Accelerated two-dimensional phase-contrast for cardiovascular MRI using deep learning-based reconstruction with complex difference estimation.

PURPOSE: To develop and validate a deep learning-based reconstruction framework for highly accelerated two-dimensional (2D) phase contrast (PC-MRI) data with accurate and precise quantitative measurements. METHODS: We propose a modified DL-ESPIRiT reconstruction framework for 2D PC-MRI, comprised of an unrolled neural network architecture with a Complex Difference estimation (CD-DL). CD-DL was trained on 155 fully sampled 2D PC-MRI pediatric clinical datasets. The fully sampled data ( n = 29 $$ n=29 $$ ) was retrospectively undersampled (6-11 × $$ \times $$ ) and reconstructed using CD-DL and a parallel imaging and compressed sensing method (PICS). Measurements of peak velocity and total flow were compared to determine the highest acceleration rate that provided accuracy and precision within ± 5 % $$ \pm 5\% $$ . Feasibility of CD-DL was demonstrated on prospectively undersampled datasets acquired in pediatric clinical patients ( n = 5 $$ n=5 $$ ) and compared to traditional parallel imaging (PI) and PICS. RESULTS: The retrospective evaluation showed that 9 × $$ \times $$ accelerated 2D PC-MRI images reconstructed with CD-DL provided accuracy and precision (bias, [95 % $$ \% $$ confidence intervals]) within ± 5 % $$ \pm 5\% $$ . CD-DL showed higher accuracy and precision compared to PICS for measurements of peak velocity (2.8 % $$ \% $$ [ - 2 . 9 $$ -2.9 $$ , 4.5] vs. 3.9 % $$ \% $$ [ - 11 . 0 $$ -11.0 $$ , 4.9]) and total flow (1.8 % $$ \% $$ [ - 3 . 9 $$ -3.9 $$ , 3.4] vs. 2.9 % $$ \% $$ [ - 7 . 1 $$ -7.1 $$ , 6.9]). The prospective feasibility study showed that CD-DL provided higher accuracy and precision than PICS for measurements of peak velocity and total flow. CONCLUSION: In a retrospective evaluation, CD-DL produced quantitative measurements of 2D PC-MRI peak velocity and total flow with ≤ 5 % $$ \le 5\% $$ error in both accuracy and precision for up to 9 × $$ \times $$ acceleration. Clinical feasibility was demonstrated using a prospective clinical deployment of our 8 × $$ \times $$ undersampled acquisition and CD-DL reconstruction in a cohort of pediatric patients.

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

PURPOSE: To develop a method for building MRI reconstruction neural networks robust to changes in signal-to-noise ratio (SNR) and trainable with a limited number of fully sampled scans. METHODS: We propose Noise2Recon, a consistency training method for SNR-robust accelerated MRI reconstruction that can use both fully sampled (labeled) and undersampled (unlabeled) scans. Noise2Recon uses unlabeled data by enforcing consistency between model reconstructions of undersampled scans and their noise-augmented counterparts. Noise2Recon was compared to compressed sensing and both supervised and self-supervised deep learning baselines. Experiments were conducted using retrospectively accelerated data from the mridata three-dimensional fast-spin-echo knee and two-dimensional fastMRI brain datasets. All methods were evaluated in label-limited settings and among out-of-distribution (OOD) shifts, including changes in SNR, acceleration factors, and datasets. An extensive ablation study was conducted to characterize the sensitivity of Noise2Recon to hyperparameter choices. RESULTS: In label-limited settings, Noise2Recon achieved better structural similarity, peak signal-to-noise ratio, and normalized-RMS error than all baselines and matched performance of supervised models, which were trained with 14 × $$ 14\times $$ more fully sampled scans. Noise2Recon outperformed all baselines, including state-of-the-art fine-tuning and augmentation techniques, among low-SNR scans and when generalizing to OOD acceleration factors. Augmentation extent and loss weighting hyperparameters had negligible impact on Noise2Recon compared to supervised methods, which may indicate increased training stability. CONCLUSION: Noise2Recon is a label-efficient reconstruction method that is robust to distribution shifts, such as changes in SNR, acceleration factors, and others, with limited or no fully sampled training data.

Free-breathing Accelerated Cardiac MRI Using Deep Learning: Validation in Children and Young Adults.

Background Obtaining ventricular volumetry and mass is key to most cardiac MRI but challenged by long multibreath-hold acquisitions. Purpose To assess the image quality and performance of a highly accelerated, free-breathing, two-dimensional cine cardiac MRI sequence incorporating deep learning (DL) reconstruction compared with reference standard balanced steady-state free precession (bSSFP). Materials and Methods A DL algorithm was developed to reconstruct custom 12-fold accelerated bSSFP cardiac MRI cine images from coil sensitivity maps using 15 iterations of separable three-dimensional convolutions and data consistency steps. The model was trained, validated, and internally tested in 10, two, and 10 adult human volunteers, respectively, based on vendor partner-supplied fully sampled bSSFP acquisitions. For prospective external clinical validation, consecutive children and young adults undergoing cardiac MRI from September through December 2019 at a single children's hospital underwent both conventional and highly accelerated short-axis bSSFP cine acquisitions in one MRI examination. Two radiologists scored overall and volumetric three-dimensional mesh image quality of all short-axis stacks on a five-point Likert scale and manually segmented endocardial and epicardial contours. Scan times and image quality were compared using the Wilcoxon rank sum test. Measurement agreement was assessed with intraclass correlation coefficient and Bland-Altman analysis. Results Fifty participants (mean age, 16 years ± 4 [standard deviation]; range, 5-30 years; 29 men) were evaluated. The mean prescribed acquisition times of accelerated scans (non-breath-held) and bSSFP (excluding breath-hold time) were 0.9 minute ± 0.3 versus 3.0 minutes ± 1.9 (P < .001). Overall and three-dimensional mesh image quality scores were, respectively, 3.8 ± 0.6 versus 4.3 ± 0.6 (P < .001) and 4.0 ± 1.0 versus 4.4 ± 0.8 (P < .001). Raters had strong agreement between all bSSFP and DL measurements, with intraclass correlation coefficients of 0.76 to 0.97, near-zero mean differences, and narrow limits of agreement. Conclusion With slightly lower image quality yet much faster speed, deep learning reconstruction may allow substantially shorter acquisition times of cardiac MRI compared with conventional balanced steady-state free precession MRI performed for ventricular volumetry. © RSNA, 2021 Online supplemental material is available for this article.

Accelerating cardiac cine MRI using a deep learning-based ESPIRiT reconstruction.

PURPOSE: To propose a novel combined parallel imaging and deep learning-based reconstruction framework for robust reconstruction of highly accelerated 2D cardiac cine MRI data. METHODS: We propose DL-ESPIRiT, an unrolled neural network architecture that utilizes an extended coil sensitivity model to address SENSE-related field-of-view (FOV) limitations in previously proposed deep learning-based reconstruction frameworks. Additionally, we propose a novel neural network design based on (2+1)D spatiotemporal convolutions to produce more accurate dynamic MRI reconstructions than conventional 3D convolutions. The network is trained on fully sampled 2D cardiac cine datasets collected from 11 healthy volunteers with IRB approval. DL-ESPIRiT is compared against a state-of-the-art parallel imaging and compressed sensing method known as l1 -ESPIRiT. The reconstruction accuracy of both methods is evaluated on retrospectively undersampled datasets (R = 12) with respect to standard image quality metrics as well as automatic deep learning-based segmentations of left ventricular volumes. Feasibility of DL-ESPIRiT is demonstrated on two prospectively undersampled datasets acquired in a single heartbeat per slice. RESULTS: The (2+1)D DL-ESPIRiT method produces higher fidelity image reconstructions when compared to l1 -ESPIRiT reconstructions with respect to standard image quality metrics (P < .001). As a result of improved image quality, segmentations made from (2+1)D DL-ESPIRiT images are also more accurate than segmentations from l1 -ESPIRiT images. CONCLUSIONS: DL-ESPIRiT synergistically combines a robust parallel imaging model and deep learning-based priors to produce high-fidelity reconstructions of retrospectively undersampled 2D cardiac cine data acquired with reduced FOV. Although a proof-of-concept is shown, further experiments are necessary to determine the efficacy of DL-ESPIRiT in prospectively undersampled data.

Free-breathing R2∗ mapping of hepatic iron overload in children using 3D multi-echo UTE cones MRI.

PURPOSE: To enable motion-robust, ungated, free-breathing R2∗ mapping of hepatic iron overload in children with 3D multi-echo UTE cones MRI. METHODS: A golden-ratio re-ordered 3D multi-echo UTE cones acquisition was developed with chemical-shift encoding (CSE). Multi-echo complex-valued source images were reconstructed via gridding and coil combination, followed by confounder-corrected R2∗ (=1/ T2∗ ) mapping. A phantom containing 15 different concentrations of gadolinium solution (0-300 mM) was imaged at 3T. 3D multi-echo UTE cones with an initial TE of 0.036 ms and Cartesian CSE-MRI (IDEAL-IQ) sequences were performed. With institutional review board approval, 85 subjects (81 pediatric patients with iron overload + 4 healthy volunteers) were imaged at 3T using 3D multi-echo UTE cones with free breathing (FB cones), IDEAL-IQ with breath holding (BH Cartesian), and free breathing (FB Cartesian). Overall image quality of R2∗ maps was scored by 2 blinded experts and compared by a Wilcoxon rank-sum test. For each pediatric subject, the paired R2∗ maps were assessed to determine if a corresponding artifact-free 15 mm region-of-interest (ROI) could be identified at a mid-liver level on both images. Agreement between resulting R2∗ quantification from FB cones and BH/FB Cartesian was assessed with Bland-Altman and linear correlation analyses. RESULTS: ROI-based regression analysis showed a linear relationship between gadolinium concentration and R2∗ in IDEAL-IQ (y = 8.83x - 52.10, R2 = 0.995) as well as in cones (y = 9.19x - 64.16, R2 = 0.992). ROI-based Bland-Altman analysis showed that the mean difference (MD) was 0.15% and the SD was 5.78%. However, IDEAL-IQ R2∗ measurements beyond 200 mM substantially deviated from a linear relationship for IDEAL-IQ (y = 5.85x + 127.61, R2 = 0.827), as opposed to cones (y = 10.87x - 166.96, R2 = 0.984). In vivo, FB cones R2∗ had similar image quality with BH and FB Cartesian in 15 and 42 cases, respectively. FB cones R2∗ had better image quality scores than BH and FB Cartesian in 3 and 21 cases, respectively, where BH/FB Cartesian exhibited severe ghosting artifacts. ROI-based Bland-Altman analyses were 2.23% (MD) and 6.59% (SD) between FB cones and BH Cartesian and were 0.21% (MD) and 7.02% (SD) between FB cones and FB Cartesian, suggesting a good agreement between FB cones and BH (FB) Cartesian R2∗ . Strong linear relationships were observed between BH Cartesian and FB cones (y = 1.00x + 1.07, R2 = 0.996) and FB Cartesian and FB cones (y = 0.98x + 1.68, R2 = 0.999). CONCLUSION: Golden-ratio re-ordered 3D multi-echo UTE Cones MRI enabled motion-robust, ungated, and free-breathing R2∗ mapping of hepatic iron overload, with comparable R2∗ measurements and image quality to BH Cartesian, and better image quality than FB Cartesian.

Near-Silent and Distortion-Free Diffusion MRI in Pediatric Musculoskeletal Disorders: Comparison With Echo Planar Imaging Diffusion.

BACKGROUND: Diffusion-weighted imaging (DWI) is common for evaluating pediatric musculoskeletal lesions, but suffers from geometric distortion and intense acoustic noise. PURPOSE: To investigate the performance of a near-silent and distortion-free DWI sequence (DW-SD) relative to standard echo-planar DWI (DW-EPI) in pediatric extremity MRI. STUDY TYPE: Prospective validation study. SUBJECTS: Thirty-nine children referred for extremity MRI. FIELD STRENGTH/SEQUENCE: DW-EPI and DW-SD, based on a rotating ultrafast sequence modified with sinusoidal diffusion preparation gradients, at 3T. ASSESSMENT: DW-SD image quality (Sanat ) was assessed from 0 (nondiagnostic) to 5 (outstanding) and comparative image quality (Scomp ) (from -2 = DW-EPI more delineated to +2 = DW-SD more delineated, 0 = same). ADC measured by DW-SD and DW-EPI were compared in bone marrow, muscle, and lesions. STATISTICAL TESTS: Wilcoxon rank-sum test and confidence interval of proportions (CIOP) were calculated for Scomp , Student's t-test, coefficient of variation (COV), and Bland-Altman analysis for ADC values, and intraclass correlation coefficient (ICC) for interreader agreement. RESULTS: DW-SD and DW-EPI ADC values for bone marrow, muscle, and lesions were not significantly different (P = 0.3, P = 0.2, and P = 0.27, respectively) and had an overall ADC COV of 14.8% (95% confidence interval: 12.3%, 16.9%) and no significant proportional bias on Bland-Altman analysis. Sanat CIOP was rated diagnostic or better (score of 3, 4, or 5) in 72-98% of cases for bone marrow, muscle, and soft tissues. DW-SD was equivalent to or preferred over DW-EPI in muscles and soft tissues, with CIOP 86-93% and 93%, respectively. Lesions were equally visualized on DW-SD and DW-EPI in 40-51%, with DW-SD preferred in 44-56% of cases. DW-SD was rated significantly better than DW-EPI across all comparative variables that included bone marrow, muscle, soft tissue, cartilage, and lesions (P < 0.05). Readers had moderate to near-perfect (ICC range = 0.45-0.85). DATA CONCLUSION: DW-SD of the extremities provided similar ADC values and improved image quality compared with conventional DW-EPI. Level of Evidence 2 Technical Efficacy Stage 2 J. MAGN. RESON. IMAGING 2021;53:504-513.

Prospective Deployment of Deep Learning in MRI: A Framework for Important Considerations, Challenges, and Recommendations for Best Practices.

Artificial intelligence algorithms based on principles of deep learning (DL) have made a large impact on the acquisition, reconstruction, and interpretation of MRI data. Despite the large number of retrospective studies using DL, there are fewer applications of DL in the clinic on a routine basis. To address this large translational gap, we review the recent publications to determine three major use cases that DL can have in MRI, namely, that of model-free image synthesis, model-based image reconstruction, and image or pixel-level classification. For each of these three areas, we provide a framework for important considerations that consist of appropriate model training paradigms, evaluation of model robustness, downstream clinical utility, opportunities for future advances, as well recommendations for best current practices. We draw inspiration for this framework from advances in computer vision in natural imaging as well as additional healthcare fields. We further emphasize the need for reproducibility of research studies through the sharing of datasets and software. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY STAGE: 2.

Diffusion-weighted double-echo steady-state with a three-dimensional cones trajectory for non-contrast-enhanced breast MRI.

The image quality limitations of echo-planar diffusion-weighted imaging (DWI) are an obstacle to its widespread adoption in the breast. Steady-state DWI is an alternative DWI method with more robust image quality but its contrast for imaging breast cancer is not well-understood. The aim of this study was to develop and evaluate diffusion-weighted double-echo steady-state imaging with a three-dimensional cones trajectory (DW-DESS-Cones) as an alternative to conventional DWI for non-contrast-enhanced MRI in the breast. This prospective study included 28 women undergoing clinically indicated breast MRI and six asymptomatic volunteers. In vivo studies were performed at 3 T and included DW-DESS-Cones, DW-DESS-Cartesian, DWI, and CE-MRI acquisitions. Phantom experiments (diffusion phantom, High Precision Devices) and simulations were performed to establish framework for contrast of DW-DESS-Cones in comparison to DWI in the breast. Motion artifacts of DW-DESS-Cones were measured with artifact-to-noise ratio in volunteers and patients. Lesion-to-fibroglandular tissue signal ratios were measured, lesions were categorized as hyperintense or hypointense, and an image quality observer study was performed in DW-DESS-Cones and DWI in patients. Effect of DW-DESS-Cones method on motion artifacts was tested by mixed-effects generalized linear model. Effect of DW-DESS-Cones on signal in phantom was tested by quadratic regression. Correlation was calculated between DW-DESS-Cones and DWI lesion-to-fibroglandular tissue signal ratios. Inter-observer agreement was assessed with Gwet's AC. Simulations predicted hyperintensity of lesions with DW-DESS-Cones but at a 3% to 67% lower degree than with DWI. Motion artifacts were reduced with DW-DESS-Cones versus DW-DESS-Cartesian (p < 0.05). Lesion-to-fibroglandular tissue signal ratios were not correlated between DW-DESS-Cones and DWI (r = 0.25, p = 0.38). Concordant hyperintensity/hypointensity was observed between DW-DESS-Cones and DWI in 11/14 lesions. DW-DESS-Cones improved sharpness, distortion, and overall image quality versus DWI. DW-DESS-Cones may be able to eliminate motion artifacts in the breast allowing for investigation of higher degrees of steady-state diffusion weighting. Malignant breast lesions in DW-DESS-Cones demonstrated hyperintensity with respect to surrounding tissue without an injection of contrast. LEVEL OF EVIDENCE: 2. TECHNICAL EFFICACY STAGE: 1.

Reconstruction of undersampled 3D non-Cartesian image-based navigators for coronary MRA using an unrolled deep learning model.

PURPOSE: To rapidly reconstruct undersampled 3D non-Cartesian image-based navigators (iNAVs) using an unrolled deep learning (DL) model, enabling nonrigid motion correction in coronary magnetic resonance angiography (CMRA). METHODS: An end-to-end unrolled network is trained to reconstruct beat-to-beat 3D iNAVs acquired during a CMRA sequence. The unrolled model incorporates a nonuniform FFT operator in TensorFlow to perform the data-consistency operation, and the regularization term is learned by a convolutional neural network (CNN) based on the proximal gradient descent algorithm. The training set includes 6,000 3D iNAVs acquired from 7 different subjects and 11 scans using a variable-density (VD) cones trajectory. For testing, 3D iNAVs from 4 additional subjects are reconstructed using the unrolled model. To validate reconstruction accuracy, global and localized motion estimates from DL model-based 3D iNAVs are compared with those extracted from 3D iNAVs reconstructed with l1 -ESPIRiT. Then, the high-resolution coronary MRA images motion corrected with autofocusing using the l1 -ESPIRiT and DL model-based 3D iNAVs are assessed for differences. RESULTS: 3D iNAVs reconstructed using the DL model-based approach and conventional l1 -ESPIRiT generate similar global and localized motion estimates and provide equivalent coronary image quality. Reconstruction with the unrolled network completes in a fraction of the time compared to CPU and GPU implementations of l1 -ESPIRiT (20× and 3× speed increases, respectively). CONCLUSIONS: We have developed a deep neural network architecture to reconstruct undersampled 3D non-Cartesian VD cones iNAVs. Our approach decreases reconstruction time for 3D iNAVs, while preserving the accuracy of nonrigid motion information offered by them for correction.

Near-silent distortionless DWI using magnetization-prepared RUFIS.

PURPOSE: To develop a near-silent and distortionless DWI (sd-DWI) sequence using magnetization-prepared rotating ultrafast imaging sequence. METHODS: A rotating ultrafast imaging sequence was modified with driven-equilibrium diffusion preparation, including eddy-current compensation methods. To compensate for the T1 recovery during readout, a phase-cycling method was used. Both compensation methods were validated in phantoms. The optimized sequence was compared with an EPI diffusion sequence for image distortion, contrast, ADC, and acoustic noise level in phantoms. The sequence was evaluated in 1 brain volunteer, 5 prostate volunteers, and 10 pediatric patients with joint diseases. RESULTS: Combination of several eddy-current compensation methods reduced the artifact to an acceptable level. Phase cycling reduced T1 recovery contamination during readout. In phantom scans, the optimized sequence generated similar image contrast to the EPI diffusion sequence, and ADC maps between the sequences were comparable; sd-DWI had significantly lower acoustic noise (P < .05). In vivo brain scan showed reduced image distortion in sd-DWI compared with the EPI diffusion, although residual motion artifact remains due to brain pulsation. The prostate scans showed that sd-DWI can provide similar ADC compared with EPI diffusion, with no image distortion. Patient scans showed that the sequence can clearly depict joint lesions. CONCLUSION: An sd-DWI sequence was developed and optimized. Compared with conventional EPI diffusion, sd-DWI provided similar diffusion contrast, accurate ADC measurement, improved image quality, and minimal ambient scanning noise. The sequence showed the ability to obtain in vivo diffusion contrast in relatively motion-free body regions, such as prostate and joint.

Rosette Trajectories Enable Ungated, Motion-Robust, Simultaneous Cardiac and Liver T2 * Iron Assessment.

BACKGROUND: Quantitative T2 * MRI is the standard of care for the assessment of iron overload. However, patient motion corrupts T2 * estimates. PURPOSE: To develop and evaluate a motion-robust, simultaneous cardiac and liver T2 * imaging approach using non-Cartesian, rosette sampling and a model-based reconstruction as compared to clinical-standard Cartesian MRI. STUDY TYPE: Prospective. PHANTOM/POPULATION: Six ferumoxytol-containing phantoms (26-288 µg/mL). Eight healthy subjects and 18 patients referred for clinically indicated iron overload assessment. FIELD STRENGTH/SEQUENCE: 1.5T, 2D Cartesian and rosette gradient echo (GRE) ASSESSMENT: GRE T2 * values were validated in ferumoxytol phantoms. In healthy subjects, test-retest and spatial coefficient of variation (CoV) analysis was performed during three breathing conditions. Cartesian and rosette T2 * were compared using correlation and Bland-Altman analysis. Images were rated by three experienced radiologists on a 5-point scale. STATISTICAL TESTS: Linear regression, analysis of variance (ANOVA), and paired Student's t-testing were used to compare reproducibility and variability metrics in Cartesian and rosette scans. The Wilcoxon rank test was used to assess reader score comparisons and reader reliability was measured using intraclass correlation analysis. RESULTS: Rosette R2* (1/T2 *) was linearly correlated with ferumoxytol concentration (r2 = 1.00) and not significantly different than Cartesian values (P = 0.16). During breath-holding, ungated rosette liver and heart T2 * had lower spatial CoV (liver: 18.4 ± 9.3% Cartesian, 8.8% ± 3.4% rosette, P = 0.02, heart: 37.7% ± 14.3% Cartesian, 13.4% ± 1.7% rosette, P = 0.001) and higher-quality scores (liver: 3.3 [3.0-3.6] Cartesian, 4.7 [4.1-4.9] rosette, P = 0.005, heart: 3.0 [2.3-3] Cartesian, 4.5 [3.8-5.0] rosette, P = 0.005) compared to Cartesian values. During free-breathing and failed breath-holding, Cartesian images had very poor to average image quality with significant artifacts, whereas rosette remained very good, with minimal artifacts (P = 0.001). DATA CONCLUSION: Rosette k-sampling with a model-based reconstruction offers a clinically useful motion-robust T2 * mapping approach for iron quantification. J. MAGN. RESON. IMAGING 2020;52:1688-1698.

Compressed Sensing: From Research to Clinical Practice with Deep Neural Networks.

Compressed sensing (CS) reconstruction methods leverage sparse structure in underlying signals to recover high-resolution images from highly undersampled measurements. When applied to magnetic resonance imaging (MRI), CS has the potential to dramatically shorten MRI scan times, increase diagnostic value, and improve overall patient experience. However, CS has several shortcomings which limit its clinical translation such as: 1) artifacts arising from inaccurate sparse modelling assumptions, 2) extensive parameter tuning required for each clinical application, and 3) clinically infeasible reconstruction times. Recently, CS has been extended to incorporate deep neural networks as a way of learning complex image priors from historical exam data. Commonly referred to as unrolled neural networks, these techniques have proven to be a compelling and practical approach to address the challenges of sparse CS. In this tutorial, we will review the classical compressed sensing formulation and outline steps needed to transform this formulation into a deep learning-based reconstruction framework. Supplementary open source code in Python will be used to demonstrate this approach with open databases. Further, we will discuss considerations in applying unrolled neural networks in the clinical setting.

Myocardial T2* mapping: influence of noise on accuracy and precision.

BACKGROUND: Pixel-wise, parametric T2* mapping is emerging as a means of automatic measurement of iron content in tissues. It enables quick, intuitive interpretation and provides the potential benefit of spatial context between tissues. However, pixel-wise mapping uses much lower SNR data to estimate T2* when compared to region-based mapping thereby decreasing both its accuracy and precision. In this study, the effects that noise has on the precision and accuracy of pixel-wise T2* mapping were investigated and techniques to mitigate those effects are proposed. METHODS: To study precision across T2* mapping techniques, a pipeline to estimate the pixel-wise standard deviation (SD) of the T2* based on the fit residuals is proposed. For validation, a Monte-Carlo analysis was performed in which T2* phantoms were scanned N = 64 times, the true SD was measured and compared to the estimated SD. To improve accuracy and precision, the automatic truncation method for mitigating noise bias was extended to pixel-wise fitting by using an SNR scaled image reconstruction and truncating low SNR measurements. Finally, the precision and accuracy of non-linear regression with and without automatic truncation, were investigated using Monte-Carlo simulations. RESULTS: Measured and estimated SD's were >99.9% correlated for non-linear regression with and without truncation. Non-linear regression with automatic truncation was shown to be the best mapping technique for improving accuracy and precision in low T2* and low SNR measurements. CONCLUSIONS: A method for applying an automatic truncation method to pixel-wise T2* mapping that reduces T2* overestimation due to noise bias was proposed. A formulation for estimating pixel-wise standard deviation (SD) maps for T2* that can serve as a quality map for interpreting images and for comparison of imaging protocols was also proposed and validated.

Free-breathing T2* mapping using respiratory motion corrected averaging.

BACKGROUND: Pixel-wise T2* maps based on breath-held segmented image acquisition are prone to ghost artifacts in instances of poor breath-holding or cardiac arrhythmia. Single shot imaging is inherently immune to ghost type artifacts. We propose a free-breathing method based on respiratory motion corrected single shot imaging with averaging to improve the signal to noise ratio. METHODS: Images were acquired using a multi-echo gradient recalled echo sequence and T2* maps were calculated at each pixel by exponential fitting. For 40 subjects (2 cohorts), two acquisition protocols were compared: (1) a breath-held, segmented acquisition, and (2) a free-breathing, single-shot multiple repetition respiratory motion corrected average. T2* measurements in the interventricular septum and liver were compared for the 2-methods in all studies with diagnostic image quality. RESULTS: In cohort 1 (N = 28) with age 51.4 ± 17.6 (m ± SD) including 1 subject with severe myocardial iron overload, there were 8 non-diagnostic breath-held studies due to poor image quality resulting from ghost artifacts caused by respiratory motion or arrhythmias. In cohort 2 (N = 12) with age 30.9 ± 7.5 (m ± SD), including 7 subjects with severe myocardial iron overload and 4 subjects with mild iron overload, a single subject was unable to breath-hold. Free-breathing motion corrected T2* maps were of diagnostic quality in all 40 subjects. T2* measurements were in excellent agreement (In cohort #1, T2*FB = 0.95 x T2*BH + 0.41, r2 = 0.93, N = 39 measurements, and in cohort #2, T2*FB = 0.98 x T2*BH + 0.05, r2 > 0.99, N = 22 measurements). CONCLUSIONS: A free-breathing approach to T2* mapping is demonstrated to produce consistently good quality maps in the presence of respiratory motion and arrhythmias.

Deep Learning Initialized Compressed Sensing (Deli-CS) in Volumetric Spatio-Temporal Subspace Reconstruction.

Introduction: Spatio-temporal MRI methods enable whole-brain multi-parametric mapping at ultra-fast acquisition times through efficient k-space encoding, but can have very long reconstruction times, which limit their integration into clinical practice. Deep learning (DL) is a promising approach to accelerate reconstruction, but can be computationally intensive to train and deploy due to the large dimensionality of spatio-temporal MRI. DL methods also need large training data sets and can produce results that don't match the acquired data if data consistency is not enforced. The aim of this project is to reduce reconstruction time using DL whilst simultaneously limiting the risk of deep learning induced hallucinations, all with modest hardware requirements. Methods: Deep Learning Initialized Compressed Sensing (Deli-CS) is proposed to reduce the reconstruction time of iterative reconstructions by "kick-starting" the iterative reconstruction with a DL generated starting point. The proposed framework is applied to volumetric multi-axis spiral projection MRF that achieves whole-brain T1 and T2 mapping at 1-mm isotropic resolution for a 2-minute acquisition. First, the traditional reconstruction is optimized from over two hours to less than 40 minutes while using more than 90% less RAM and only 4.7 GB GPU memory, by using a memory-efficient GPU implementation. The Deli-CS framework is then implemented and evaluated against the above reconstruction. Results: Deli-CS achieves comparable reconstruction quality with 50% fewer iterations bringing the full reconstruction time to 20 minutes. Conclusion: Deli-CS reduces the reconstruction time of subspace reconstruction of volumetric spatio-temporal acquisitions by providing a warm start to the iterative reconstruction algorithm.

Deep Learning-Based Water-Fat Separation from Dual-Echo Chemical Shift-Encoded Imaging.

Conventional water-fat separation approaches suffer long computational times and are prone to water/fat swaps. To solve these problems, we propose a deep learning-based dual-echo water-fat separation method. With IRB approval, raw data from 68 pediatric clinically indicated dual echo scans were analyzed, corresponding to 19382 contrast-enhanced images. A densely connected hierarchical convolutional network was constructed, in which dual-echo images and corresponding echo times were used as input and water/fat images obtained using the projected power method were regarded as references. Models were trained and tested using knee images with 8-fold cross validation and validated on out-of-distribution data from the ankle, foot, and arm. Using the proposed method, the average computational time for a volumetric dataset with ~400 slices was reduced from 10 min to under one minute. High fidelity was achieved (correlation coefficient of 0.9969, l1 error of 0.0381, SSIM of 0.9740, pSNR of 58.6876) and water/fat swaps were mitigated. I is of particular interest that metal artifacts were substantially reduced, even when the training set contained no images with metallic implants. Using the models trained with only contrast-enhanced images, water/fat images were predicted from non-contrast-enhanced images with high fidelity. The proposed water-fat separation method has been demonstrated to be fast, robust, and has the added capability to compensate for metal artifacts.

Upstream Machine Learning in Radiology.

Machine learning (ML) and Artificial intelligence (AI) has the potential to dramatically improve radiology practice at multiple stages of the imaging pipeline. Most of the attention has been garnered by applications focused on improving the end of the pipeline: image interpretation. However, this article reviews how AI/ML can be applied to improve upstream components of the imaging pipeline, including exam modality selection, hardware design, exam protocol selection, data acquisition, image reconstruction, and image processing. A breadth of applications and their potential for impact is shown across multiple imaging modalities, including ultrasound, computed tomography, and MRI.

DIAGNOSTIC IMAGE QUALITY ASSESSMENT AND CLASSIFICATION IN MEDICAL IMAGING: OPPORTUNITIES AND CHALLENGES.

Magnetic Resonance Imaging (MRI) suffers from several artifacts, the most common of which are motion artifacts. These artifacts often yield images that are of non-diagnostic quality. To detect such artifacts, images are prospectively evaluated by experts for their diagnostic quality, which necessitates patient-revisits and rescans whenever non-diagnostic quality scans are encountered. This motivates the need to develop an automated framework capable of accessing medical image quality and detecting diagnostic and non-diagnostic images. In this paper, we explore several convolutional neural network-based frameworks for medical image quality assessment and investigate several challenges therein.

Unsupervised clustering method to convert high-resolution magnetic resonance volumes to three-dimensional acoustic models for full-wave ultrasound simulations.

Simulations of acoustic wave propagation, including both the forward and the backward propagations of the wave (also known as full-wave simulations), are increasingly utilized in ultrasound imaging due to their ability to more accurately model important acoustic phenomena. Realistic anatomic models, particularly those of the abdominal wall, are needed to take full advantage of the capabilities of these simulation tools. We describe a method for converting fat-water-separated magnetic resonance imaging (MRI) volumes to anatomical models for ultrasound simulations. These acoustic models are used to map acoustic imaging parameters, such as speed of sound and density, to grid points in an ultrasound simulation. The tissues of these models are segmented from the MRI volumes into five primary classes of tissue in the human abdominal wall (skin, fat, muscle, connective tissue, and nontissue). This segmentation is achieved using an unsupervised machine learning algorithm, fuzzy c-means clustering (FCM), on a multiscale feature representation of the MRI volumes. We describe an automated method for utilizing FCM weights to produce a model that achieves ∼ 90 % agreement with manual segmentation. Two-dimensional (2-D) and three-dimensional (3-D) full-wave nonlinear ultrasound simulations are conducted, demonstrating the utility of realistic 3-D abdominal wall models over previously available 2-D abdominal wall models.