High angular resolution susceptibility imaging and estimation of fiber orientation distribution functions in primate brain.

Uncovering brain-tissue microstructure including axonal characteristics is a major neuroimaging research focus. Within this scope, anisotropic properties of magnetic susceptibility in white matter have been successfully employed to estimate primary axonal trajectories using mono-tensorial models. However, anisotropic susceptibility has not yet been considered for modeling more complex fiber structures within a voxel, such as intersecting bundles, or an estimation of orientation distribution functions (ODFs). This information is routinely obtained by high angular resolution diffusion imaging (HARDI) techniques. In applications to fixed tissue, however, diffusion-weighted imaging suffers from an inherently low signal-to-noise ratio and limited spatial resolution, leading to high demands on the performance of the gradient system in order to mitigate these limitations. In the current work, high angular resolution susceptibility imaging (HARSI) is proposed as a novel, phase-based methodology to estimate ODFs. A multiple gradient-echo dataset was acquired in an entire fixed chimpanzee brain at 61 orientations by reorienting the specimen in the magnetic field. The constant solid angle method was adapted for estimating phase-based ODFs. HARDI data were also acquired for comparison. HARSI yielded information on whole-brain fiber architecture, including identification of peaks of multiple bundles that resembled features of the HARDI results. Distinct differences between both methods suggest that susceptibility properties may offer complementary microstructural information. These proof-of-concept results indicate a potential to study the axonal organization in post-mortem primate and human brain at high resolution.

Deep learning-based Accelerated and Noise-Suppressed Estimation (DANSE) of quantitative Gradient-Recalled Echo (qGRE) magnetic resonance imaging metrics associated with human brain neuronal structure and hemodynamic properties.

The purpose of the current study was to introduce a Deep learning-based Accelerated and Noise-Suppressed Estimation (DANSE) method for reconstructing quantitative maps of biological tissue cellular-specific, R2t*, and hemodynamic-specific, R2', metrics of quantitative gradient-recalled echo (qGRE) MRI. The DANSE method adapts a supervised learning paradigm to train a convolutional neural network for robust estimation of R2t* and R2' maps with significantly reduced sensitivity to noise and the adverse effects of macroscopic (B0 ) magnetic field inhomogeneities directly from the gradient-recalled echo (GRE) magnitude images. The R2t* and R2' maps for training were generated by means of a voxel-by-voxel fitting of a previously developed biophysical quantitative qGRE model accounting for tissue, hemodynamic, and B0 -inhomogeneities contributions to multigradient-echo GRE signal using a nonlinear least squares (NLLS) algorithm. We show that the DANSE model efficiently estimates the aforementioned qGRE maps and preserves all the features of the NLLS approach with significant improvements including noise suppression and computation speed (from many hours to seconds). The noise-suppression feature of DANSE is especially prominent for data with low signal-to-noise ratio (SNR ~ 50-100), where DANSE-generated R2t* and R2' maps had up to three times smaller errors than that of the NLLS method. The DANSE method enables fast reconstruction of qGRE maps with significantly reduced sensitivity to noise and magnetic field inhomogeneities. The DANSE method does not require any information about field inhomogeneities during application. It exploits spatial and gradient echo time-dependent patterns in the GRE data and previously gained knowledge from the biophysical model, thus producing high quality qGRE maps, even in environments with high noise levels. These features along with fast computational speed can lead to broad qGRE clinical and research applications.

Learning-based motion artifact removal networks for quantitative R2∗ mapping.

PURPOSE: To introduce two novel learning-based motion artifact removal networks (LEARN) for the estimation of quantitative motion- and B0 -inhomogeneity-corrected R2∗ maps from motion-corrupted multi-Gradient-Recalled Echo (mGRE) MRI data. METHODS: We train two convolutional neural networks (CNNs) to correct motion artifacts for high-quality estimation of quantitative B0 -inhomogeneity-corrected R2∗ maps from mGRE sequences. The first CNN, LEARN-IMG, performs motion correction on complex mGRE images, to enable the subsequent computation of high-quality motion-free quantitative R2∗ (and any other mGRE-enabled) maps using the standard voxel-wise analysis or machine learning-based analysis. The second CNN, LEARN-BIO, is trained to directly generate motion- and B0 -inhomogeneity-corrected quantitative R2∗ maps from motion-corrupted magnitude-only mGRE images by taking advantage of the biophysical model describing the mGRE signal decay. RESULTS: We show that both CNNs trained on synthetic MR images are capable of suppressing motion artifacts while preserving details in the predicted quantitative R2∗ maps. Significant reduction of motion artifacts on experimental in vivo motion-corrupted data has also been achieved by using our trained models. CONCLUSION: Both LEARN-IMG and LEARN-BIO can enable the computation of high-quality motion- and B0 -inhomogeneity-corrected R2∗ maps. LEARN-IMG performs motion correction on mGRE images and relies on the subsequent analysis for the estimation of R2∗ maps, while LEARN-BIO directly performs motion- and B0 -inhomogeneity-corrected R2∗ estimation. Both LEARN-IMG and LEARN-BIO jointly process all the available gradient echoes, which enables them to exploit spatial patterns available in the data. The high computational speed of LEARN-BIO is an advantage that can lead to a broader clinical application.

Tissue damage detected by quantitative gradient echo MRI correlates with clinical progression in non-relapsing progressive MS.

BACKGROUND: Imaging biomarkers of progressive multiple sclerosis (MS) are needed. Quantitative gradient recalled echo (qGRE) magnetic resonance imaging (MRI) evaluates microstructural tissue damage in MS. OBJECTIVE: To evaluate qGRE-derived R2t* as an imaging biomarker of MS progression compared with atrophy and lesion burden. METHODS: Twenty-three non-relapsing progressive MS (PMS), 22 relapsing-remitting MS (RRMS), and 18 healthy control participants underwent standard MS physical and cognitive neurological assessments and imaging with qGRE, FLAIR, and MPRAGE at 3T. PMS subjects were tested clinically and imaged every 9 months over 45 months. Imaging measures included lesion burden, atrophy, and R2t* in cortical gray matter (GM), deep GM, and normal-appearing white matter (NAWM). Longitudinal analysis of clinical performance and imaging biomarkers in PMS subjects was conducted via linear models with subject as repeated, within-subject factor. Relationship between imaging biomarkers and clinical scores was assessed by Spearman rank correlation. RESULTS: R2t* reductions correlated with neurological impairment cross-sectionally and longitudinally. PMS patients with clinically defined disease progression (N = 13) showed faster decrease of R2t* in NAWM and deep GM compared with the clinically stable PMS group (N = 10). Importantly, tissue damage measured by R2t* outperformed lesion burden and atrophy as a biomarker of progression during the study period. CONCLUSION: qGRE-derived R2t* is a potential imaging biomarker of MS progression.

An analysis of the correlation between diabetic retinopathy and preretinal oxygen tension using three-dimensional spoiled gradient-recalled echo sequence imaging.

BACKGROUND: The aims of this study were to evaluate the levels of preretinal oxygen tension in patients with diabetes who did not have hypertension by using three-dimensional spoiled gradient-recalled (3D-SPGR) echo sequence imaging and to explore the correlation between diabetic retinopathy (DR) and changes in preretinal oxygen tension. METHOD: This study involved 15 patients with type 2 diabetes without hypertension, who were divided into a diabetic retinopathy (DR) group (n = 10 eyes) and a diabetic non-retinopathy (NDR) group (n = 20 eyes), according to the results of a fundus photography test. Another healthy control group (n = 14 eyes) also participated in the study. The preretinal vitreous optic disc area, nasal side, and temporal side signal intensity of the eyes was assessed before and after oxygen inhalation with the use of 3D-SPGR echo magnetic resonance imaging (MRI). The signal acquisition time was 10, 20, 30, 40, and 50 min after oxygen inhalation. RESULTS: The results showed that, in the DR and NDR groups, the preretinal vitreous oxygen tension increased rapidly at 10 min after oxygen inhalation and peaked at 30-40 min, and the increased slope of the DR group was higher than that of the NDR group. The oxygen tension of the preretinal vitreous gradually increased after oxygen inhalation, and the difference between the DR and NDR groups and the control group was statistically significant (P < 0.05). The preretinal vitreous oxygen tension was higher in the optic disc, temporal side, and nasal side in the NDR group than in the control group, and the difference was statistically significant (P < 0.05). The maximum slope ratios of the optic disc and the temporal side of the DR group were greater than those of the control group, and the difference was statistically significant (P < 0.05). CONCLUSION: Three-dimensional-SPGR echo MRI sequencing technology is useful for detecting preretinal oxygen tension levels in patients with diabetes. It can be used as one of the functional and imaging observation indicators for the early diagnosis of DR.

Decompose quantitative susceptibility mapping (QSM) to sub-voxel diamagnetic and paramagnetic components based on gradient-echo MRI data.

PURPOSE: A method named DECOMPOSE-QSM is developed to decompose bulk susceptibility measured with QSM into sub-voxel paramagnetic and diamagnetic components based on a three-pool complex signal model. METHODS: Multi-echo gradient echo signal is modeled as a summation of three weighted exponentials corresponding to three types of susceptibility sources: reference susceptibility, diamagnetic and paramagnetic susceptibility relative to the reference. Paramagnetic component susceptibility (PCS) and diamagnetic component susceptibility (DCS) maps are constructed to represent the sub-voxel compartments by solving for linear and nonlinear parameters in the model. RESULTS: Numerical forward simulation and phantom validation confirmed the ability of DECOMPOSE-QSM to separate the mixture of paramagnetic and diamagnetic components. The PCS obtained from temperature-variant brainstem imaging follows the Curie's Law, which further validated the model and the solver. Initial in vivo investigation of human brain images showed the ability to extract sub-voxel PCS and DCS sources that produce visually enhanced contrast between brain structures comparing to threshold QSM.

In vivo evaluation of heme and non-heme iron content and neuronal density in human basal ganglia.

Non-heme iron is an important element supporting the structure and functioning of biological tissues. Imbalance in non-heme iron can lead to different neurological disorders. Several MRI approaches have been developed for iron quantification relying either on the relaxation properties of MRI signal or measuring tissue magnetic susceptibility. Specific quantification of the non-heme iron can, however, be constrained by the presence of the heme iron in the deoxygenated blood and contribution of cellular composition. The goal of this paper is to introduce theoretical background and experimental MRI method allowing disentangling contributions of heme and non-heme irons simultaneously with evaluation of tissue neuronal density in the iron-rich basal ganglia. Our approach is based on the quantitative Gradient Recalled Echo (qGRE) MRI technique that allows separation of the total R2* metric characterizing decay of GRE signal into tissue-specific (R2t*) and the baseline blood oxygen level-dependent (BOLD) contributions. A combination with the QSM data (also available from the qGRE signal phase) allowed further separation of the tissue-specific R2t* metric in a cell-specific and non-heme-iron-specific contributions. It is shown that the non-heme iron contribution to R2t* relaxation can be described with the previously developed Gaussian Phase Approximation (GPA) approach. qGRE data were obtained from 22 healthy control participants (ages 26-63 years). Results suggest that the ferritin complexes are aggregated in clusters with an average radius about 100nm comprising approximately 2600 individual ferritin units. It is also demonstrated that the concentrations of heme and non-heme iron tend to increase with age. The strongest age effect was seen in the pallidum region, where the highest age-related non-heme iron accumulation was observed.

Robust water-fat separation based on deep learning model exploring multi-echo nature of mGRE.

PURPOSE: To design a new deep learning network for fast and accurate water-fat separation by exploring the correlations between multiple echoes in multi-echo gradient-recalled echo (mGRE) sequence and evaluate the generalization capabilities of the network for different echo times, field inhomogeneities, and imaging regions. METHODS: A new multi-echo bidirectional convolutional residual network (MEBCRN) was designed to separate water and fat images in a fast and accurate manner for the mGRE data. This new MEBCRN network contains 2 main modules, the first 1 is the feature extraction module, which learns the correlations between consecutive echoes, and the other one is the water-fat separation module that processes the feature information extracted from the feature extraction module. The multi-layer feature fusion (MLFF) mechanism and residual structure were adopted in the water-fat separation module to increase separation accuracy and robustness. Moreover, we trained the network using in vivo abdomen images and tested it on the abdomen, knee, and wrist images. RESULTS: The results showed that the proposed network could separate water and fat images accurately. The comparison of the proposed network and other deep learning methods shows the advantage in both quantitative metrics and robustness for different TEs, field inhomogeneities, and images acquired for various imaging regions. CONCLUSION: The proposed network could learn the correlations between consecutive echoes and separate water and fat images effectively. The deep learning method has certain generalization capabilities for TEs and field inhomogeneity. Although the network was trained only in vivo abdomen images, it could be applied for different imaging regions.

Magnetic resonance elastography for estimating in vivo stiffness of the abdominal aorta using cardiac-gated spin-echo echo-planar imaging: a feasibility study.

INTRODUCTION: Magnetic resonance elastography (MRE)-derived aortic stiffness is a potential biomarker for multiple cardiovascular diseases. Currently, gradient-recalled echo (GRE) MRE is a widely accepted technique to estimate aortic stiffness. However, multi-slice GRE MRE requires multiple breath-holds (BHs), which can be challenging for patients who cannot consistently hold their breath. The aim of this study was to investigate the feasibility of a multi-slice spin-echo echo-planar imaging (SE-EPI) MRE sequence for quantifying in vivo aortic stiffness using a free-breathing (FB) protocol and a single-BH protocol. METHOD: On Scanner 1, 25 healthy subjects participated in the validation of FB SE-EPI against FB GRE. On Scanner 2, another 15 healthy subjects were recruited to compare FB SE-EPI with single-BH SE-EPI. Among all volunteers, five participants were studied on both scanners to investigate the inter-scanner reproducibility of FB SE-EPI aortic MRE. Bland-Altman analysis, Lin's concordance correlation coefficient (LCCC) and coefficient of variation (COV) were evaluated. The phase-difference signal-to-noise ratios (PD SNR) were compared. RESULTS: Aortic MRE using FB SE-EPI and FB GRE yielded similar stiffnesses (paired t-test, P = 0.19), with LCCC = 0.97. The FB SE-EPI measurements were reproducible (intra-scanner LCCC = 0.96) and highly repeatable (LCCC = 0.99). The FB SE-EPI MRE was also reproducible across different scanners (inter-scanner LCCC = 0.96). Single-BH SE-EPI scans yielded similar stiffness to FB SE-EPI scans (LCCC = 0.99) and demonstrated a low COV of 2.67% across five repeated measurements. CONCLUSION: Multi-slice SE-EPI aortic MRE using an FB protocol or a single-BH protocol is reproducible and repeatable with advantage over multi-slice FB GRE in reducing acquisition time. Additionally, FB SE-EPI MRE provides a potential alternative to BH scans for patients who have challenges in holding their breath.

Simultaneous time of flight-MRA and T2* imaging for cerebrovascular MRI.

PURPOSE: 3D multi-echo gradient-recalled echo (ME-GRE) can simultaneously generate time-of-flight magnetic resonance angiography (pTOF) in addition to T2*-based susceptibility-weighted images (SWI). We assessed the clinical performance of pTOF generated from a 3D ME-GRE acquisition compared with conventional TOF-MRA (cTOF). METHODS: Eighty consecutive children were retrospectively identified who obtained 3D ME-GRE alongside cTOF. Two blinded readers independently assessed pTOF derived from 3D ME-GRE and compared them with cTOF. A 5-point Likert scale was used to rank lesion conspicuity and to assess for diagnostic confidence. RESULTS: Across 80 pediatric neurovascular pathologies, a similar number of lesions were reported on pTOF and cTOF (43-40%, respectively, p > 0.05). Rating of lesion conspicuity was higher with cTOF (4.5 ± 1.0) as compared with pTOF (4.0 ± 0.7), but this was not significantly different (p = 0.06). Diagnostic confidence was rated higher with cTOF (4.8 ± 0.5) than that of pTOF (3.7 ± 0.6; p < 0.001). Overall, the inter-rater agreement between two readers for lesion count on pTOF was classified as almost perfect (κ = 0.98, 96% CI 0.8-1.0). CONCLUSIONS: In this study, TOF-MRA simultaneously generated in addition to SWI from 3D MR-GRE can serve as a diagnostic adjunct, particularly for proximal vessel disease and when conventional TOF-MRA images are absent.

The brains of aged mice are characterized by altered tissue diffusion properties and cerebral microbleeds.

BACKGROUND: Brain aging is a major risk factor in the progression of cognitive diseases including Alzheimer's disease (AD) and vascular dementia. We investigated a mouse model of brain aging up to 24 months old (mo). METHODS: A high field (11.7T) MRI protocol was developed to characterize specific features of brain aging including the presence of cerebral microbleeds (CMBs), morphology of grey and white matter, and tissue diffusion properties. Mice were selected from age categories of either young (3 mo), middle-aged (18 mo), or old (24 mo) and fed normal chow over the duration of the study. Mice were imaged in vivo with multimodal MRI, including conventional T2-weighted (T2W) and T2*-weighted (T2*W) imaging, followed by ex vivo diffusion-weighted imaging (DWI) and T2*W MR-microscopy to enhance the detection of microstructural features. RESULTS: Structural changes observed in the mouse brain with aging included reduced cortical grey matter volume and enlargement of the brain ventricles. A remarkable age-related change in the brains was the development of CMBs found starting at 18 mo and increasing in total volume at 24 mo, primarily in the thalamus. CMBs presence was confirmed with high resolution ex vivo MRI and histology. DWI detected further brain tissue changes in the aged mice including reduced fractional anisotropy, increased radial diffusion, increased mean diffusion, and changes in the white matter fibers visualized by color-coded tractography, including around a large cortical CMB. CONCLUSIONS: The mouse is a valuable model of age-related vascular contributions to cognitive impairment and dementia (VCID). In composite, these methods and results reveal brain aging in older mice as a multifactorial process including CMBs and tissue diffusion alterations that can be well characterized by high field MRI.

Deep learning using a biophysical model for robust and accelerated reconstruction of quantitative, artifact-free and denoised R2* images.

PURPOSE: To introduce a novel deep learning method for Robust and Accelerated Reconstruction (RoAR) of quantitative and B0-inhomogeneity-corrected R2* maps from multi-gradient recalled echo (mGRE) MRI data. METHODS: RoAR trains a convolutional neural network (CNN) to generate quantitative R2∗ maps free from field inhomogeneity artifacts by adopting a self-supervised learning strategy given (a) mGRE magnitude images, (b) the biophysical model describing mGRE signal decay, and (c) preliminary-evaluated F-function accounting for contribution of macroscopic B0 field inhomogeneities. Importantly, no ground-truth R2* images are required and F-function is only needed during RoAR training but not application. RESULTS: We show that RoAR preserves all features of R2* maps while offering significant improvements over existing methods in computation speed (seconds vs. hours) and reduced sensitivity to noise. Even for data with SNR = 5 RoAR produced R2* maps with accuracy of 22% while voxel-wise analysis accuracy was 47%. For SNR = 10 the RoAR accuracy increased to 17% vs. 24% for direct voxel-wise analysis. CONCLUSIONS: RoAR is trained to recognize the macroscopic magnetic field inhomogeneities directly from the input magnitude-only mGRE data and eliminate their effect on R2∗ measurements. RoAR training is based on the biophysical model and does not require ground-truth R2* maps. Since RoAR utilizes signal information not just from individual voxels but also accounts for spatial patterns of the signals in the images, it reduces the sensitivity of R2* maps to the noise in the data. These features plus high computational speed provide significant benefits for the potential usage of RoAR in clinical settings.

Comparison of technical failure of MR elastography for measuring liver stiffness between gradient-recalled echo and spin-echo echo-planar imaging: A systematic review and meta-analysis.

BACKGROUND: Magnetic resonance elastography (MRE) using a gradient-recalled echo (GRE) or a recently available spin-echo echo-planar imaging (SE-EPI) sequence is a promising noninvasive method for measuring liver stiffness. However, it sometimes fails to measure stiffness values, thereby resulting in technical failures. PURPOSE: To assess and compare technical failures of MRE for measuring liver stiffness between GRE and SE-EPI sequences. STUDY TYPE: Systematic review and meta-analysis. POPULATION: Eight studies with both GRE and SE-EPI, 22 studies with only GRE, one study with only SE-EPI. FIELD STRENGTH/SEQUENCE: Either 1.5 or 3T MRE using GRE and/or SE-EPI. ASSESSMENT: Through an Ovid-MEDLINE and EMBASE database search, original articles investigating the proportion of MRE technical failures in the measurement of liver stiffness published up until October 2018 were screened and selected. STATISTICAL ANALYSIS: The pooled proportions of technical failures under GRE and SE-EPI were calculated using random-effects meta-analysis of single proportions and inverse variance for calculating weights. Subgroup analyses were performed to explore the covariates affecting heterogeneity. Head-to-head comparisons of technical failure between the sequences were conducted with eight MRE studies using both GRE and SE-EPI. RESULTS: The pooled proportion of technical failure under GRE MRE was 5.8% (95% confidence interval [CI], 4.6-7.4%), and a subgroup analysis showed higher technical failure rates at 3T than at 1.5T. The pooled proportion of technical failure under SE-EPI MRE was 2.0% (95% CI, 1.3-3.4%), without significant differences (P = 0.38-0.89) being observed in the subgroup analyses. In the eight studies comparing the two sequences, failure was more frequently observed with GRE than with SE-EPI (9.4% vs. 1.9%; P < 0.01). DATA CONCLUSION: MRE conducted with SE-EPI sequences showed a lower technical failure rate than GRE sequences. With GRE sequences, a magnetic field of 3T was associated with higher technical failure rates than was 1.5T. LEVEL OF EVIDENCE: 1 Technical Efficacy Stage: 3 J. Magn. Reson. Imaging 2020;51:1086-1102.

PET-guided repeat transsphenoidal surgery for previously deemed unresectable lateral disease in acromegaly.

OBJECTIVE: The object of this study was to determine if revision transsphenoidal surgery (TSS), guided by 11C-methionine PET/CT coregistered with volumetric MRI (Met-PET/MRCR), can lead to remission in patients with persistent acromegaly due to a postoperative lateral disease remnant. METHODS: The authors identified 9 patients with persistent acromegaly following primary intervention (TSS ± medical therapy ± radiotherapy) in whom further surgery had initially been discounted because of equivocal MRI findings with suspected lateral sellar and/or parasellar disease (cases with clear Knosp grade 4 disease were excluded). All patients underwent Met-PET/MRCR. Scan findings were used by the pituitary multidisciplinary team to inform decision-making regarding repeat surgery. Revision TSS was performed with wide lateral exploration as guided by the PET findings. Endocrine reassessment was performed at 6-10 weeks after surgery, with longitudinal follow-up thereafter. RESULTS: Met-PET/MRCR revealed focal tracer uptake in the lateral sellar and/or parasellar region(s) in all 9 patients, which correlated with sites of suspected residual tumor on volumetric MRI. At surgery, tumor was identified and resected in 5 patients, although histological analysis confirmed somatotroph tumor in only 4 cases. In the other 4 patients, no definite tumor was seen, but equivocal tissue was removed. Despite the uncertainty at surgery, all patients showed immediate significant improvements in clinical and biochemical parameters. In the 8 patients for whom long-term follow-up data were available, insulin-like growth factor 1 (IGF-1) was ≤ 1.2 times the upper limit of normal (ULN) in all subjects and ≤ 1 times the ULN in 6 subjects, and these findings have been maintained for up to 28 months (median 8 months, mean 13 months) with no requirement for adjunctive medical therapy or radiotherapy. No patient suffered any additional pituitary deficit or other complication of surgery. CONCLUSIONS: This study provides proof of concept that Met-PET/MRCR can be helpful in the evaluation of residual lateral sellar/parasellar disease in persistent acromegaly and facilitate targeted revision TSS in a subgroup of patients.

Robust water-fat separation for multi-echo gradient-recalled echo sequence using convolutional neural network.

PURPOSE: To accurately separate water and fat signals for bipolar multi-echo gradient-recalled echo sequence using a convolutional neural network (CNN). METHODS: A CNN architecture was designed and trained using the relationship between multi-echo images from the bipolar multi-echo gradient-recalled echo sequence and artifact-free water-fat-separated images. The artifact-free water-fat-separated images for training the CNN were obtained from multiple signals with different TEs by using iterative decomposition of water and fat with echo asymmetry and the least-squares estimation method, in which multiple signals at different TEs were acquired using a single-echo gradient-recalled echo sequence. We also proposed a data augmentation method using a synthetic field inhomogeneity to generate multi-echo signals, including various bipolar multi-echo gradient-recalled echo artifacts so that the CNN could prevent overfitting and increase the separation accuracy. We trained the CNN using in vivo knee images and tested it using in vivo knee, head, and ankle images. RESULTS: In vivo imaging results showed that the proposed CNN could separate water-fat images accurately. Although the proposed CNN was trained using only in vivo knee images, the proposed CNN could also separate water-fat images of different imaging regions. The proposed data augmentation method could prevent overfitting even with a limited number of training data sets and make the method robust to magnetic field inhomogeneities. CONCLUSION: The proposed CNN could obtain water-fat-separated images from the multi-echo images acquired from the bipolar multi-echo gradient-recalled echo sequence, which included artifacts from the bipolar gradients.

Biophysically motivated efficient estimation of the spatially isotropic R2* component from a single gradient-recalled echo measurement.

PURPOSE: To propose and validate an efficient method, based on a biophysically motivated signal model, for removing the orientation-dependent part of R2* using a single gradient-recalled echo (GRE) measurement. METHODS: The proposed method utilized a temporal second-order approximation of the hollow-cylinder-fiber model, in which the parameter describing the linear signal decay corresponded to the orientation-independent part of R2* . The estimated parameters were compared to the classical, mono-exponential decay model for R2* in a sample of an ex vivo human optic chiasm (OC). The OC was measured at 16 distinct orientations relative to the external magnetic field using GRE at 7T. To show that the proposed signal model can remove the orientation dependence of R2* , it was compared to the established phenomenological method for separating R2* into orientation-dependent and -independent parts. RESULTS: Using the phenomenological method on the classical signal model, the well-known separation of R2* into orientation-dependent and -independent parts was verified. For the proposed model, no significant orientation dependence in the linear signal decay parameter was observed. CONCLUSIONS: Since the proposed second-order model features orientation-dependent and -independent components at distinct temporal orders, it can be used to remove the orientation dependence of R2* using only a single GRE measurement.

Visualization of the normal appendix in children: feasibility of a single contrast-enhanced radial gradient recalled echo MRI sequence.

BACKGROUND: Magnetic resonance imaging (MRI) assessment for appendicitis is limited by exam time and patient cooperation. The radially sampled 3-dimensional (3-D) T1-weighted, gradient recalled echo sequence (radial GRE) is a free-breathing, motion robust sequence that may be useful in evaluating appendicitis in children. OBJECTIVE: To compare the rate of detection of the normal appendix with contrast-enhanced radial GRE versus contrast-enhanced 3-D GRE and a multi-sequence study including contrast-enhanced radial GRE. MATERIALS AND METHODS: This was a retrospective study of patients ages 7-18 years undergoing abdominal-pelvic contrast-enhanced MRI between Jan. 1, 2012, and April 1, 2016. Visualization of the appendix was assessed by consensus between two pediatric radiologists. The rate of detection of the appendix for each sequence and combination of sequences was compared using a McNemar test. RESULTS: The rate of detection of the normal appendix on contrast-enhanced radial GRE was significantly higher than on contrast-enhanced 3-D GRE (76% vs. 57.3%, P=0.003). The rate of detection of the normal appendix with multi-sequence MRI including contrast-enhanced radial GRE was significantly higher than on contrast-enhanced 3-D GRE (81.3% vs. 57%, P<0.001). There was no significant difference between the rate of detection of the normal appendix on contrast-enhanced radial GRE alone and multi-sequence MRI including contrast-enhanced radial GRE (76% vs. 81.3%, P=0.267). CONCLUSION: Contrast-enhanced radial GRE allows superior detection of the normal appendix compared to contrast-enhanced 3-D GRE. The rate of detection of the normal appendix on contrast-enhanced radial GRE alone is nearly as good as when the contrast-enhanced radial GRE is interpreted with additional sequences.

Comparison of gradient-recalled echo and spin-echo echo-planar imaging MR elastography in staging liver fibrosis: a meta-analysis.

OBJECTIVES: To compare the diagnostic performance of gradient-recalled echo-based magnetic resonance elastography (GRE-MRE) and spin-echo echo-planar imaging-based MRE (SE-EPI-MRE) in liver fibrosis staging. METHODS: A systematic literature search was performed to identify studies involving the performance of MRE for the diagnosis of liver fibrosis. Pooled sensitivity, specificity, positive and negative likelihood ratios, the diagnostic odds ratio, and a summary receiver operating characteristic (ROC) curve were estimated by using a bivariate random effects model. Subgroup analyses were performed between different study characteristics. RESULTS: Twenty-six studies with a total of 3,200 patients were included in the meta-analysis. Pooled sensitivity and specificity of GRE-MRE and SE-EPI-MRE did not differ significantly. The area under the summary ROC curve for stage diagnosis of any (F ≥ 1), significant (F ≥ 2), advanced (F ≥ 3), and cirrhosis (F = 4) on GRE-MRE and SE-EPI-MRE were 0.93 versus 0.94, 0.95 versus 0.94, 0.94 versus 0.95, and 0.92 versus 0.93, respectively. Substantial heterogeneity was detected for both sequences. CONCLUSION: Both GRE and SE-EPI-MRE show high sensitivity and specificity for detection of each stage of liver fibrosis, without significant differences. Magnetic resonance elastography (MRE) may be useful for noninvasive evaluation of liver fibrosis in chronic liver disease. KEY POINTS: • Pooled sensitivity and specificity of GRE-MRE and SE-EPI-MRE did not differ significantly. • GRE-MRE and SE-EPI-MRE were highly accurate for detecting all stages of fibrosis. • Due to better agreement and repeatability, GRE-MRE should be used first. • In case of failure on GRE-MRE, SE-EPI-MRE should be used.

Comparison of Standard Breath-Held, Free-Breathing, and Compressed Sensing 2D Gradient-Recalled Echo MR Elastography Techniques for Evaluating Liver Stiffness.

OBJECTIVE: The objective of our study was to compare standard breath-held, free-breathing, and compressed sensing (CS) accelerated 2D gradient-recalled echo (GRE) MR elastography (MRE) techniques for measuring liver stiffness. SUBJECTS AND METHODS: Twenty-five adult volunteers (20 without liver disease and five with liver disease) underwent MRE with five different 2D GRE techniques: breath-held; free-breathing; and CS with acceleration factors of 1.5, 2, and 3. Four axial images were obtained through the mid liver with each technique, and liver stiffness measurements (in kilopascals) were made by three blinded readers using elastograms with 95% confidence maps. Liver stiffness measurements were compared between techniques using repeated-measures ANOVA. Absolute agreement between techniques was evaluated for each reader using single-measure intraclass correlation coefficients (ICCs). Bland-Altman analyses were performed to assess bias between techniques using breath-held MRE as the reference. RESULTS: Individual subject mean liver stiffness values ranged from 1.35 to 5.25 kPa for the population. There was no significant difference in mean liver stiffness values between MRE techniques for readers 1 and 3 (p > 0.05). A significant difference (p = 0.02) due to higher stiffness measurements using free-breathing MRE was observed for reader 2. There was excellent absolute agreement between MRE techniques for each reader (all ICCs > 0.940). Bias between techniques ranged from -0.102 to 0.089 kPa for reader 1, -0.119 to 0.121 kPa for reader 2, and -0.074 to 0.085 kPa for reader 3. CONCLUSION: Free-breathing and CS accelerated 2D GRE MRE techniques yield similar results to the conventional breath-held technique with only slight bias and may be useful in pediatric and adult patients with limited ability to breath-hold.

Liver MR Elastography at 3 T: Agreement Across Pulse Sequences and Effect of Liver R2* on Image Quality.

OBJECTIVE: The objectives of our study were to compare MR elastography (MRE) based on gradient-recalled echo (GRE) imaging with spin-echo echo-planar imaging (SEEPI) and rapid fractional (RF)-GRE MRE sequences at 3 T in terms of liver stiffness (LS) and image quality and to evaluate the effect of liver R2* on image quality. MATERIALS AND METHODS: Eighty-one patients underwent 3-T liver MRE with GRE, SE-EPI, and RF-GRE sequences performed in variable order in this study. LS and ROI areas on the LS 95% confidence maps were compared among the three sequences. The relationship between liver R2* and ROI area was investigated. RESULTS: There was no significant difference in mean LS among the three sequences (p = 0.49). Mean ROI area was significantly larger for RF-GRE (18,213 ± 9292 [SD] mm2) than for GRE (13,196 ± 8149 mm2) and SE-EPI (12,896 ± 8656 mm2) (p < 0.0001). Liver R2* was significantly higher among patients with one or more failed sequences (mean ± SD, 116 ± 76 s-1) than for patients with no failed sequences (59 ± 26 s-1) (p = 0.001). Technical failure rates were 10% (8/81), 4% (3/81), and 2% (2/81) for GRE, SE-EPI, and RF-GRE, respectively. Among patients with iron overload (R2* ≥ 100 s-1), there was a trend toward larger ROI area for SE-EPI (p = 0.09). CONCLUSION: SE-EPI-and RF-GRE-based MRE sequences provide equivalent measures of LS compared with GRE-based MRE, and both have lower technical failure rates. The RF-GRE sequence yielded the largest measurable area of LS. Among patients with iron overload, there was a trend toward larger measurable area of LS for the SE-EPI sequence.