Deep-Learning Methods for Parallel Magnetic Resonance Imaging Reconstruction: A Survey of the Current Approaches, Trends, and Issues

Deep learning MR reconstruction in knees and ankles in children and young adults. Is it ready for clinical use?

OBJECTIVE

To evaluate the diagnostic performance and image quality of accelerated Turbo Spin Echo sequences using deep-learning (DL) reconstructions compared to conventional sequences in knee and ankle MRIs of children and young adults.

MATERIALS AND METHODS

IRB-approved prospective study consisting of 49 MRIs from 48 subjects (10 males, mean age 16.4 years, range 7-29 years), with each MRI consisting of both conventional and DL sequences. Sequences were evaluated blindly to determine predictive values, sensitivity, and specificity of DL sequences using conventional sequences and knee arthroscopy (if available) as references. Physeal patency and appearance were evaluated. Qualitative parameters were compared. Presence of undesired image alterations was assessed.

RESULTS

The prevalence of abnormal findings in the knees and ankles were 11.7% (75/640), and 11.5% (19/165), respectively. Using conventional sequences as reference, sensitivity and specificity of DL sequences in knees were 90.7% and 99.3%, and in ankles were 100.0% and 100.0%. Using arthroscopy as reference, sensitivity and specificity of DL sequences were 80.0% and 95.8%, and of conventional sequences were 80.0% and 97.9%. Agreement of physeal status was 100.0%. DL sequences were qualitatively "same-or-better" compared to conventional (p < 0.032), except for pixelation artifact for the PDFS sequence (p = 0.233). No discrete image alteration was identified in the knee DL sequences. In the ankle, we identified one DL artifact involving a tendon (0.8%, 1/125). DL sequences were faster than conventional sequences by a factor of 2 (p < 0.001).

CONCLUSION

In knee and ankle MRIs, DL sequences provided similar diagnostic performance and "same-or-better" image quality than conventional sequences at half the acquisition time.

Clinical 7 Tesla magnetic resonance imaging: Impact and patient value in neurological disorders

Magnetic resonance imaging (MRI) is a cornerstone of non-invasive diagnostics and treatment monitoring, particularly for diseases of the central nervous system. Although 1.5- and 3 Tesla (T) field strengths remain the clinical standard, the advent of 7 T MRI represents a transformative step forward, offering superior spatial resolution, contrast, and sensitivity for visualizing neuroanatomy, metabolism, and function. Recent innovations, including parallel transmission and deep learning-based reconstruction, have resolved many prior technical challenges of 7 T MRI, enabling its routine clinical use. This review examines the diagnostic impact, patient value, and practical considerations of 7 T MRI, emphasizing its role in facilitating earlier diagnoses and improving care in conditions, such as amyotrophic lateral sclerosis (ALS), epilepsy, multiple sclerosis (MS), dementia, parkinsonism, tumors, and vascular diseases. Based on insights from over 1200 clinical scans with a second-generation 7 T system, the review highlights disease-specific biomarkers such as the motor band sign in ALS and the new diagnostic markers in MS, the central vein sign, and paramagnetic rim lesions. The unparalleled ability of 7 T MRI to study neurological diseases ex vivo at ultra-high resolution is also explored, offering new opportunities to understand pathophysiology and identify novel treatment targets. Additionally, the review provides a clinical perspective on patient handling and safety considerations, addressing challenges and practicalities associated with clinical 7 T MRI. By bridging research and clinical practice, 7 T MRI has the potential to redefine neuroimaging and advance the understanding and management of complex neurological disorders.

Advancing clinical MRI exams with artificial intelligence: Japan's contributions and future prospects

In this narrative review, we review the applications of artificial intelligence (AI) into clinical magnetic resonance imaging (MRI) exams, with a particular focus on Japan's contributions to this field. In the first part of the review, we introduce the various applications of AI in optimizing different aspects of the MRI process, including scan protocols, patient preparation, image acquisition, image reconstruction, and postprocessing techniques. Additionally, we examine AI's growing influence in clinical decision-making, particularly in areas such as segmentation, radiation therapy planning, and reporting assistance. By emphasizing studies conducted in Japan, we highlight the nation's contributions to the advancement of AI in MRI. In the latter part of the review, we highlight the characteristics that make Japan a unique environment for the development and implementation of AI in MRI examinations. Japan's healthcare landscape is distinguished by several key factors that collectively create a fertile ground for AI research and development. Notably, Japan boasts one of the highest densities of MRI scanners per capita globally, ensuring widespread access to the exam. Japan's national health insurance system plays a pivotal role by providing MRI scans to all citizens irrespective of socioeconomic status, which facilitates the collection of inclusive and unbiased imaging data across a diverse population. Japan's extensive health screening programs, coupled with collaborative research initiatives like the Japan Medical Imaging Database (J-MID), enable the aggregation and sharing of large, high-quality datasets. With its technological expertise and healthcare infrastructure, Japan is well-positioned to make meaningful contributions to the MRI-AI domain. The collaborative efforts of researchers, clinicians, and technology experts, including those in Japan, will continue to advance the future of AI in clinical MRI, potentially leading to improvements in patient care and healthcare efficiency.

Single-shot echo planar time-resolved imaging for multi-echo functional MRI and distortion-free diffusion imaging

PURPOSE

To develop a single-shot SNR-efficient distortion-free multi-echo imaging technique for dynamic imaging applications.

METHODS

Echo planar time-resolved imaging (EPTI) was first introduced as a multi-shot technique for distortion-free multi-echo imaging. This work aims to develop single-shot EPTI (ss-EPTI) to achieve improved robustness to motion/physiological noise, increased temporal resolution, and higher SNR efficiency. A new spatiotemporal encoding that enables reduced phase-encoding blips and minimized echo spacing under the single-shot regime was developed, which improves sampling efficiency and enhances spatiotemporal correlation in the k-TE space for improved reconstruction. A continuous readout with minimized deadtime was employed to optimize SNR efficiency. Moreover, k-TE partial Fourier and simultaneous multi-slice acquisition were integrated for further acceleration.

RESULTS

ss-EPTI provided distortion-free imaging with densely sampled multi-echo images at standard resolutions (e.g., ˜1.25 to 3 mm) in a single-shot. Improved SNR efficiency was observed in ss-EPTI due to improved motion/physiological-noise robustness and efficient continuous readout. Its ability to eliminate dynamic distortions-geometric changes across dynamics due to field changes induced by physiological variations or eddy currents-further improved the data's temporal stability. For multi-echo fMRI, ss-EPTI's multi-echo images recovered signal dropout in short-T 2 * $$ {\mathrm{T}}_2^{\ast } $$ regions and provided TE-dependent functional information to distinguish non-BOLD noise for further tSNR improvement. For diffusion MRI, it achieved shortened TEs for improved SNR and provided images free from both B0-induced and diffusion-encoding-dependent eddy-current-induced distortions with multi-TE diffusion metrics.

CONCLUSION

ss-EPTI provides SNR-efficient distortion-free multi-echo imaging with comparable temporal resolutions to ss-EPI, offering a new acquisition tool for dynamic imaging.

The scientific evidence of commercial AI products for MRI acceleration: a systematic review

OBJECTIVES

This study explores the methods employed by commercially available AI products to accelerate MRI protocols and investigates the strength of their diagnostic image quality assessment.

MATERIALS AND METHODS

All commercial AI products for MRI acceleration were identified from the exhibitors presented at the RSNA 2023 and ECR 2024 annual meetings. Peer-reviewed scientific articles describing validation of clinical performance were searched for each product. Information was extracted regarding the MRI acceleration technique, achieved acceleration, diagnostic performance metrics, test cohort, and hallucinatory artifacts. The strength of the diagnostic image quality was assessed using scientific evidence levels ranging from "product's technical feasibility for clinical purposes" to "product's economic impact on society".

RESULTS

Out of 1046 companies, 14 products of 14 companies were included. No scientific articles were found for four products (29%). For the remaining ten products (71%), 21 articles were retrieved. Four acceleration methods were identified: noise reduction, raw data reconstruction, personalized scanning protocols, and synthetic image generation. Only a limited number of articles prospectively demonstrated impact on patient outcomes (n = 4, 19%), and no articles discussed an evaluation in a prospective cohort of > 100 patients or performed an economic analysis. None of the articles performed an analysis of hallucinatory artifacts.

CONCLUSION

Currently, commercially available AI products for MRI acceleration can be categorized into four main methods. The acceleration methods lack prospective scientific evidence on clinical performance in large cohorts and economic analysis, which would help to get a better insight into their diagnostic performance and enable safe and effective clinical implementation.

KEY POINTS

Question There is a growing interest in AI products that reduce MRI scan time, but an overview of these methods and their scientific evidence is missing. Findings Only a limited number of articles (n = 4, 19%) prospectively demonstrated the impact of the software for accelerating MRI on diagnostic performance metrics. Clinical relevance Although various commercially available products shorten MRI acquisition time, more studies in large cohorts are needed to get a better insight into the diagnostic performance of AI-constructed MRI.

Physics-guided self-supervised learning: Demonstration for generalized RF pulse design

PURPOSE

To introduce a new method for generalized RF pulse design using physics-guided self-supervised learning (GPS), which uses the Bloch equations as the guiding physics model.

THEORY AND METHODS

The GPS framework consists of a neural network module and a physics module, where the physics module is a Bloch simulator for MRI applications. For RF pulse design, the neural network module maps an input target profile to an RF pulse, which is subsequently loaded into the physics module. Through the supervision of the physics module, the neural network module designs an RF pulse corresponding to the target profile. GPS was applied to design 1D selective,B 1 $$ {B}_1 $$ -insensitive, saturation, and multidimensional RF pulses, each conventionally requiring dedicated design algorithms. We further demonstrate our method's flexibility and versatility by compensating for experimental and scanner imperfections through online adaptation.

RESULTS

Both simulations and experiments show that GPS can design a variety of RF pulses with corresponding profiles that agree well with the target input. Despite these verifications, GPS-designed pulses have unique differences compared to conventional designs, such as achievingB 1 $$ {B}_1 $$ -insensitivity using different mechanisms and using non-sampled regions of the conventional design to lower its peak power. Experiments, both ex vivo and in vivo, further verify that it can also be used for online adaptation to correct system imperfections, such asB 0 $$ {B}_0 $$ /B 1 + $$ {B}_1^{+} $$ inhomogeneity.

CONCLUSION

This work demonstrates the generalizability, versatility, and flexibility of the GPS method for designing RF pulses and showcases its utility in several applications.

Unsupervised reconstruction of accelerated cardiac cine MRI using neural fields

BACKGROUND

Cardiac cine MRI is the gold standard for cardiac functional assessment, but the inherently slow acquisition process creates the necessity of reconstruction approaches for accelerated undersampled acquisitions. Several regularization approaches that exploit spatial-temporal redundancy have been proposed to reconstruct undersampled cardiac cine MRI. More recently, methods based on supervised deep learning have been also proposed to further accelerate acquisition and reconstruction. However, these techniques rely on usually large dataset for training, which are not always available and might introduce biases.

METHODS

In this work we propose NF-cMRI, an unsupervised approach based on implicit neural field representations for cardiac cine MRI. We evaluate our method in in-vivo undersampled golden-angle radial multi-coil acquisitions for undersampling factors of 13x, 17x and 26x.

RESULTS

The proposed method achieves excellent scores in sharpness and robustness to artifacts and comparable or improved spatial-temporal depiction than state-of-the-art conventional and unsupervised deep learning reconstruction techniques.

CONCLUSIONS

We have demonstrated NF-cMRI potential for cardiac cine MRI reconstruction with highly undersampled data.

Deep-ER: Deep Learning ECCENTRIC Reconstruction for fast high-resolution neurometabolic imaging

INTRODUCTION

Altered neurometabolism is an important pathological mechanism in many neurological diseases and brain cancer, which can be mapped non-invasively by Magnetic Resonance Spectroscopic Imaging (MRSI). Advanced MRSI using non-cartesian compressed-sense acquisition enables fast high-resolution metabolic imaging but has lengthy reconstruction times that limits throughput and needs expert user interaction. Here, we present a robust and efficient Deep Learning reconstruction embedded in a physical model within an end-to-end automated processing pipeline to obtain high-quality metabolic maps.

METHODS

Fast high-resolution whole-brain metabolic imaging was performed at 3.4 mm3 isotropic resolution with acquisition times between 4:11-9:21 min:s using ECCENTRIC pulse sequence on a 7T MRI scanner. Data were acquired in a high-resolution phantom and 27 human participants, including 22 healthy volunteers and 5 glioma patients. A deep neural network using recurring interlaced convolutional layers with joint dual-space feature representation was developed for deep learning ECCENTRIC reconstruction (Deep-ER). 21 subjects were used for training and 6 subjects for testing. Deep-ER performance was compared to iterative compressed sensing Total Generalized Variation reconstruction using image and spectral quality metrics.

RESULTS

Deep-ER demonstrated 600-fold faster reconstruction than conventional methods, providing improved spatial-spectral quality and metabolite quantification with 12%-45% (P<0.05) higher signal-to-noise and 8%-50% (P<0.05) smaller Cramer-Rao lower bounds. Metabolic images clearly visualize glioma tumor heterogeneity and boundary. Deep-ER generalizes reliably to unseen data.

CONCLUSION

Deep-ER provides efficient and robust reconstruction for sparse-sampled MRSI. The accelerated acquisition-reconstruction MRSI is compatible with high-throughput imaging workflow. It is expected that such improved performance will facilitate basic and clinical MRSI applications for neuroscience and precision medicine.

Accelerating multi-coil MR image reconstruction using weak supervision

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

Multi-task magnetic resonance imaging reconstruction using meta-learning

Using single-task deep learning methods to reconstruct Magnetic Resonance Imaging (MRI) data acquired with different imaging sequences is inherently challenging. The trained deep learning model typically lacks generalizability, and the dissimilarity among image datasets with different types of contrast leads to suboptimal learning performance. This paper proposes a meta-learning approach to efficiently learn image features from multiple MRI datasets. Our algorithm can perform multi-task learning to simultaneously reconstruct MRI images acquired using different imaging sequences with various image contrasts. We have developed a proximal gradient descent-inspired optimization method to learn image features across image and k-space domains, ensuring high-performance learning for every image contrast. Meanwhile, meta-learning, a "learning-to-learn" process, is incorporated into our framework to improve the learning of mutual features embedded in multiple image contrasts. The experimental results reveal that our proposed multi-task meta-learning approach surpasses state-of-the-art single-task learning methods at high acceleration rates. Our meta-learning consistently delivers accurate and detailed reconstructions, achieves the lowest pixel-wise errors, and significantly enhances qualitative performance across all tested acceleration rates. We have demonstrated the ability of our new meta-learning reconstruction method to successfully reconstruct highly-undersampled k-space data from multiple MRI datasets simultaneously, outperforming other compelling reconstruction methods previously developed for single-task learning.

Data- and Physics-Driven Deep Learning Based Reconstruction for Fast MRI: Fundamentals and Methodologies

Magnetic Resonance Imaging (MRI) is a pivotal clinical diagnostic tool, yet its extended scanning times often compromise patient comfort and image quality, especially in volumetric, temporal and quantitative scans. This review elucidates recent advances in MRI acceleration via data and physics-driven models, leveraging techniques from algorithm unrolling models, enhancement-based methods, and plug-and-play models to the emerging full spectrum of generative model-based methods. We also explore the synergistic integration of data models with physics-based insights, encompassing the advancements in multi-coil hardware accelerations like parallel imaging and simultaneous multi-slice imaging, and the optimization of sampling patterns. We then focus on domain-specific challenges and opportunities, including image redundancy exploitation, image integrity, evaluation metrics, data heterogeneity, and model generalization. This work also discusses potential solutions and future research directions, with an emphasis on the role of data harmonization and federated learning for further improving the general applicability and performance of these methods in MRI reconstruction.

Cardiac MR image reconstruction using cascaded hybrid dual domain deep learning framework

Recovering diagnostic-quality cardiac MR images from highly under-sampled data is a current research focus, particularly in addressing cardiac and respiratory motion. Techniques such as Compressed Sensing (CS) and Parallel Imaging (pMRI) have been proposed to accelerate MRI data acquisition and improve image quality. However, these methods have limitations in high spatial-resolution applications, often resulting in blurring or residual artifacts. Recently, deep learning-based techniques have gained attention for their accuracy and efficiency in image reconstruction. Deep learning-based MR image reconstruction methods are divided into two categories: (a) single domain methods (image domain learning and k-space domain learning) and (b) cross/dual domain methods. Single domain methods, which typically use U-Net in either the image or k-space domain, fail to fully exploit the correlation between these domains. This paper introduces a dual-domain deep learning approach that incorporates multi-coil data consistency (MCDC) layers for reconstructing cardiac MR images from 1-D Variable Density (VD) random under-sampled data. The proposed hybrid dual-domain deep learning models integrate data from both the domains to improve image quality, reduce artifacts, and enhance overall robustness and accuracy of the reconstruction process. Experimental results demonstrate that the proposed methods outperform than conventional deep learning and CS techniques, as evidenced by higher Structural Similarity Index (SSIM), lower Root Mean Square Error (RMSE), and higher Peak Signal-to-Noise Ratio (PSNR).

MRI at low field: A review of software solutions for improving SNR

Low magnetic field magnetic resonance imaging (MRI) (B 0 $$ {B}_0 $$  < 1 T) is regaining interest in the magnetic resonance (MR) community as a complementary, more flexible, and cost-effective approach to MRI diagnosis. Yet, the impaired signal-to-noise ratio (SNR) per square root of time, or SNR efficiency, leading in turn to prolonged acquisition times, still challenges its relevance at the clinical level. To address this, researchers investigate various hardware and software solutions to improve SNR efficiency at low field, including the leveraging of latest advances in computing hardware. However, there may not be a single recipe for improving SNR at low field, and it is key to embrace the challenges and limitations of each proposed solution. In other words, suitable solutions depend on the final objective or application envisioned for a low-field scanner and, more importantly, on the characteristics of a specific lowB 0 $$ {B}_0 $$ field. In this review, we aim to provide an overview on software solutions to improve SNR efficiency at low field. First, we cover techniques for efficient k-space sampling and reconstruction. Then, we present post-acquisition techniques that enhance MR images such as denoising and super-resolution. In addition, we summarize recently introduced electromagnetic interference cancellation approaches showing great promises when operating in shielding-free environments. Finally, we discuss the advantages and limitations of these approaches that could provide directions for future applications.

Computationally Efficient Implicit Training Strategy for Unrolled Networks (IMUNNE): A Preliminary Analysis Using Accelerated Real-Time Cardiac Cine MRI

OBJECTIVE

Highly-undersampled, dynamic MRI reconstruction, particularly in multi-coil scenarios, is a challenging inverse problem. Unrolled networks achieve state-of-the-art performance in MRI reconstruction but suffer from long training times and extensive GPU memory cost.

METHODS

In this work, we propose a novel training strategy for IMplicit UNrolled NEtworks (IMUNNE) for highly-undersampled, multi-coil dynamic MRI reconstruction. It formulates the MRI reconstruction problem as an implicit fixed-point equation and leverages gradient approximation for backpropagation, enabling training of deep architectures with fixed memory cost. This study represents the first application of implicit network theory in the context of real-time cine MRI. The proposed method is evaluated using a prospectively undersampled, real-time cine dataset using radial k-space sampling, comprising balanced steady-state free precession (b-SSFP) readouts. Experiments include a hyperparameter search, head-to-head comparisons with a complex U-Net (CU-Net) and an alternating unrolled network (Alt-UN), and an analysis of robustness under noise perturbations; peak signal-to-noise ratio, structural similarity index, normalized root mean-square error, spatio-temporal entropic difference, and a blur metric were used.

RESULTS

IMUNNE produced significantly and slightly better image quality compared to CU-Net and Alt-UN, respectively. Compared with Alt-UN, IMUNNE significantly reduced training and inference times, making it a promising approach for highly-accelerated, multi-coil real-time cine MRI reconstruction.

CONCLUSION

IMUNNE strategy successfully applies unrolled networks to image reconstruction of highly-accelerated, real-time radial cine MRI.

SIGNIFICANCE

Implicit training enables rapid, high-quality, and cost-effective CMR exams by reducing training and inference times and lowering memory cost associated with advanced reconstruction methods.

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.

Machine Learning and Deep Learning Applications in Magnetic Particle Imaging

In recent years, magnetic particle imaging (MPI) has emerged as a promising imaging technique depicting high sensitivity and spatial resolution. It originated in the early 2000s where it proposed a new approach to challenge the low spatial resolution achieved by using relaxometry in order to measure the magnetic fields. MPI presents 2D and 3D images with high temporal resolution, non-ionizing radiation, and optimal visual contrast due to its lack of background tissue signal. Traditionally, the images were reconstructed by the conversion of signal from the induced voltage by generating system matrix and X-space based methods. Because image reconstruction and analyses play an integral role in obtaining precise information from MPI signals, newer artificial intelligence-based methods are continuously being researched and developed upon. In this work, we summarize and review the significance and employment of machine learning and deep learning models for applications with MPI and the potential they hold for the future. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY: Stage 1.

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.

Deep learning methods for 3D magnetic resonance image denoising, bias field and motion artifact correction: a comprehensive review

Magnetic resonance imaging (MRI) provides detailed structural information of the internal body organs and soft tissue regions of a patient in clinical diagnosis for disease detection, localization, and progress monitoring. MRI scanner hardware manufacturers incorporate various post-acquisition image-processing techniques into the scanner's computer software tools for different post-processing tasks. These tools provide a final image of adequate quality and essential features for accurate clinical reporting and predictive interpretation for better treatment planning. Different post-acquisition image-processing tasks for MRI quality enhancement include noise removal, motion artifact reduction, magnetic bias field correction, and eddy electric current effect removal. Recently, deep learning (DL) methods have shown great success in many research fields, including image and video applications. DL-based data-driven feature-learning approaches have great potential for MR image denoising and image-quality-degrading artifact correction. Recent studies have demonstrated significant improvements in image-analysis tasks using DL-based convolutional neural network techniques. The promising capabilities and performance of DL techniques in various problem-solving domains have motivated researchers to adapt DL methods to medical image analysis and quality enhancement tasks. This paper presents a comprehensive review of DL-based state-of-the-art MRI quality enhancement and artifact removal methods for regenerating high-quality images while preserving essential anatomical and physiological feature maps without destroying important image information. Existing research gaps and future directions have also been provided by highlighting potential research areas for future developments, along with their importance and advantages in medical imaging.

Zero-Shot Adaptation for Approximate Posterior Sampling of Diffusion Models in Inverse Problems

Diffusion models have emerged as powerful generative techniques for solving inverse problems. Despite their success in a variety of inverse problems in imaging, these models require many steps to converge, leading to slow inference time. Recently, there has been a trend in diffusion models for employing sophisticated noise schedules that involve more frequent iterations of timesteps at lower noise levels, thereby improving image generation and convergence speed. However, application of these ideas for solving inverse problems with diffusion models remain challenging, as these noise schedules do not perform well when using empirical tuning for the forward model log-likelihood term weights. To tackle these challenges, we propose zero-shot approximate posterior sampling (ZAPS) that leverages connections to zero-shot physics-driven deep learning. ZAPS fixes the number of sampling steps, and uses zero-shot training with a physics-guided loss function to learn log-likelihood weights at each irregular timestep. We apply ZAPS to the recently proposed diffusion posterior sampling method as baseline, though ZAPS can also be used with other posterior sampling diffusion models. We further approximate the Hessian of the logarithm of the prior using a diagonalization approach with learnable diagonal entries for computational efficiency. These parameters are optimized over a fixed number of epochs with a given computational budget. Our results for various noisy inverse problems, including Gaussian and motion deblurring, inpainting, and super-resolution show that ZAPS reduces inference time, provides robustness to irregular noise schedules and improves reconstruction quality. Code is available at https://github.com/ualcalar17/ZAPS.

Comparison of image quality and lesion conspicuity between conventional and deep learning reconstruction in gadoxetic acid-enhanced liver MRI

OBJECTIVE

To compare the image quality and lesion conspicuity of conventional vs deep learning (DL)-based reconstructed three-dimensional T1-weighted images in gadoxetic acid-enhanced liver magnetic resonance imaging (MRI).

METHODS

This prospective study (NCT05182099) enrolled participants scheduled for gadoxetic acid-enhanced liver MRI due to suspected focal liver lesions (FLLs) who provided signed informed consent. A liver MRI was conducted using a 3-T scanner. T1-weighted images were reconstructed using both conventional and DL-based (AIRTM Recon DL 3D) reconstruction algorithms. Three radiologists independently reviewed the image quality and lesion conspicuity on a 5-point scale.

RESULTS

Fifty participants (male = 36, mean age 62 ± 11 years) were included for image analysis. The DL-based reconstruction showed significantly higher image quality than conventional images in all phases (3.71-4.40 vs 3.37-3.99, p < 0.001 for all), as well as significantly less noise and ringing artifacts than conventional images (p < 0.05 for all), while also showing significantly altered image texture (p < 0.001 for all). Lesion conspicuity was significantly higher in DL-reconstructed images than in conventional images in the arterial phase (2.15 [95% confidence interval: 1.78, 2.52] vs 2.03 [1.65, 2.40], p = 0.036), but no significant difference was observed in the portal venous phase and hepatobiliary phase (p > 0.05 for all). There was no significant difference in the figure-of-merit (0.728 in DL vs 0.709 in conventional image, p = 0.474).

CONCLUSION

DL reconstruction provided higher-quality three-dimensional T1-weighted imaging than conventional reconstruction in gadoxetic acid-enhanced liver MRI.

CRITICAL RELEVANCE STATEMENT

DL reconstruction of 3D T1-weighted images improves image quality and arterial phase lesion conspicuity in gadoxetic acid-enhanced liver MRI compared to conventional reconstruction.

KEY POINTS

DL reconstruction is feasible for 3D T1-weighted images across different spatial resolutions and phases. DL reconstruction showed superior image quality with reduced noise and ringing artifacts. Hepatic anatomic structures were more conspicuous on DL-reconstructed images.

Effect of MR head coil geometry on deep-learning-based MR image reconstruction

PURPOSE

To investigate whether parallel imaging-imposed geometric coil constraints can be relaxed when using a deep learning (DL)-based image reconstruction method as opposed to a traditional non-DL method.

THEORY AND METHODS

Traditional and DL-based MR image reconstruction approaches operate in fundamentally different ways: Traditional methods solve a system of equations derived from the image data whereas DL methods use data/target pairs to learn a generalizable reconstruction model. Two sets of head coil profiles were evaluated: (1) 8-channel and (2) 32-channel geometries. A DL model was compared to conjugate gradient SENSE (CG-SENSE) and L1-wavelet compressed sensing (CS) through quantitative metrics and visual assessment as coil overlap was increased.

RESULTS

Results were generally consistent between experiments. As coil overlap increased, there was a significant (p < 0.001) decrease in performance in most cases for all methods. The decrease was most pronounced for CG-SENSE, and the DL models significantly outperformed (p < 0.001) their non-DL counterparts in all scenarios. CS showed improved robustness to coil overlap and signal-to-noise ratio (SNR) versus CG-SENSE, but had quantitatively and visually poorer reconstructions characterized by blurriness as compared to DL. DL showed virtually no change in performance across SNR and very small changes across coil overlap.

CONCLUSION

The DL image reconstruction method produced images that were robust to coil overlap and of higher quality than CG-SENSE and CS. This suggests that geometric coil design constraints can be relaxed when using DL reconstruction methods.

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.

Towards retrospective motion correction and reconstruction for clinical 3D brain MRI protocols with a reference contrast

OBJECT

In a typical MR session, several contrasts are acquired. Due to the sequential nature of the data acquisition process, the patient may experience some discomfort at some point during the session, and start moving. Hence, it is quite common to have MR sessions where some contrasts are well-resolved, while other contrasts exhibit motion artifacts. Instead of repeating the scans that are corrupted by motion, we introduce a reference-guided retrospective motion correction scheme that takes advantage of the motion-free scans, based on a generalized rigid registration routine.

MATERIALS AND METHODS

We focus on various existing clinical 3D brain protocols at 1.5 Tesla MRI based on Cartesian sampling. Controlled experiments with three healthy volunteers and three levels of motion are performed.

RESULTS

Radiological inspection confirms that the proposed method consistently ameliorates the corrupted scans. Furthermore, for the set of specific motion tests performed in this study, the quality indexes based on PSNR and SSIM shows only a modest decrease in correction quality as a function of motion complexity.

DISCUSSION

While the results on controlled experiments are positive, future applications to patient data will ultimately clarify whether the proposed correction scheme satisfies the radiological requirements.

SPICER: Self-supervised learning for MRI with automatic coil sensitivity estimation and reconstruction

PURPOSE

To introduce a novel deep model-based architecture (DMBA), SPICER, that uses pairs of noisy and undersampled k-space measurements of the same object to jointly train a model for MRI reconstruction and automatic coil sensitivity estimation.

METHODS

SPICER consists of two modules to simultaneously reconstructs accurate MR images and estimates high-quality coil sensitivity maps (CSMs). The first module, CSM estimation module, uses a convolutional neural network (CNN) to estimate CSMs from the raw measurements. The second module, DMBA-based MRI reconstruction module, forms reconstructed images from the input measurements and the estimated CSMs using both the physical measurement model and learned CNN prior. With the benefit of our self-supervised learning strategy, SPICER can be efficiently trained without any fully sampled reference data.

RESULTS

We validate SPICER on both open-access datasets and experimentally collected data, showing that it can achieve state-of-the-art performance in highly accelerated data acquisition settings (up to10 × $$ 10\times $$ ). Our results also highlight the importance of different modules of SPICER-including the DMBA, the CSM estimation, and the SPICER training loss-on the final performance of the method. Moreover, SPICER can estimate better CSMs than pre-estimation methods especially when the ACS data is limited.

CONCLUSION

Despite being trained on noisy undersampled data, SPICER can reconstruct high-quality images and CSMs in highly undersampled settings, which outperforms other self-supervised learning methods and matches the performance of the well-known E2E-VarNet trained on fully sampled ground-truth data.

Deep learning reconstructed T2-weighted Dixon imaging of the spine: Impact on acquisition time and image quality

PURPOSE

To assess the image quality and impact on acquisition time of a novel deep learning based T2 Dixon sequence (T2DL) of the spine.

METHODS

This prospective, single center study included n = 44 consecutive patients with a clinical indication for lumbar MRI at our university radiology department between September 2022 and March 2023. MRI examinations were performed on 1.5-T and 3-T scanners (MAGNETOM Aera and Vida; Siemens Healthineers, Erlangen, Germany) using dedicated spine coils. The MR study protocol consisted of our standard clinical protocol, including a T2 weighted standard Dixon sequence (T2std) and an additional T2DL acquisition. The latter used a conventional sampling pattern with a higher parallel acceleration factor. The individual contrasts acquired for Dixon water-fat separation were then reconstructed using a dedicated research application. After reconstruction of the contrast images from k-space data, a conventional water-fat separation was performed to provide derived water images. Two readers with 6 and 4 years of experience in interpreting MSK imaging, respectively, analyzed the images in a randomized fashion. Regarding overall image quality, banding artifacts, artifacts, sharpness, noise, and diagnostic confidence were analyzed using a 5-point Likert scale (from 1 = non-diagnostic to 5 = excellent image quality). Statistical analyses included the Wilcoxon signed-rank test and weighted Cohen's kappa statistics.

RESULTS

Forty-four patients (mean age 53 years (±18), male sex: 39 %) were prospectively included. Thirty-one examinations were performed on 1.5 T and 13 examinations on 3 T scanners. A sequence was successfully acquired in all patients. The total acquisition time of T2DL was 93 s at 1.5-T and 86 s at 3-T, compared to 235 s, and 257 s, respectively for T2std (reduction of acquisition time: 60.4 % at 1.5-T, and 66.5 % at 3-T; p < 0.01). Overall image quality was rated equal for both sequences (median T2DL: 5[3 -5], and median T2std: 5 [2 -5]; p = 0.57). T2DL showed significantly reduced noise levels compared to T2std (5 [4 -5] versus 4 [3 -4]; p < 0.001). In addition, sharpness was rated to be significantly higher in T2DL (5 [4 -5] versus 4 [3 -5]; p < 0.001). Although T2DL displayed significantly more banding artifacts (5 [2 -5] versus 5 [4 -5]; p < 0.001), no significant impact on readers diagnostic confidence between sequences was noted (T2std: 5 [2 -5], and T2DL: 5 [3 -5]; p = 0.61). Substantial inter-reader and intrareader agreement was observed for T2DL overall image quality (κ: 0.77, and κ: 0.8, respectively).

CONCLUSION

T2DL is feasible, yields an image quality comparable to the reference standard while substantially reducing the acquisition time.

The "hidden noise" problem in MR image reconstruction

PURPOSE

The performance of modern image reconstruction methods is commonly judged using quantitative error metrics like root mean squared-error and the structural similarity index, which are calculated by comparing reconstructed images against fully sampled reference data. In practice, the reference data will contain noise and is not a true gold standard. In this work, we demonstrate that the "hidden noise" present in reference data can substantially confound standard approaches for ranking different image reconstruction results.

METHODS

Using both experimental and simulated k-space data and several different image reconstruction techniques, we examined whether there was correlation between performance metrics obtained with typical noisy reference data versus those obtained with higher-quality reference data.

RESULTS

For conventional performance metrics, the reconstructions that matched best with the higher-quality reference data were substantially different from the reconstructions that matched best with typical noisy reference data. This leads to suboptimal reconstruction results if the performance with respect to noisy reference data is used to select which reconstruction methods/parameters to employ. These issues were reduced when employing alternative error metrics that better account for noise.

CONCLUSION

Reference data containing hidden noise can substantially mislead the ranking of image reconstruction methods when using conventional error metrics, but this issue can be mitigated with alternative error metrics.

Enhancing quality and speed in database-free neural network reconstructions of undersampled MRI with SCAMPI

PURPOSE

We present SCAMPI (Sparsity Constrained Application of deep Magnetic resonance Priors for Image reconstruction), an untrained deep Neural Network for MRI reconstruction without previous training on datasets. It expands the Deep Image Prior approach with a multidomain, sparsity-enforcing loss function to achieve higher image quality at a faster convergence speed than previously reported methods.

METHODS

Two-dimensional MRI data from the FastMRI dataset with Cartesian undersampling in phase-encoding direction were reconstructed for different acceleration rates for single coil and multicoil data.

RESULTS

The performance of our architecture was compared to state-of-the-art Compressed Sensing methods and ConvDecoder, another untrained Neural Network for two-dimensional MRI reconstruction. SCAMPI outperforms these by better reducing undersampling artifacts and yielding lower error metrics in multicoil imaging. In comparison to ConvDecoder, the U-Net architecture combined with an elaborated loss-function allows for much faster convergence at higher image quality. SCAMPI can reconstruct multicoil data without explicit knowledge of coil sensitivity profiles. Moreover, it is a novel tool for reconstructing undersampled single coil k-space data.

CONCLUSION

Our approach avoids overfitting to dataset features, that can occur in Neural Networks trained on databases, because the network parameters are tuned only on the reconstruction data. It allows better results and faster reconstruction than the baseline untrained Neural Network approach.

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.

Evaluation of deep-learning TSE images in clinical musculoskeletal imaging

In this study, we compared the fat-saturated (FS) and non-FS turbo spin echo (TSE) magnetic resonance imaging knee sequences reconstructed conventionally (conventional-TSE) against a deep learning-based reconstruction of accelerated TSE (DL-TSE) scans. A total of 232 conventional-TSE and DL-TSE image pairs were acquired for comparison. For each consenting patient, one of the clinically acquired conventional-TSE proton density-weighted sequences in the sagittal or coronal planes (FS and non-FS), or in the axial plane (non-FS), was repeated using a research DL-TSE sequence. The DL-TSE reconstruction resulted in an image resolution that increased by at least 45% and scan times that were up to 52% faster compared to the conventional TSE. All images were acquired on a MAGNETOM Vida 3T scanner (Siemens Healthineers AG, Erlangen, Germany). The reporting radiologists, blinded to the acquisition time, were requested to qualitatively compare the DL-TSE against the conventional-TSE reconstructions. Despite having a faster acquisition time, the DL-TSE was rated to depict smaller structures better for 139/232 (60%) cases, equivalent for 72/232 (31%) cases and worse for 21/232 (9%) cases compared to the conventional-TSE. Overall, the radiologists preferred the DL-TSE reconstruction in 124/232 (53%) cases and stated no preference, implying equivalence, for 65/232 (28%) cases. DL-TSE reconstructions enabled faster acquisition times while enhancing spatial resolution and preserving the image contrast. From these results, the DL-TSE provided added or comparable clinical value and utility in less time. DL-TSE offers the opportunity to further reduce the overall examination time and improve patient comfort with no loss in diagnostic accuracy.

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

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

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.

Large-scale 3D non-Cartesian coronary MRI reconstruction using distributed memory-efficient physics-guided deep learning with limited training data

OBJECT

To enable high-quality physics-guided deep learning (PG-DL) reconstruction of large-scale 3D non-Cartesian coronary MRI by overcoming challenges of hardware limitations and limited training data availability.

MATERIALS AND METHODS

While PG-DL has emerged as a powerful image reconstruction method, its application to large-scale 3D non-Cartesian MRI is hindered by hardware limitations and limited availability of training data. We combine several recent advances in deep learning and MRI reconstruction to tackle the former challenge, and we further propose a 2.5D reconstruction using 2D convolutional neural networks, which treat 3D volumes as batches of 2D images to train the network with a limited amount of training data. Both 3D and 2.5D variants of the PG-DL networks were compared to conventional methods for high-resolution 3D kooshball coronary MRI.

RESULTS

Proposed PG-DL reconstructions of 3D non-Cartesian coronary MRI with 3D and 2.5D processing outperformed all conventional methods both quantitatively and qualitatively in terms of image assessment by an experienced cardiologist. The 2.5D variant further improved vessel sharpness compared to 3D processing, and scored higher in terms of qualitative image quality.

DISCUSSION

PG-DL reconstruction of large-scale 3D non-Cartesian MRI without compromising image size or network complexity is achieved, and the proposed 2.5D processing enables high-quality reconstruction with limited training data.

Improving quantitative MRI using self-supervised deep learning with model reinforcement: Demonstration for rapid T1 mapping

PURPOSE

This paper proposes a novel self-supervised learning framework that uses model reinforcement, REference-free LAtent map eXtraction with MOdel REinforcement (RELAX-MORE), for accelerated quantitative MRI (qMRI) reconstruction. The proposed method uses an optimization algorithm to unroll an iterative model-based qMRI reconstruction into a deep learning framework, enabling accelerated MR parameter maps that are highly accurate and robust.

METHODS

Unlike conventional deep learning methods which require large amounts of training data, RELAX-MORE is a subject-specific method that can be trained on single-subject data through self-supervised learning, making it accessible and practically applicable to many qMRI studies. Using quantitativeT 1 $$ {\mathrm{T}}_1 $$ mapping as an example, the proposed method was applied to the brain, knee and phantom data.

RESULTS

The proposed method generates high-quality MR parameter maps that correct for image artifacts, removes noise, and recovers image features in regions of imperfect image conditions. Compared with other state-of-the-art conventional and deep learning methods, RELAX-MORE significantly improves efficiency, accuracy, robustness, and generalizability for rapid MR parameter mapping.

CONCLUSION

This work demonstrates the feasibility of a new self-supervised learning method for rapid MR parameter mapping, that is readily adaptable to the clinical translation of qMRI.

The beating heart: artificial intelligence for cardiovascular application in the clinic

Artificial intelligence (AI) integration in cardiac magnetic resonance imaging presents new and exciting avenues for advancing patient care, automating post-processing tasks, and enhancing diagnostic precision and outcomes. The use of AI significantly streamlines the examination workflow through the reduction of acquisition and postprocessing durations, coupled with the automation of scan planning and acquisition parameters selection. This has led to a notable improvement in examination workflow efficiency, a reduction in operator variability, and an enhancement in overall image quality. Importantly, AI unlocks new possibilities to achieve spatial resolutions that were previously unattainable in patients. Furthermore, the potential for low-dose and contrast-agent-free imaging represents a stride toward safer and more patient-friendly diagnostic procedures. Beyond these benefits, AI facilitates precise risk stratification and prognosis evaluation by adeptly analysing extensive datasets. This comprehensive review article explores recent applications of AI in the realm of cardiac magnetic resonance imaging, offering insights into its transformative potential in the field.

Comparative analysis of image quality and interchangeability between standard and deep learning-reconstructed T2-weighted spine MRI

RATIONALE AND OBJECTIVES

MRI reconstruction of undersampled data using a deep learning (DL) network has been recently performed as part of accelerated imaging. Herein, we compared DL-reconstructed T2-weighted image (T2-WI) to conventional T2-WI regarding image quality and degenerative lesion detection.

MATERIALS AND METHODS

Sixty-two patients underwent C-spine (n = 27) or L-spine (n = 35) MRIs, including conventional and DL-reconstructed T2-WI. Image quality was assessed with non-uniformity measurement and 4-scale grading of structural visibility. Three readers (R1, R2, R3) independently assessed the presence and types of degenerative lesions. Student t-test was used to compare non-uniformity measurements. Interprotocol and interobserver agreement of structural visibility was analyzed with Wilcoxon signed-rank test and weighted-κ values, respectively. The diagnostic equivalence of degenerative lesion detection between two protocols was assessed with interchangeability test.

RESULTS

The acquisition time of DL-reconstructed images was reduced to about 21-58% compared to conventional images. Non-uniformity measurement was insignificantly different between the two images (p-value = 0.17). All readers rated DL-reconstructed images as showing the same or superior structural visibility compared to conventional images. Significantly improved visibility was observed at disk margin of C-spine (R1, p < 0.001; R2, p = 0.04) and dorsal root ganglia (R1, p = 0.03; R3, p = 0.02) and facet joint (R1, p = 0.04; R2, p < 0.001; R3, p = 0.03) of L-spine. Interobserver agreements of image quality were variable in each structure. Clinical interchangeability between two protocols for degenerative lesion detection was verified showing <5% in the upper bounds of 95% confidence intervals of agreement rate differences.

CONCLUSIONS

DL-reconstructed T2-WI demonstrates comparable image quality and diagnostic performance with conventional T2-WI in spine imaging, with reduced acquisition time.

Scan-Specific Self-Supervised Bayesian Deep Non-Linear Inversion for Undersampled MRI Reconstruction

Magnetic resonance imaging is subject to slow acquisition times due to the inherent limitations in data sampling. Recently, supervised deep learning has emerged as a promising technique for reconstructing sub-sampled MRI. However, supervised deep learning requires a large dataset of fully-sampled data. Although unsupervised or self-supervised deep learning methods have emerged to address the limitations of supervised deep learning approaches, they still require a database of images. In contrast, scan-specific deep learning methods learn and reconstruct using only the sub-sampled data from a single scan. Here, we introduce Scan-Specific Self-Supervised Bayesian Deep Non-Linear Inversion (DNLINV) that does not require an auto calibration scan region. DNLINV utilizes a Deep Image Prior-type generative modeling approach and relies on approximate Bayesian inference to regularize the deep convolutional neural network. We demonstrate our approach on several anatomies, contrasts, and sampling patterns and show improved performance over existing approaches in scan-specific calibrationless parallel imaging and compressed sensing.

Zero-DeepSub: Zero-shot deep subspace reconstruction for rapid multiparametric quantitative MRI using 3D-QALAS

PURPOSE

To develop and evaluate methods for (1) reconstructing 3D-quantification using an interleaved Look-Locker acquisition sequence with T2 preparation pulse (3D-QALAS) time-series images using a low-rank subspace method, which enables accurate and rapid T1 and T2 mapping, and (2) improving the fidelity of subspace QALAS by combining scan-specific deep-learning-based reconstruction and subspace modeling.

THEORY AND METHODS

A low-rank subspace method for 3D-QALAS (i.e., subspace QALAS) and zero-shot deep-learning subspace method (i.e., Zero-DeepSub) were proposed for rapid and high fidelity T1 and T2 mapping and time-resolved imaging using 3D-QALAS. Using an ISMRM/NIST system phantom, the accuracy and reproducibility of the T1 and T2 maps estimated using the proposed methods were evaluated by comparing them with reference techniques. The reconstruction performance of the proposed subspace QALAS using Zero-DeepSub was evaluated in vivo and compared with conventional QALAS at high reduction factors of up to nine-fold.

RESULTS

Phantom experiments showed that subspace QALAS had good linearity with respect to the reference methods while reducing biases and improving precision compared to conventional QALAS, especially for T2 maps. Moreover, in vivo results demonstrated that subspace QALAS had better g-factor maps and could reduce voxel blurring, noise, and artifacts compared to conventional QALAS and showed robust performance at up to nine-fold acceleration with Zero-DeepSub, which enabled whole-brain T1, T2, and PD mapping at 1 mm isotropic resolution within 2 min of scan time.

CONCLUSION

The proposed subspace QALAS along with Zero-DeepSub enabled high fidelity and rapid whole-brain multiparametric quantification and time-resolved imaging.

Accelerated MRI reconstructions via variational network and feature domain learning

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

Speeding Up and Improving Image Quality in Glioblastoma MRI Protocol by Deep Learning Image Reconstruction

A fully diagnostic MRI glioma protocol is key to monitoring therapy assessment but is time-consuming and especially challenging in critically ill and uncooperative patients. Artificial intelligence demonstrated promise in reducing scan time and improving image quality simultaneously. The purpose of this study was to investigate the diagnostic performance, the impact on acquisition acceleration, and the image quality of a deep learning optimized glioma protocol of the brain. Thirty-three patients with histologically confirmed glioblastoma underwent standardized brain tumor imaging according to the glioma consensus recommendations on a 3-Tesla MRI scanner. Conventional and deep learning-reconstructed (DLR) fluid-attenuated inversion recovery, and T2- and T1-weighted contrast-enhanced Turbo spin echo images with an improved in-plane resolution, i.e., super-resolution, were acquired. Two experienced neuroradiologists independently evaluated the image datasets for subjective image quality, diagnostic confidence, tumor conspicuity, noise levels, artifacts, and sharpness. In addition, the tumor volume was measured in the image datasets according to Response Assessment in Neuro-Oncology (RANO) 2.0, as well as compared between both imaging techniques, and various clinical-pathological parameters were determined. The average time saving of DLR sequences was 30% per MRI sequence. Simultaneously, DLR sequences showed superior overall image quality (all p < 0.001), improved tumor conspicuity and image sharpness (all p < 0.001, respectively), and less image noise (all p < 0.001), while maintaining diagnostic confidence (all p > 0.05), compared to conventional images. Regarding RANO 2.0, the volume of non-enhancing non-target lesions (p = 0.963), enhancing target lesions (p = 0.993), and enhancing non-target lesions (p = 0.951) did not differ between reconstruction types. The feasibility of the deep learning-optimized glioma protocol was demonstrated with a 30% reduction in acquisition time on average and an increased in-plane resolution. The evaluated DLR sequences improved subjective image quality and maintained diagnostic accuracy in tumor detection and tumor classification according to RANO 2.0.

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

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

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

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

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

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

New-Generation 0.55 T MRI of the Knee-Initial Clinical Experience and Comparison With 3 T MRI

OBJECTIVES

The aim of this study was to compare the detection rate of and reader confidence in 0.55 T knee magnetic resonance imaging (MRI) findings with 3 T knee MRI in patients with acute trauma and knee pain.

MATERIALS AND METHODS

In this prospective study, 0.55 T and 3 T knee MRI of 25 symptomatic patients (11 women; median age, 38 years) with suspected internal derangement of the knee was obtained in 1 setting. On the 0.55 T system, a commercially available deep learning image reconstruction algorithm was used (Deep Resolve Gain and Deep Resolve Sharp; Siemens Healthineers), which was not available on the 3 T system. Two board-certified radiologists reviewed all images independently and graded image quality parameters, noted MRI findings and their respective reporting confidence level for the presence or absence, as well as graded the bone, cartilage, meniscus, ligament, and tendon lesions. Image quality and reader confidence levels were compared ( P < 0.05 = significant), and clinical findings were correlated between 0.55 T and 3 T MRI by calculation of the intraclass correlation coefficient (ICC).

RESULTS

Image quality was rated higher at 3 T compared with 0.55 T studies (each P ≤ 0.017). Agreement between 0.55 T and 3 T MRI for the detection and grading of bone marrow edema and fractures, ligament and tendon lesions, high-grade meniscus and cartilage lesions, Baker cysts, and joint effusions was perfect for both readers. Overall identification and grading of cartilage and meniscal lesions showed good agreement between high- and low-field MRI (each ICC > 0.76), with lower agreement for low-grade cartilage (ICC = 0.77) and meniscus lesions (ICC = 0.49). There was no difference in readers' confidence levels for reporting lesions of bone, ligaments, tendons, Baker cysts, and joint effusions between 0.55 T and 3 T (each P > 0.157). Reader reporting confidence was higher for cartilage and meniscal lesions at 3 T (each P < 0.041).

CONCLUSIONS

New-generation 0.55 T knee MRI, with deep learning-aided image reconstruction, allows for reliable detection and grading of joint lesions in symptomatic patients, but it showed limited accuracy and reader confidence for low-grade cartilage and meniscal lesions in comparison with 3 T MRI.

UNCERTAINTY-GUIDED PHYSICS-DRIVEN DEEP LEARNING RECONSTRUCTION VIA CYCLIC MEASUREMENT CONSISTENCY

Physics-driven deep learning (PD-DL) techniques have recently emerged as a powerful means for improved computational imaging, including in MRI applications. These methods use the physics information by incorporating the known forward model for data fidelity, while performing regularization using neural networks. There has been substantial progress in the training of PD-DL reconstruction methods, ranging from simple supervised learning to more practical self-supervised learning and generative models that allow training without reference data. Similarly, efforts have been made to characterize the errors associated with PD-DL methods via uncertainty quantification, mostly focusing on generative models. In this work, we devise an uncertainty estimation process that primarily focuses on the data fidelity component of PD-DL by characterizing the cyclic consistency between different forward models. Subsequently, we use this uncertainty estimate to guide the training of the PD-DL method. Results show that the proposed uncertainty-guided PD-DL strategy improves reconstruction quality.

MRI of Temporomandibular Joint Disorders: A Comparative Study of 0.55 T and 1.5 T MRI

OBJECTIVES

Temporomandibular disorders (TMDs) are common and may cause persistent functional limitations and pain. Magnetic resonance imaging (MRI) at 1.5 and 3 T is commonly applied for the evaluation of the temporomandibular joint (TMJ). No evidence is available regarding the feasibility of modern low-field MRI for the assessment of TMDs. The objective of this prospective study was to evaluate the image quality (IQ) of 0.55 T MRI in direct comparison with 1.5 T MRI.

MATERIALS AND METHODS

Seventeen patients (34 TMJs) with suspected intraarticular TMDs were enrolled, and both 0.55 and 1.5 T MRI were performed on the same day. Two senior readers independently evaluated the IQ focusing on the conspicuity of disc morphology (DM), disc position (DP), and osseous joint morphology (OJM) for each joint. We analyzed the IQ and degree of artifacts using a 4-point Likert scale (LS) at both field strengths. A fully sufficient IQ was defined as an LS score of ≥3. Nonparametric Wilcoxon test for related samples was used for statistical comparison.

RESULTS

The median IQ for the DM and OJM at 0.55 T was inferior to that at 1.5 T (DM: 3 [interquartile range {IQR}, 3-4] vs 4 [IQR, 4-4]; OJM: 3 [IQR, 3-4] vs 4 [IQR 4-4]; each P < 0.001). For DP, the IQ was comparable (4 [IQR 3-4] vs 4 [IQR 4-4]; P > 0.05). A sufficient diagnostic IQ was maintained for the DM, DP, and OJM in 92% of the cases at 0.55 T and 100% at 1.5 T. Minor image artifacts (LS score of ≥3) were more prevalent at 0.55 T (29%) than at 1.5 T (12%).

CONCLUSIONS

Magnetic resonance imaging of the TMJ at 0.55 T yields a lower IQ than does MRI at 1.5 T but maintains sufficient diagnostic confidence in the majority of patients. Further improvements are needed for reliable clinical application.

Single-shot Echo Planar Time-resolved Imaging for multi-echo functional MRI and distortion-free diffusion imaging

Purpose

To develop EPTI, a multi-shot distortion-free multi-echo imaging technique, into a single-shot acquisition to achieve improved robustness to motion and physiological noise, increased temporal resolution, and high SNR efficiency for dynamic imaging applications.

Methods

A new spatiotemporal encoding was developed to achieve single-shot EPTI by enhancing spatiotemporal correlation in k-t space. The proposed single-shot encoding improves reconstruction conditioning and sampling efficiency, with additional optimization under various accelerations to achieve optimized performance. To achieve high SNR efficiency, continuous readout with minimized deadtime was employed that begins immediately after excitation and extends for an SNR-optimized length. Moreover, k-t partial Fourier and simultaneous multi-slice acquisition were integrated to further accelerate the acquisition and achieve high spatial and temporal resolution.

Results

We demonstrated that ss-EPTI achieves higher tSNR efficiency than multi-shot EPTI, and provides distortion-free imaging with densely-sampled multi-echo images at resolutions ~1.25-3 mm at 3T and 7T-with high SNR efficiency and with comparable temporal resolutions to ss-EPI. The ability of ss-EPTI to eliminate dynamic distortions common in EPI also further improves temporal stability. For fMRI, ss-EPTI also provides early-TE images (e.g., 2.9ms) to recover signal-intensity and functional-sensitivity dropout in challenging regions. The multi-echo images provide TE-dependent information about functional fluctuations, successfully distinguishing noise-components from BOLD signals and further improving tSNR. For diffusion MRI, ss-EPTI provides high-quality distortion-free diffusion images and multi-echo diffusion metrics.

Conclusion

ss-EPTI provides distortion-free imaging with high image quality, rich multi-echo information, and enhanced efficiency within comparable temporal resolution to ss-EPI, offering a robust and efficient acquisition for dynamic imaging.

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.

McSTRA: A multi-branch cascaded swin transformer for point spread function-guided robust MRI reconstruction

Deep learning MRI reconstruction methods are often based on Convolutional neural network (CNN) models; however, they are limited in capturing global correlations among image features due to the intrinsic locality of the convolution operation. Conversely, the recent vision transformer models (ViT) are capable of capturing global correlations by applying self-attention operations on image patches. Nevertheless, the existing transformer models for MRI reconstruction rarely leverage the physics of MRI. In this paper, we propose a novel physics-based transformer model titled, the Multi-branch Cascaded Swin Transformers (McSTRA) for robust MRI reconstruction. McSTRA combines several interconnected MRI physics-related concepts with the Swin transformers: it exploits global MRI features via the shifted window self-attention mechanism; it extracts MRI features belonging to different spectral components via a multi-branch setup; it iterates between intermediate de-aliasing and data consistency via a cascaded network with intermediate loss computations; furthermore, we propose a point spread function-guided positional embedding generation mechanism for the Swin transformers which exploit the spread of the aliasing artifacts for effective reconstruction. With the combination of all these components, McSTRA outperforms the state-of-the-art methods while demonstrating robustness in adversarial conditions such as higher accelerations, noisy data, different undersampling protocols, out-of-distribution data, and abnormalities in anatomy.