Beyond the Conventional Structural MRI: Clinical Application of Deep Learning Image Reconstruction and Synthetic MRI of the Brain.

ABSTRACT: Recent technological advancements have revolutionized routine brain magnetic resonance imaging (MRI) sequences, offering enhanced diagnostic capabilities in intracranial disease evaluation. This review explores 2 pivotal breakthrough areas: deep learning reconstruction (DLR) and quantitative MRI techniques beyond conventional structural imaging. DLR using deep neural networks facilitates accelerated imaging with improved signal-to-noise ratio and spatial resolution, enhancing image quality with short scan times. DLR focuses on supervised learning applied to clinical implementation and applications. Quantitative MRI techniques, exemplified by 2D multidynamic multiecho, 3D quantification using interleaved Look-Locker acquisition sequences with T2 preparation pulses, and magnetic resonance fingerprinting, enable precise calculation of brain-tissue parameters and further advance diagnostic accuracy and efficiency. Potential DLR instabilities and quantification and bias limitations will be discussed. This review underscores the synergistic potential of DLR and quantitative MRI, offering prospects for improved brain imaging beyond conventional methods.

Fast Whole-Brain MR Multi-Parametric Mapping with Scan-Specific Self-Supervised Networks.

Quantification of tissue parameters using MRI is emerging as a powerful tool in clinical diagnosis and research studies. The need for multiple long scans with different acquisition parameters prohibits quantitative MRI from reaching widespread adoption in routine clinical and research exams. Accelerated parameter mapping techniques leverage parallel imaging, signal modelling and deep learning to offer more practical quantitative MRI acquisitions. However, the achievable acceleration and the quality of maps are often limited. Joint MAPLE is a recent state-of-the-art multi-parametric and scan-specific parameter mapping technique with promising performance at high acceleration rates. It synergistically combines parallel imaging, model-based and machine learning approaches for joint mapping of T 1 , T 2 * , proton density and the field inhomogeneity. However, Joint MAPLE suffers from prohibitively long reconstruction time to estimate the maps from a multi-echo, multi-flip angle (MEMFA) dataset at high resolution in a scan-specific manner. In this work, we propose a faster version of Joint MAPLE which retains the mapping performance of the original version. Coil compression, random slice selection, parameter-specific learning rates and transfer learning are synergistically combined in the proposed framework. It speeds-up the reconstruction time up to 700 times than the original version and processes a whole-brain MEMFA dataset in 21 minutes on average, which originally requires ~260 hours for Joint MAPLE. The mapping performance of the proposed framework is ~2-fold better than the standard and the state-of-the-art evaluated reconstruction techniques on average in terms of the root mean squared error.

Time-division multiplexing (TDM) sequence removes bias in T2 estimation and relaxation-diffusion measurements.

PURPOSE: To compare the performance of multi-echo (ME) and time-division multiplexing (TDM) sequences for accelerated relaxation-diffusion MRI (rdMRI) acquisition and to examine their reliability in estimating accurate rdMRI microstructure measures. METHOD: The ME, TDM, and the reference single-echo (SE) sequences with six TEs were implemented using Pulseq with single-band (SB) and multi-band 2 (MB2) acceleration factors. On a diffusion phantom, the image intensities of the three sequences were compared, and the differences were quantified using the normalized RMS error (NRMSE). Shinnar-Le Roux (SLR) pulses were implemented for the SB-ME and SB-SE sequences to investigate the impact of slice profiles on ME sequences. For the in-vivo brain scan, besides the image intensity comparison and T2-estimates, different methods were used to assess sequence-related effects on microstructure estimation, including the relaxation diffusion imaging moment (REDIM) and the maximum-entropy relaxation diffusion distribution (MaxEnt-RDD). RESULTS: TDM performance was similar to the gold standard SE acquisition, whereas ME showed greater biases (3-4× larger NRMSEs for phantom, 2× for in-vivo). T2 values obtained from TDM closely matched SE, whereas ME sequences underestimated the T2 relaxation time. TDM provided similar diffusion and relaxation parameters as SE using REDIM, whereas SB-ME exhibited a 60% larger bias in the map and on average 3.5× larger bias in the covariance between relaxation-diffusion coefficients. CONCLUSION: Our analysis demonstrates that TDM provides a more accurate estimation of relaxation-diffusion measurements while accelerating the acquisitions by a factor of 2 to 3.

Time-division multiplexing (TDM) sequence removes bias in T2 estimation and relaxation-diffusion measurements.

Purpose: To compare the performance of multi-echo (ME) and time-division multiplexing (TDM) sequences for accelerated relaxation-diffusion MRI (rdMRI) acquisition and to examine their reliability in estimating accurate rdMRI microstructure measures. Method: The ME, TDM, and the reference single-echo (SE) sequences with six echo times (TE) were implemented using Pulseq with single-band (SB-) and multi-band 2 (MB2-) acceleration factors. On a diffusion phantom, the image intensities of the three sequences were compared, and the differences were quantified using the normalized root mean squared error (NRMSE). For the in-vivo brain scan, besides the image intensity comparison and T2-estimates, different methods were used to assess sequence-related effects on microstructure estimation, including the relaxation diffusion imaging moment (REDIM) and the maximum-entropy relaxation diffusion distribution (MaxEnt-RDD). Results: TDM performance was similar to the gold standard SE acquisition, whereas ME showed greater biases (3-4× larger NRMSEs for phantom, 2× for in-vivo). T2 values obtained from TDM closely matched SE, whereas ME sequences underestimated the T2 relaxation time. TDM provided similar diffusion and relaxation parameters as SE using REDIM, whereas SB-ME exhibited a 60% larger bias in the map and on average 3.5× larger bias in the covariance between relaxation-diffusion coefficients. Conclusion: Our analysis demonstrates that TDM provides a more accurate estimation of relaxation-diffusion measurements while accelerating the acquisitions by a factor of 2 to 3.

Chaos and COSMOS-Considerations on QSM methods with multiple and single orientations and effects from local anisotropy.

PURPOSE: Field-to-susceptibility inversion in quantitative susceptibility mapping (QSM) is ill-posed and needs numerical stabilization through either regularization or oversampling by acquiring data at three or more object orientations. Calculation Of Susceptibility through Multiple Orientations Sampling (COSMOS) is an established oversampling approach and regarded as QSM gold standard. It achieves a well-conditioned inverse problem, requiring rotations by 0°, 60° and 120° in the yz-plane. However, this is impractical in vivo, where head rotations are typically restricted to a range of ±25°. Non-ideal sampling degrades the conditioning with residual streaking artifacts whose mitigation needs further regularization. Moreover, susceptibility anisotropy in white matter is not considered in the COSMOS model, which may introduce additional bias. The current work presents a thorough investigation of these effects in primate brain. METHODS: Gradient-recalled echo (GRE) data of an entire fixed chimpanzee brain were acquired at 7 T (350 µm resolution, 10 orientations) including ideal COSMOS sampling and realistic rotations in vivo. Comparisons of the results included ideal COSMOS, in-vivo feasible acquisitions with 3-8 orientations and single-orientation iLSQR QSM. RESULTS: In-vivo feasible and optimal COSMOS yielded high-quality susceptibility maps with increased SNR resulting from averaging multiple acquisitions. COSMOS reconstructions from non-ideal rotations about a single axis required additional L2-regularization to mitigate residual streaking artifacts. CONCLUSION: In view of unconsidered anisotropy effects, added complexity of the reconstruction, and the general challenge of multi-orientation acquisitions, advantages of sub-optimal COSMOS schemes over regularized single-orientation QSM appear limited in in-vivo settings.

DIMOND: DIffusion Model OptimizatioN with Deep Learning.

Diffusion magnetic resonance imaging is an important tool for mapping tissue microstructure and structural connectivity non-invasively in the in vivo human brain. Numerous diffusion signal models are proposed to quantify microstructural properties. Nonetheless, accurate estimation of model parameters is computationally expensive and impeded by image noise. Supervised deep learning-based estimation approaches exhibit efficiency and superior performance but require additional training data and may be not generalizable. A new DIffusion Model OptimizatioN framework using physics-informed and self-supervised Deep learning entitled "DIMOND" is proposed to address this problem. DIMOND employs a neural network to map input image data to model parameters and optimizes the network by minimizing the difference between the input acquired data and synthetic data generated via the diffusion model parametrized by network outputs. DIMOND produces accurate diffusion tensor imaging results and is generalizable across subjects and datasets. Moreover, DIMOND outperforms conventional methods for fitting sophisticated microstructural models including the kurtosis and NODDI model. Importantly, DIMOND reduces NODDI model fitting time from hours to minutes, or seconds by leveraging transfer learning. In summary, the self-supervised manner, high efficacy, and efficiency of DIMOND increase the practical feasibility and adoption of microstructure and connectivity mapping in clinical and neuroscientific applications.

Reduced cross-scanner variability using vendor-agnostic sequences for single-shell diffusion MRI.

PURPOSE: To reduce the inter-scanner variability of diffusion MRI (dMRI) measures between scanners from different vendors by developing a vendor-neutral dMRI pulse sequence using the open-source vendor-agnostic Pulseq platform. METHODS: We implemented a standard EPI based dMRI sequence in Pulseq. We tested it on two clinical scanners from different vendors (Siemens Prisma and GE Premier), systematically evaluating and comparing the within- and inter-scanner variability across the vendors, using both the vendor-provided and Pulseq dMRI sequences. Assessments covered both a diffusion phantom and three human subjects, using standard error (SE) and Lin's concordance correlation to measure the repeatability and reproducibility of standard DTI metrics including fractional anisotropy (FA) and mean diffusivity (MD). RESULTS: Identical dMRI sequences were executed on both scanners using Pulseq. On the phantom, the Pulseq sequence showed more than a 2.5× reduction in SE (variability) across Siemens and GE scanners. Furthermore, Pulseq sequences exhibited markedly reduced SE in-vivo, maintaining scan-rescan repeatability while delivering lower variability in FA and MD (more than 50% reduction in cortical/subcortical regions) compared to vendor-provided sequences. CONCLUSION: The Pulseq diffusion sequence reduces the cross-scanner variability for both phantom and in-vivo data, which will benefit multi-center neuroimaging studies and improve the reproducibility of neuroimaging studies.

Clinical Evaluation of a 2-Minute Ultrafast Brain MR Protocol for Evaluation of Acute Pathology in the Emergency and Inpatient Settings.

BACKGROUND AND PURPOSE: The use of MR imaging in emergency settings has been limited by availability, long scan times, and sensitivity to motion. This study assessed the diagnostic performance of an ultrafast brain MR imaging protocol for evaluation of acute intracranial pathology in the emergency department and inpatient settings. MATERIALS AND METHODS: Sixty-six adult patients who underwent brain MR imaging in the emergency department and inpatient settings were included in the study. All patients underwent both the reference and the ultrafast brain MR protocols. Both brain MR imaging protocols consisted of T1-weighted, T2/T2*-weighted, FLAIR, and DWI sequences. The ultrafast MR images were reconstructed by using a machine-learning assisted framework. All images were reviewed by 2 blinded neuroradiologists. RESULTS: The average acquisition time was 2.1 minutes for the ultrafast brain MR protocol and 10 minutes for the reference brain MR protocol. There was 98.5% agreement on the main clinical diagnosis between the 2 protocols. In head-to-head comparison, the reference protocol was preferred in terms of image noise and geometric distortion (P < .05 for both). The ultrafast ms-EPI protocol was preferred over the reference protocol in terms of reduced motion artifacts (P < .01). Overall diagnostic quality was not significantly different between the 2 protocols (P > .05). CONCLUSIONS: The ultrafast brain MR imaging protocol provides high accuracy for evaluating acute pathology while only requiring a fraction of the scan time. Although there was greater image noise and geometric distortion on the ultrafast brain MR protocol images, there was significant reduction in motion artifacts with similar overall diagnostic quality between the 2 protocols.

Cross-vendor multiparametric mapping of the human brain using 3D-QALAS: A multicenter and multivendor study.

PURPOSE: To evaluate a vendor-agnostic multiparametric mapping scheme based on 3D quantification using an interleaved Look-Locker acquisition sequence with a T2 preparation pulse (3D-QALAS) for whole-brain T1, T2, and proton density (PD) mapping. METHODS: This prospective, multi-institutional study was conducted between September 2021 and February 2022 using five different 3T systems from four prominent MRI vendors. The accuracy of this technique was evaluated using a standardized MRI system phantom. Intra-scanner repeatability and inter-vendor reproducibility of T1, T2, and PD values were evaluated in 10 healthy volunteers (6 men; mean age ± SD, 28.0 ± 5.6 y) who underwent scan-rescan sessions on each scanner (total scans = 100). To evaluate the feasibility of 3D-QALAS, nine patients with multiple sclerosis (nine women; mean age ± SD, 48.2 ± 11.5 y) underwent imaging examination on two 3T MRI systems from different manufacturers. RESULTS: Quantitative maps obtained with 3D-QALAS showed high linearity (R2 = 0.998 and 0.998 for T1 and T2, respectively) with respect to reference measurements. The mean intra-scanner coefficients of variation for each scanner and structure ranged from 0.4% to 2.6%. The mean structure-wise test-retest repeatabilities were 1.6%, 1.1%, and 0.7% for T1, T2, and PD, respectively. Overall, high inter-vendor reproducibility was observed for all parameter maps and all structure measurements, including white matter lesions in patients with multiple sclerosis. CONCLUSION: The vendor-agnostic multiparametric mapping technique 3D-QALAS provided reproducible measurements of T1, T2, and PD for human tissues within a typical physiological range using 3T scanners from four different MRI manufacturers.

Recommended implementation of quantitative susceptibility mapping for clinical research in the brain: A consensus of the ISMRM electro-magnetic tissue properties study group.

This article provides recommendations for implementing QSM for clinical brain research. It is a consensus of the International Society of Magnetic Resonance in Medicine, Electro-Magnetic Tissue Properties Study Group. While QSM technical development continues to advance rapidly, the current QSM methods have been demonstrated to be repeatable and reproducible for generating quantitative tissue magnetic susceptibility maps in the brain. However, the many QSM approaches available have generated a need in the neuroimaging community for guidelines on implementation. This article outlines considerations and implementation recommendations for QSM data acquisition, processing, analysis, and publication. We recommend that data be acquired using a monopolar 3D multi-echo gradient echo (GRE) sequence and that phase images be saved and exported in Digital Imaging and Communications in Medicine (DICOM) format and unwrapped using an exact unwrapping approach. Multi-echo images should be combined before background field removal, and a brain mask created using a brain extraction tool with the incorporation of phase-quality-based masking. Background fields within the brain mask should be removed using a technique based on SHARP or PDF, and the optimization approach to dipole inversion should be employed with a sparsity-based regularization. Susceptibility values should be measured relative to a specified reference, including the common reference region of the whole brain as a region of interest in the analysis. The minimum acquisition and processing details required when reporting QSM results are also provided. These recommendations should facilitate clinical QSM research and promote harmonized data acquisition, analysis, and reporting.

Joint MAPLE: Accelerated joint T1 and T 2 * $$ {{\mathrm{T}}_2}

PURPOSE: Quantitative MRI finds important applications in clinical and research studies. However, it is encoding intensive and may suffer from prohibitively long scan times. Accelerated MR parameter mapping techniques have been developed to help address these challenges. Here, an accelerated joint T1, T 2 * $$ {{\mathrm{T}}_2}^{\ast } $$ , frequency and proton density mapping technique with scan-specific self-supervised network reconstruction is proposed to synergistically combine parallel imaging, model-based, and deep learning approaches to speed up parameter mapping. METHODS: Proposed framework, Joint MAPLE, includes parallel imaging, signal modeling, and data consistency blocks which are optimized jointly in a combined loss function. A scan-specific self-supervised reconstruction is embedded into the framework, which takes advantage of multi-contrast data from a multi-echo, multi-flip angle, gradient echo acquisition. RESULTS: In comparison with parallel reconstruction techniques powered by low-rank methods, emerging scan specific networks, and model-based T 2 * $$ {{\mathrm{T}}_2}^{\ast } $$ estimation approaches, the proposed framework reduces the reconstruction error in parameter maps by approximately two-fold on average at acceleration rates as high as R = 16 with uniform sampling. It can outperform evaluated parallel reconstruction techniques up to four-fold on average in the presence of challenging sub-sampling masks. It is observed that Joint MAPLE performs well at extreme acceleration rates of R = 25 and R = 36 with error values less than 20%. CONCLUSION: Joint MAPLE enables higher fidelity parameter estimation at high acceleration rates by synergistically combining parallel imaging and model-based parameter mapping and exploiting multi-echo, multi-flip angle datasets. Utilizing a scan-specific self-supervised reconstruction obviates the need for large data sets for training while improving the parameter estimation ability.

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.

Eddy current-induced artifact correction in high b-value ex vivo human brain diffusion MRI with dynamic field monitoring.

PURPOSE: To investigate whether spatiotemporal magnetic field monitoring can correct pronounced eddy current-induced artifacts incurred by strong diffusion-sensitizing gradients up to 300 mT/m used in high b-value diffusion-weighted (DW) EPI. METHODS: A dynamic field camera equipped with 16 1 H NMR field probes was first used to characterize field perturbations caused by residual eddy currents from diffusion gradients waveforms in a 3D multi-shot EPI sequence on a 3T Connectom scanner for different gradient strengths (up to 300 mT/m), diffusion directions, and shots. The efficacy of dynamic field monitoring-based image reconstruction was demonstrated on high-gradient strength, submillimeter resolution whole-brain ex vivo diffusion MRI. A 3D multi-shot image reconstruction framework was developed that incorporated the nonlinear phase evolution measured with the dynamic field camera. RESULTS: Phase perturbations in the readout induced by residual eddy currents from strong diffusion gradients are highly nonlinear in space and time, vary among diffusion directions, and interfere significantly with the image encoding gradients, changing the k-space trajectory. During the readout, phase modulations between odd and even EPI echoes become non-static and diffusion encoding direction-dependent. Superior reduction of ghosting and geometric distortion was achieved with dynamic field monitoring compared to ghosting reduction approaches such as navigator- and structured low-rank-based methods or MUSE followed by image-based distortion correction with the FSL tool "eddy." CONCLUSION: Strong eddy current artifacts characteristic of high-gradient strength DW-EPI can be well corrected with dynamic field monitoring-based image reconstruction.