI k-space interpolation

BACKGROUND AND OBJECTIVE

High-quality reconstruction of MRI images from under-sampled 'k-space' data, which is in the Fourier domain, is crucial for shortening MRI acquisition times and ensuring superior temporal resolution. Over recent years, a wealth of deep neural network (DNN) methods have emerged, aiming to tackle the complex, ill-posed inverse problem linked to this process. However, their instability against variations in the acquisition process and anatomical distribution exposes a deficiency in the generalization of relevant physical models within these DNN architectures. The goal of our work is to enhance the generalization capabilities of DNN methods for k-space interpolation by introducing 'MA-RECON', an innovative mask-aware DNN architecture and associated training method.

METHODS

Unlike preceding approaches, our 'MA-RECON' architecture encodes not only the observed data but also the under-sampling mask within the model structure. It implements a tailored training approach that leverages data generated with a variety of under-sampling masks to stimulate the model's generalization of the under-sampled MRI reconstruction problem. Therefore, effectively represents the associated inverse problem, akin to the classical compressed sensing approach.

RESULTS

The benefits of our MA-RECON approach were affirmed through rigorous testing with the widely accessible fastMRI dataset. Compared to standard DNN methods and DNNs trained with under-sampling mask augmentation, our approach demonstrated superior generalization capabilities. This resulted in a considerable improvement in robustness against variations in both the acquisition process and anatomical distribution, especially in regions with pathology.

CONCLUSION

In conclusion, our mask-aware strategy holds promise for enhancing the generalization capacity and robustness of DNN-based methodologies for MRI reconstruction from undersampled k-space data. Code is available in the following link: https://github.com/nitzanavidan/PD_Recon.

A Systematic Review and Identification of the Challenges of Deep Learning Techniques for Undersampled Magnetic Resonance Image Reconstruction

Deep learning (DL) in magnetic resonance imaging (MRI) shows excellent performance in image reconstruction from undersampled k-space data. Artifact-free and high-quality MRI reconstruction is essential for ensuring accurate diagnosis, supporting clinical decision-making, enhancing patient safety, facilitating efficient workflows, and contributing to the validity of research studies and clinical trials. Recently, deep learning has demonstrated several advantages over conventional MRI reconstruction methods. Conventional methods rely on manual feature engineering to capture complex patterns and are usually computationally demanding due to their iterative nature. Conversely, DL methods use neural networks with hundreds of thousands of parameters and automatically learn relevant features and representations directly from the data. Nevertheless, there are some limitations to DL-based techniques concerning MRI reconstruction tasks, such as the need for large, labeled datasets, the possibility of overfitting, and the complexity of model training. Researchers are striving to develop DL models that are more efficient, adaptable, and capable of providing valuable information for medical practitioners. We provide a comprehensive overview of the current developments and clinical uses by focusing on state-of-the-art DL architectures and tools used in MRI reconstruction. This study has three objectives. Our main objective is to describe how various DL designs have changed over time and talk about cutting-edge tactics, including their advantages and disadvantages. Hence, data pre- and post-processing approaches are assessed using publicly available MRI datasets and source codes. Secondly, this work aims to provide an extensive overview of the ongoing research on transformers and deep convolutional neural networks for rapid MRI reconstruction. Thirdly, we discuss several network training strategies, like supervised, unsupervised, transfer learning, and federated learning for rapid and efficient MRI reconstruction. Consequently, this article provides significant resources for future improvement of MRI data pre-processing and fast image reconstruction.

Learning ADC maps from accelerated radial k-space diffusion-weighted MRI in mice using a deep CNN-transformer model

PURPOSE

To accelerate radially sampled diffusion weighted spin-echo (Rad-DW-SE) acquisition method for generating high quality ADC maps.

METHODS

A deep learning method was developed to generate accurate ADC maps from accelerated DWI data acquired with the Rad-DW-SE method. The deep learning method integrates convolutional neural networks (CNNs) with vision transformers to generate high quality ADC maps from accelerated DWI data, regularized by a monoexponential ADC model fitting term. A model was trained on DWI data of 147 mice and evaluated on DWI data of 36 mice, with acceleration factors of 4× and 8× compared to the original acquisition parameters.

RESULTS

Ablation studies and experimental results have demonstrated that the proposed deep learning model generates higher quality ADC maps from accelerated DWI data than alternative deep learning methods under comparison when their performance is quantified in whole images as well as in regions of interest, including tumors, kidneys, and muscles.

CONCLUSIONS

The deep learning method with integrated CNNs and transformers provides an effective means to accurately compute ADC maps from accelerated DWI data acquired with the Rad-DW-SE method.

Hierarchical neural architecture search with adaptive global-local feature learning for Magnetic Resonance Image reconstruction

Neural architecture search (NAS) has been introduced into the design of deep neural network architectures for Magnetic Resonance Imaging (MRI) reconstruction since NAS-based methods can acquire the complex network architecture automatically without professional designing experience and improve the model's generalization ability. However, current NAS-based MRI reconstruction methods suffer from a lack of efficient operators in the search space, which leads to challenges in effectively recovering high-frequency details. This limitation is primarily due to the prevalent use of convolution operators in the current search space, which struggle to capture both global and local features of MR images simultaneously, resulting in insufficient information utilization. To address this issue, a generative adversarial network (GAN) based model is proposed to reconstruct the MR image from under-sampled K-space data. Firstly, parameterized global and local feature learning modules at multiple scales are added into the search space to improve the capability of recovering high-frequency details. Secondly, to mitigate the increased search time caused by the augmented search space, a hierarchical NAS is designed to learn the global-local feature learning modules that enable the reconstruction network to learn global and local information of MR images at different scales adaptively. Thirdly, to reduce the number of network parameters and computational complexity, the standard operations in global-local feature learning modules are replaced with lightweight operations. Finally, experiments on several publicly available brain MRI image datasets evaluate the performance of the proposed method. Compared to the state-of-the-art MRI reconstruction methods, the proposed method yields better reconstruction results in terms of peak signal-to-noise ratio and structural similarity at a lower computational cost. Additionally, our reconstruction results are validated through a brain tumor classification task, affirming the practicability of the proposed method. Our code is available at https://github.com/wwHwo/HNASMRI.

AliasNet: Alias artefact suppression network for accelerated phase-encode MRI

Sparse reconstruction is an important aspect of MRI, helping to reduce acquisition time and improve spatial-temporal resolution. Popular methods are based mostly on compressed sensing (CS), which relies on the random sampling of k-space to produce incoherent (noise-like) artefacts. Due to hardware constraints, 1D Cartesian phase-encode under-sampling schemes are popular for 2D CS-MRI. However, 1D under-sampling limits 2D incoherence between measurements, yielding structured aliasing artefacts (ghosts) that may be difficult to remove assuming a 2D sparsity model. Reconstruction algorithms typically deploy direction-insensitive 2D regularisation for these direction-associated artefacts. Recognising that phase-encode artefacts can be separated into contiguous 1D signals, we develop two decoupling techniques that enable explicit 1D regularisation and leverage the excellent 1D incoherence characteristics. We also derive a combined 1D + 2D reconstruction technique that takes advantage of spatial relationships within the image. Experiments conducted on retrospectively under-sampled brain and knee data demonstrate that combination of the proposed 1D AliasNet modules with existing 2D deep learned (DL) recovery techniques leads to an improvement in image quality. We also find AliasNet enables a superior scaling of performance compared to increasing the size of the original 2D network layers. AliasNet therefore improves the regularisation of aliasing artefacts arising from phase-encode under-sampling, by tailoring the network architecture to account for their expected appearance. The proposed 1D + 2D approach is compatible with any existing 2D DL recovery technique deployed for this application.

DERnet: a deep neural network for end-to-end reconstruction in magnetic particle imaging

Objective. Magnetic particle imaging (MPI) shows potential for contributing to biomedical research and clinical practice. However, MPI images are effectively affected by noise in the signal as its reconstruction is an ill-posed inverse problem. Thus, effective reconstruction method is required to reduce the impact of the noise while mapping signals to MPI images. Traditional methods rely on the hand-crafted data-consistency (DC) term and regularization term based on spatial priors to achieve noise-reducing and reconstruction. While these methods alleviate the ill-posedness and reduce noise effects, they may be difficult to fully capture spatial features.Approach. In this study, we propose a deep neural network for end-to-end reconstruction (DERnet) in MPI that emulates the DC term and regularization term using the feature mapping subnetwork and post-processing subnetwork, respectively, but in a data-driven manner. By doing so, DERnet can better capture signal and spatial features without relying on hand-crafted priors and strategies, thereby effectively reducing noise interference and achieving superior reconstruction quality.Main results. Our data-driven method outperforms the state-of-the-art algorithms with an improvement of 0.9-8.8 dB in terms of peak signal-to-noise ratio under various noise levels. The result demonstrates the advantages of our approach in suppressing noise interference. Furthermore, DERnet can be employed for measured data reconstruction with improved fidelity and reduced noise. In conclusion, our proposed method offers performance benefits in reducing noise interference and enhancing reconstruction quality by effectively capturing signal and spatial features.Significance. DERnet is a promising candidate method to improve MPI reconstruction performance and facilitate its more in-depth biomedical application.

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.

Universal generative modeling in dual domains for dynamic MRI

Dynamic magnetic resonance image reconstruction from incomplete k-space data has generated great research interest due to its ability to reduce scan time. Nevertheless, the reconstruction problem remains a thorny issue due to its ill posed nature. Recently, diffusion models, especially score-based generative models, have demonstrated great potential in terms of algorithmic robustness and flexibility of utilization. Moreover, a unified framework through the variance exploding stochastic differential equation is proposed to enable new sampling methods and further extend the capabilities of score-based generative models. Therefore, by taking advantage of the unified framework, we propose a k-space and image dual-domain collaborative universal generative model (DD-UGM), which combines the score-based prior with a low-rank regularization penalty to reconstruct highly under-sampled measurements. More precisely, we extract prior components from both image and k-space domains via a universal generative model and adaptively handle these prior components for faster processing while maintaining good generation quality. Experimental comparisons demonstrate the noise reduction and detail preservation abilities of the proposed method. Moreover, DD-UGM can reconstruct data of different frames by only training a single frame image, which reflects the flexibility of the proposed model.

Harnessing artificial intelligence in radiology to augment population health

This review article serves to highlight radiological services as a major cost driver for the healthcare sector, and the potential improvements in productivity and cost savings that can be generated by incorporating artificial intelligence (AI) into the radiology workflow, referencing Singapore healthcare as an example. More specifically, we will discuss the opportunities for AI in lowering healthcare costs and supporting transformational shifts in our care model in the following domains: predictive analytics for optimising throughput and appropriate referrals, computer vision for image enhancement (to increase scanner efficiency and decrease radiation exposure) and pattern recognition (to aid human interpretation and worklist prioritisation), natural language processing and large language models for optimising reports and text data-mining. In the context of preventive health, we will discuss how AI can support population level screening for major disease burdens through opportunistic screening and democratise expertise to increase access to radiological services in primary and community care.

Noise2Recon: Enabling SNR-robust MRI reconstruction with semi-supervised and self-supervised learning

PURPOSE

To develop a method for building MRI reconstruction neural networks robust to changes in signal-to-noise ratio (SNR) and trainable with a limited number of fully sampled scans.

METHODS

We propose Noise2Recon, a consistency training method for SNR-robust accelerated MRI reconstruction that can use both fully sampled (labeled) and undersampled (unlabeled) scans. Noise2Recon uses unlabeled data by enforcing consistency between model reconstructions of undersampled scans and their noise-augmented counterparts. Noise2Recon was compared to compressed sensing and both supervised and self-supervised deep learning baselines. Experiments were conducted using retrospectively accelerated data from the mridata three-dimensional fast-spin-echo knee and two-dimensional fastMRI brain datasets. All methods were evaluated in label-limited settings and among out-of-distribution (OOD) shifts, including changes in SNR, acceleration factors, and datasets. An extensive ablation study was conducted to characterize the sensitivity of Noise2Recon to hyperparameter choices.

RESULTS

In label-limited settings, Noise2Recon achieved better structural similarity, peak signal-to-noise ratio, and normalized-RMS error than all baselines and matched performance of supervised models, which were trained with14 × $$ 14\times $$ more fully sampled scans. Noise2Recon outperformed all baselines, including state-of-the-art fine-tuning and augmentation techniques, among low-SNR scans and when generalizing to OOD acceleration factors. Augmentation extent and loss weighting hyperparameters had negligible impact on Noise2Recon compared to supervised methods, which may indicate increased training stability.

CONCLUSION

Noise2Recon is a label-efficient reconstruction method that is robust to distribution shifts, such as changes in SNR, acceleration factors, and others, with limited or no fully sampled training data.

Spatial-Intensity Transforms for Medical Image-to-Image Translation

Image-to-image translation has seen major advances in computer vision but can be difficult to apply to medical images, where imaging artifacts and data scarcity degrade the performance of conditional generative adversarial networks. We develop the spatial-intensity transform (SIT) to improve output image quality while closely matching the target domain. SIT constrains the generator to a smooth spatial transform (diffeomorphism) composed with sparse intensity changes. SIT is a lightweight, modular network component that is effective on various architectures and training schemes. Relative to unconstrained baselines, this technique significantly improves image fidelity, and our models generalize robustly to different scanners. Additionally, SIT provides a disentangled view of anatomical and textural changes for each translation, making it easier to interpret the model's predictions in terms of physiological phenomena. We demonstrate SIT on two tasks: predicting longitudinal brain MRIs in patients with various stages of neurodegeneration, and visualizing changes with age and stroke severity in clinical brain scans of stroke patients. On the first task, our model accurately forecasts brain aging trajectories without supervised training on paired scans. On the second task, it captures associations between ventricle expansion and aging, as well as between white matter hyperintensities and stroke severity. As conditional generative models become increasingly versatile tools for visualization and forecasting, our approach demonstrates a simple and powerful technique for improving robustness, which is critical for translation to clinical settings. Source code is available at github.com/clintonjwang/spatial-intensity-transforms.

Overcoming the Challenges in the Development and Implementation of Artificial Intelligence in Radiology: A Comprehensive Review of Solutions Beyond Supervised Learning

Artificial intelligence (AI) in radiology is a rapidly developing field with several prospective clinical studies demonstrating its benefits in clinical practice. In 2022, the Korean Society of Radiology held a forum to discuss the challenges and drawbacks in AI development and implementation. Various barriers hinder the successful application and widespread adoption of AI in radiology, such as limited annotated data, data privacy and security, data heterogeneity, imbalanced data, model interpretability, overfitting, and integration with clinical workflows. In this review, some of the various possible solutions to these challenges are presented and discussed; these include training with longitudinal and multimodal datasets, dense training with multitask learning and multimodal learning, self-supervised contrastive learning, various image modifications and syntheses using generative models, explainable AI, causal learning, federated learning with large data models, and digital twins.

AI-assisted accelerated MRI of the ankle: clinical practice assessment

BACKGROUND

High-spatial resolution magnetic resonance imaging (MRI) is essential for imaging ankle joints. However, the clinical application of fast spin-echo sequences remains limited by their lengthy acquisition time. Artificial intelligence-assisted compressed sensing (ACS) technology has been recently introduced as an integrative acceleration solution. We compared ACS-accelerated 3-T ankle MRI to conventional methods of compressed sensing (CS) and parallel imaging (PI) .

METHODS

We prospectively included 2 healthy volunteers and 105 patients with ankle pain. ACS acceleration factors for ankle protocol of T1-, T2-, and proton density (PD)-weighted sequences were optimized in a pilot study on healthy volunteers (acceleration factor 3.2-3.3×). Images of patients acquired using ACS and conventional acceleration methods were compared in terms of acquisition times, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), subjective image quality, and diagnostic agreement. Shapiro-Wilk test, Cohen κ, intraclass correlation coefficient, and one-way ANOVA with post hoc tests (Tukey or Dunn) were used.

RESULTS

ACS acceleration reduced the acquisition times of T1-, T2-, and PD-weighted sequences by 32-43%, compared with conventional CS and PI, while maintaining image quality (mostly higher SNR with p < 0.004 and higher CNR with p < 0.047). The diagnostic agreement between ACS and conventional sequences was rated excellent (κ = 1.00).

CONCLUSIONS

The optimum ACS acceleration factors for ankle MRI were found to be 3.2-3.3× protocol. The ACS allows faster imaging, yielding similar image quality and diagnostic performance.

RELEVANCE STATEMENT

AI-assisted compressed sensing significantly accelerates ankle MRI times while preserving image quality and diagnostic precision, potentially expediting patient diagnoses and improving clinical workflows.

KEY POINTS

• AI-assisted compressed sensing (ACS) significantly reduced scan duration for ankle MRI. • Similar image quality achieved by ACS compared to conventional acceleration methods. • A high agreement by three acceleration methods in the diagnosis of ankle lesions was observed.

Adaptive channel-modulated personalized federated learning for magnetic resonance image reconstruction

Magnetic resonance imaging (MRI) is extensively utilized in clinical practice for diagnostic purposes, owing to its non-invasive nature and remarkable ability to provide detailed characterization of soft tissues. However, its drawback lies in the prolonged scanning time. To accelerate MR imaging, how to reconstruct MR images from under-sampled data quickly and accurately has drawn intensive research interest; it, however, remains a challenging task. While some deep learning models have achieved promising performance in MRI reconstruction, these models usually require a substantial quantity of paired data for training, which proves challenging to gather and share owing to high scanning costs and data privacy concerns. Federated learning (FL) is a potential tool to alleviate these difficulties. It enables multiple clinical clients to collaboratively train a global model without compromising privacy. However, it is extremely challenging to fit a single model to diverse data distributions of different clients. Moreover, existing FL algorithms treat the features of each channel equally, lacking discriminative learning ability across feature channels, and hence hindering their representational capability. In this study, we propose a novel Adaptive Channel-Modulated Federal learning framework for personalized MRI reconstruction, dubbed as ACM-FedMRI. Specifically, considering each local client may focus on features in different channels, we first design a client-specific hypernetwork to guide the channel selection operation in order to optimize the extracted features. Additionally, we introduce a performance-based channel decoupling scheme, which dynamically separates the global model at the channel level to facilitate personalized adjustments based on the performance of individual clients. This approach eliminates the need for heuristic design of specific personalization layers. Extensive experiments on four datasets under two different settings show that our ACM-FedMRI achieves outstanding results compared to other cutting-edge federated learning techniques in the field of MRI reconstruction.

Inflating 2D convolution weights for efficient generation of 3D medical images

BACKGROUND AND OBJECTIVE

The generation of three-dimensional (3D) medical images has great application potential since it takes into account the 3D anatomical structure. Two problems prevent effective training of a 3D medical generative model: (1) 3D medical images are expensive to acquire and annotate, resulting in an insufficient number of training images, and (2) a large number of parameters are involved in 3D convolution.

METHODS

We propose a novel GAN model called 3D Split&Shuffle-GAN. To address the 3D data scarcity issue, we first pre-train a two-dimensional (2D) GAN model using abundant image slices and inflate the 2D convolution weights to improve the initialization of the 3D GAN. Novel 3D network architectures are proposed for both the generator and discriminator of the GAN model to significantly reduce the number of parameters while maintaining the quality of image generation. Several weight inflation strategies and parameter-efficient 3D architectures are investigated.

RESULTS

Experiments on both heart (Stanford AIMI Coronary Calcium) and brain (Alzheimer's Disease Neuroimaging Initiative) datasets show that our method leads to improved 3D image generation quality (14.7 improvements on Frchet inception distance) with significantly fewer parameters (only 48.5% of the baseline method).

CONCLUSIONS

We built a parameter-efficient 3D medical image generation model. Due to the efficiency and effectiveness, it has the potential to generate high-quality 3D brain and heart images for real use cases.

CS-MRI Reconstruction Using an Improved GAN with Dilated Residual Networks and Channel Attention Mechanism

Compressed sensing (CS) MRI has shown great potential in enhancing time efficiency. Deep learning techniques, specifically generative adversarial networks (GANs), have emerged as potent tools for speedy CS-MRI reconstruction. Yet, as the complexity of deep learning reconstruction models increases, this can lead to prolonged reconstruction time and challenges in achieving convergence. In this study, we present a novel GAN-based model that delivers superior performance without the model complexity escalating. Our generator module, built on the U-net architecture, incorporates dilated residual (DR) networks, thus expanding the network's receptive field without increasing parameters or computational load. At every step of the downsampling path, this revamped generator module includes a DR network, with the dilation rates adjusted according to the depth of the network layer. Moreover, we have introduced a channel attention mechanism (CAM) to distinguish between channels and reduce background noise, thereby focusing on key information. This mechanism adeptly combines global maximum and average pooling approaches to refine channel attention. We conducted comprehensive experiments with the designed model using public domain MRI datasets of the human brain. Ablation studies affirmed the efficacy of the modified modules within the network. Incorporating DR networks and CAM elevated the peak signal-to-noise ratios (PSNR) of the reconstructed images by about 1.2 and 0.8 dB, respectively, on average, even at 10× CS acceleration. Compared to other relevant models, our proposed model exhibits exceptional performance, achieving not only excellent stability but also outperforming most of the compared networks in terms of PSNR and SSIM. When compared with U-net, DR-CAM-GAN's average gains in SSIM and PSNR were 14% and 15%, respectively. Its MSE was reduced by a factor that ranged from two to seven. The model presents a promising pathway for enhancing the efficiency and quality of CS-MRI reconstruction.

Emerging Trends in Fast MRI Using Deep-Learning Reconstruction on Undersampled k-Space Data: A Systematic Review

Magnetic Resonance Imaging (MRI) is an essential medical imaging modality that provides excellent soft-tissue contrast and high-resolution images of the human body, allowing us to understand detailed information on morphology, structural integrity, and physiologic processes. However, MRI exams usually require lengthy acquisition times. Methods such as parallel MRI and Compressive Sensing (CS) have significantly reduced the MRI acquisition time by acquiring less data through undersampling k-space. The state-of-the-art of fast MRI has recently been redefined by integrating Deep Learning (DL) models with these undersampled approaches. This Systematic Literature Review (SLR) comprehensively analyzes deep MRI reconstruction models, emphasizing the key elements of recently proposed methods and highlighting their strengths and weaknesses. This SLR involves searching and selecting relevant studies from various databases, including Web of Science and Scopus, followed by a rigorous screening and data extraction process using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. It focuses on various techniques, such as residual learning, image representation using encoders and decoders, data-consistency layers, unrolled networks, learned activations, attention modules, plug-and-play priors, diffusion models, and Bayesian methods. This SLR also discusses the use of loss functions and training with adversarial networks to enhance deep MRI reconstruction methods. Moreover, we explore various MRI reconstruction applications, including non-Cartesian reconstruction, super-resolution, dynamic MRI, joint learning of reconstruction with coil sensitivity and sampling, quantitative mapping, and MR fingerprinting. This paper also addresses research questions, provides insights for future directions, and emphasizes robust generalization and artifact handling. Therefore, this SLR serves as a valuable resource for advancing fast MRI, guiding research and development efforts of MRI reconstruction for better image quality and faster data acquisition.

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.