Prospectively accelerated dynamic speech magnetic resonance imaging at 3 T using a self-navigated spiral-based manifold regularized scheme.

This work develops and evaluates a self-navigated variable density spiral (VDS)-based manifold regularization scheme to prospectively improve dynamic speech magnetic resonance imaging (MRI) at 3 T. Short readout duration spirals (1.3-ms long) were used to minimize sensitivity to off-resonance. A custom 16-channel speech coil was used for improved parallel imaging of vocal tract structures. The manifold model leveraged similarities between frames sharing similar vocal tract postures without explicit motion binning. The self-navigating capability of VDS was leveraged to learn the Laplacian structure of the manifold. Reconstruction was posed as a sensitivity-encoding-based nonlocal soft-weighted temporal regularization scheme. Our approach was compared with view-sharing, low-rank, temporal finite difference, extra dimension-based sparsity reconstruction constraints. Undersampling experiments were conducted on five volunteers performing repetitive and arbitrary speaking tasks at different speaking rates. Quantitative evaluation in terms of mean square error over moving edges was performed in a retrospective undersampling experiment on one volunteer. For prospective undersampling, blinded image quality evaluation in the categories of alias artifacts, spatial blurring, and temporal blurring was performed by three experts in voice research. Region of interest analysis at articulator boundaries was performed in both experiments to assess articulatory motion. Improved performance with manifold reconstruction constraints was observed over existing constraints. With prospective undersampling, a spatial resolution of 2.4 × 2.4 mm2 /pixel and a temporal resolution of 17.4 ms/frame for single-slice imaging, and 52.2 ms/frame for concurrent three-slice imaging, were achieved. We demonstrated implicit motion binning by analyzing the mechanics of the Laplacian matrix. Manifold regularization demonstrated superior image quality scores in reducing spatial and temporal blurring compared with all other reconstruction constraints. While it exhibited faint (nonsignificant) alias artifacts that were similar to temporal finite difference, it provided statistically significant improvements compared with the other constraints. In conclusion, the self-navigated manifold regularized scheme enabled robust high spatiotemporal resolution dynamic speech MRI at 3 T.

Reducing MRI-guided radiotherapy planning and delivery times via efficient leaf sequencing and segment shape optimization algorithms.

Objective.Extended treatment session times are an operational limitation in magnetic resonance imaging guided adaptive radiotherapy (MRIgRT). In this study a novel leaf sequencing algorithm called optimal fluence levels (OFL) and an optimization algorithm called pseudo gradient descent (PGD) are evaluated with respect to plan quality, beam complexity, and the ability to reduce treatment session times on the Elekta Unity MRIgRT system.Approach.Ten total patients were evaluated on this Institutional Review Board approved study: three with prostate cancer, three with oligometastases, two with pancreatic cancer, and two with liver cancer. Plans were generated using the clinical Monaco Hyperion optimizer and leaf sequencer and then re-optimized using OFL and PGD (OFL + PGD) while holding all IMRT constraints and planning parameters constant. All plans were normalized to ensure 95% of the PTV received the prescription dose. A paired t-test was used to evaluate statistical significance.Main Results.Plan quality in terms of dosimetric OAR sparing was found to be equivalent between the OFL + PGD and conventional Monaco Hyperion optimizer plans. The OFL + PGD plans had a reduction in optimization time of 51.4% ± 5.0% (p = 0.002) and reduction in treatment delivery time of 10.6% ± 7.5% (p = 0.005). OFL + PGD generated plans had on average 13.2% ± 12.6% fewer multi-leaf collimator (MLC) segments (p = 0.009) and 0.1 ± 0.1 lower plan averaged beam modulation (PM) (p = 0.004) relative to the Monaco Hyperion plans.Significance.The OFL + PGD algorithms more quickly generate Unity treatment plans that are faster to deliver than with the conventional approach and without compromising dosimetric plan quality. This is likely due to a delivery complexity reduction enabled by OFL + PGD relative to the Monaco Hyperion plans.

DYNAMIC IMAGING USING DEEP BILINEAR UNSUPERVISED LEARNING (DEBLUR).

Bilinear models such as low-rank and compressed sensing, which decompose the dynamic data to spatial and temporal factors, are powerful and memory efficient tools for the recovery of dynamic MRI data. These methods rely on sparsity and energy compaction priors on the factors to regularize the recovery. Motivated by deep image prior, we introduce a novel bilinear model, whose factors are regularized using convolutional neural networks. To reduce the run time, we initialize the CNN parameters by pre-training them on pre-acquired data with longer acquistion time. Since fully sampled data is not available, pretraining is performed on undersampled data in an unsupervised fashion. We use sparsity regularization of the network parameters to minimize the overfitting of the network to measurement noise. Our experiments on on free-breathing and ungated cardiac CINE data acquired using a navigated golden-angle gradient-echo radial sequence show the ability of our method to provide reduced spatial blurring as compared to low-rank and SToRM reconstructions.

DEEP GENERATIVE STORM MODEL FOR DYNAMIC IMAGING.

We introduce a novel generative smoothness regularization on manifolds (SToRM) model for the recovery of dynamic image data from highly undersampled measurements. The proposed generative framework represents the image time series as a smooth non-linear function of low-dimensional latent vectors that capture the cardiac and respiratory phases. The non-linear function is represented using a deep convolutional neural network (CNN). Unlike the popular CNN approaches that require extensive fully-sampled training data that is not available in this setting, the parameters of the CNN generator as well as the latent vectors are jointly estimated from the undersampled measurements using stochastic gradient descent. We penalize the norm of the gradient of the generator to encourage the learning of a smooth surface/manifold, while temporal gradients of the latent vectors are penalized to encourage the time series to be smooth. The main benefits of the proposed scheme are (a) the quite significant reduction in memory demand compared to the analysis based SToRM model, and (b) the spatial regularization brought in by the CNN model. We also introduce efficient progressive approaches to minimize the computational complexity of the algorithm.

Dynamic Imaging Using a Deep Generative SToRM (Gen-SToRM) Model.

We introduce a generative smoothness regularization on manifolds (SToRM) model for the recovery of dynamic image data from highly undersampled measurements. The model assumes that the images in the dataset are non-linear mappings of low-dimensional latent vectors. We use the deep convolutional neural network (CNN) to represent the non-linear transformation. The parameters of the generator as well as the low-dimensional latent vectors are jointly estimated only from the undersampled measurements. This approach is different from traditional CNN approaches that require extensive fully sampled training data. We penalize the norm of the gradients of the non-linear mapping to constrain the manifold to be smooth, while temporal gradients of the latent vectors are penalized to obtain a smoothly varying time-series. The proposed scheme brings in the spatial regularization provided by the convolutional network. The main benefit of the proposed scheme is the improvement in image quality and the orders-of-magnitude reduction in memory demand compared to traditional manifold models. To minimize the computational complexity of the algorithm, we introduce an efficient progressive training-in-time approach and an approximate cost function. These approaches speed up the image reconstructions and offers better reconstruction performance.

Three-Dimensional GRE T1ρ mapping of the brain using tailored variable flip-angle scheduling.

PURPOSE: To introduce a new approach called tailored variable flip-angle (VFA) scheduling for SNR-efficient 3D T1ρ mapping of the brain using a magnetization-prepared gradient-echo sequence. METHODS: Simulations were used to assess the relative SNR efficiency, quantitative accuracy, and spatial blurring of tailored VFA scheduling for T1ρ mapping of brain tissue compared with magnetization-prepared angle-modulated partitioned k-space spoiled gradient-echo snapshots (MAPSS), a state-of-the-art technique for accurate 3D gradient-echo T1ρ mapping. Simulations were also used to calculate optimal imaging parameters for tailored VFA scheduling versus MAPSS, without and with nulling of CSF. Four participants were imaged at 3T MRI to demonstrate the feasibility of tailored VFA scheduling for T1ρ mapping of the brain. Using MAPSS as a reference standard, in vivo data were used to validate the relative SNR efficiency and quantitative accuracy of the new approach. RESULTS: Tailored VFA scheduling can provide a 2-fold to 4-fold gain in the SNR of the resulting T1ρ map as compared with MAPSS when using identical sequence parameters while limiting T1ρ quantification errors to 2% or less. In vivo whole-brain 3D T1ρ maps acquired with tailored VFA scheduling had superior SNR efficiency than is achievable with MAPSS, and the SNR efficiency improved with a greater number of views per segment. CONCLUSIONS: Tailored VFA scheduling is an SNR-efficient GRE technique for 3D T1ρ mapping of the brain that provides increased flexibility in choice of imaging parameters compared with MAPSS, which may benefit a variety of applications.

MULTI-ECHO RECOVERY WITH FIELD INHOMOGENEITY COMPENSATION USING STRUCTURED LOW-RANK MATRIX COMPLETION.

Echo-planar imaging (EPI), which is the main workhorse of functional MRI, suffers from field inhomogeneity-induced geometric distortions. The amount of distortion is proportional to the readout duration, which restricts the maximum achievable spatial resolution. The spatially varying nature of the T 2 * decay makes it challenging for EPI schemes with a single echo time to obtain good sensitivity to functional activations in different brain regions. Despite the use of parallel MRI and multislice acceleration, the number of different echo times that can be acquired in a reasonable TR is limited. The main focus of this work is to introduce a rosette-based acquisition scheme and a structured low-rank reconstruction algorithm to overcome the above challenges. The proposed scheme exploits the exponential structure of the time series to recover distortion-free images from several echoes simultaneously.

Pulmonary ventilation imaging in asthma and cystic fibrosis using oxygen-enhanced 3D radial ultrashort echo time MRI.

BACKGROUND: A previous study demonstrated the feasibility of using 3D radial ultrashort echo time (UTE) oxygen-enhanced MRI (UTE OE-MRI) for functional imaging of healthy human lungs. The repeatability of quantitative measures from UTE OE-MRI needs to be established prior to its application in clinical research. PURPOSE: To evaluate repeatability of obstructive patterns in asthma and cystic fibrosis (CF) with UTE OE-MRI with isotropic spatial resolution and full chest coverage. STUDY TYPE: Volunteer and patient repeatability. POPULATION: Eighteen human subjects (five asthma, six CF, and seven normal subjects). FIELD STRENGTH/SEQUENCE: Respiratory-gated free-breathing 3D radial UTE (80 µs) sequence at 1.5T. ASSESSMENT: Two 3D radial UTE volumes were acquired sequentially under normoxic and hyperoxic conditions. A subset of subjects underwent repeat acquisitions on either the same day or ≤15 days apart. Asthma and CF subjects also underwent spirometry. A workflow including deformable registration and retrospective lung density correction was used to compute 3D isotropic percent signal enhancement (PSE) maps. Median PSE (MPSE) and ventilation defect percent (VDP) of the lung were measured from the PSE map. STATISTICAL TESTS: The relations between MPSE, VDP, and spirometric measures were assessed using Spearman correlations. The test-retest repeatability was evaluated using Bland-Altman analysis and intraclass correlation coefficients (ICC). RESULTS: Ventilation measures in normal subjects (MPSE = 8.0%, VDP = 3.3%) were significantly different from those in asthma (MPSE = 6.0%, P = 0.042; VDP = 21.7%, P = 0.018) and CF group (MPSE = 4.5%, P = 0.0006; VDP = 27.2%, P = 0.002). MPSE correlated significantly with forced expiratory lung volume in 1 second percent predicted (ρ = 0.72, P = 0.017). The ICC of the test-retest VDP and MPSE were both ≥0.90. In all subject groups, an anterior/posterior gradient was observed with higher MPSE and lower VDP in the posterior compared to anterior regions (P ≤ 0.0021 for all comparisons). DATA CONCLUSION: 3D radial UTE OE-MRI supports quantitative differentiation of diseased vs. healthy lungs using either whole lung VDP or MPSE with excellent test-retest repeatability. LEVEL OF EVIDENCE: 2 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2018;47:1287-1297.

Regional Heterogeneity of Lobar Ventilation in Asthma Using Hyperpolarized Helium-3 MRI.

RATIONALE AND OBJECTIVES: To determine lobar ventilation patterns in asthmatic lungs with hyperpolarized 3He magnetic resonance imaging (HP 3He MRI). MATERIALS AND METHODS: Eighty-two subjects (14 normal, 48 mild-to-moderate asthma, and 20 severe asthma) underwent HP 3He MRI, computed tomography (CT), and pulmonary function testing. After registering proton to 3He images, we segmented the lungs from proton MRI and further segmented the five lung lobes (right upper lobe [RUL], right middle lobe [RML], and right lower lobe [RLL]; left upper lobe and left lower lobe [LLL]) by referring to the lobar segmentation from CT. We classified the gas volume into four signal intensity levels as follows: ventilation defect percent (VDP), low ventilation percent, medium ventilation percent, and high ventilation percent. The local signal intensity variations in the ventilated volume were estimated using heterogeneity score (Hs). We compared each ventilation level and Hs measured in the whole lung and lobar regions across the three subject groups. RESULTS: In mild-to-moderate asthma, the RML and RUL showed significantly greater VDP than the two lower lobes (RLL and LLL) (P ≤ .047). In severe asthma, the pattern was more variable with the VDP in the RUL significantly greater than in the RLL (P = .026). In both asthma groups, the lower lobes (RLL and LLL) showed significantly higher high ventilation percent and Hs compared to the three upper lobes (all P ≤ .015). CONCLUSIONS: In asthma, the RML and RUL showed greater ventilation abnormalities, and the RLL and LLL were more highly ventilated with greater local heterogeneity. These findings may facilitate guided bronchoscopic sampling and localized airway treatment in future studies.

Semiautomated Ventilation Defect Quantification in Exercise-induced Bronchoconstriction Using Hyperpolarized Helium-3 Magnetic Resonance Imaging: A Repeatability Study.

RATIONALE AND OBJECTIVES: This study aimed to compare the performance of a semiautomated ventilation defect segmentation approach, adaptive K-means, with manual segmentation of hyperpolarized helium-3 magnetic resonance imaging in subjects with exercise-induced bronchoconstriction (EIB). MATERIALS AND METHODS: Six subjects with EIB underwent hyperpolarized helium-3 magnetic resonance imaging and spirometry tests at baseline, post exercise, and recovery over two separate visits. Ventilation defects were analyzed by two methods. First, two independent readers manually segmented ventilation defects. Second, defects were quantified by an adaptive K-means method that corrected for coil sensitivity, applied a vesselness filter to estimate pulmonary vasculature, and segmented defects adaptively based on the overall low-intensity signals in the lungs. These two methods were then compared in four aspects: (1) ventilation defect percent (VDP) measurements, (2) correlation between spirometric measures and measured VDP, (3) regional VDP variations pre- and post exercise challenge, and (4) Dice coefficient for spatial agreement. RESULTS: The adaptive K-means method was ~5 times faster, and the measured VDP bias was under 2%. The correlation between predicted forced expiratory volume in 1 second over forced vital capacity and VDP measured by adaptive K-means (ρ = -0.64, P <0.0001) and by the manual method (ρ = -0.63, P <0.0001) yielded almost identical 95% confidence intervals. Neither method of measuring VDP indicated apical/basal or anterior dependence in this small study cohort. CONCLUSIONS: Compared to the manual method, the adaptive K-means method provided faster, reproducible, comparable measures of VDP in EIB and may be applied to a variety of lung diseases.

Functional imaging of the lungs with gas agents.

This review focuses on the state-of-the-art of the three major classes of gas contrast agents used in magnetic resonance imaging (MRI)-hyperpolarized (HP) gas, molecular oxygen, and fluorinated gas--and their application to clinical pulmonary research. During the past several years there has been accelerated development of pulmonary MRI. This has been driven in part by concerns regarding ionizing radiation using multidetector computed tomography (CT). However, MRI also offers capabilities for fast multispectral and functional imaging using gas agents that are not technically feasible with CT. Recent improvements in gradient performance and radial acquisition methods using ultrashort echo time (UTE) have contributed to advances in these functional pulmonary MRI techniques. The relative strengths and weaknesses of the main functional imaging methods and gas agents are compared and applications to measures of ventilation, diffusion, and gas exchange are presented. Functional lung MRI methods using these gas agents are improving our understanding of a wide range of chronic lung diseases, including chronic obstructive pulmonary disease, asthma, and cystic fibrosis in both adults and children.

Oxygen-enhanced 3D radial ultrashort echo time magnetic resonance imaging in the healthy human lung.

The purpose of this work was to use 3D radial ultrashort echo time (UTE) MRI to perform whole-lung oxygen-enhanced (OE) imaging in humans. Eight healthy human subjects underwent two 3D radial UTE MRI acquisitions (TE = 0.08 ms): one while breathing 21% O2 and the other while breathing 100% O2. Scans were each performed over 5 min of free breathing, using prospective respiratory gating. For comparison purposes, conventional echo time (TE = 2.1 ms) images were acquired simultaneously during each acquisition using a radial " outward-inward" k-space trajectory. 3D percent OE maps were generated from these images. 3D OE maps showing lung signal enhancement were generated successfully in seven subjects (technical failure in one subject). Mean percent signal enhancement was 6.6% ± 1.8%, near the value predicted by theory of 6.3%. No significant enhancement was seen using the conventional echo time data, confirming the importance of UTE for this acquisition strategy. 3D radial UTE MRI shows promise as a method for OE MRI that enables whole-lung coverage and isotropic spatial resolution, in comparison to existing 2D OE methods, which rely on a less time-efficient inversion recovery pulse sequence. These qualities may help OE MRI become a viable low-cost method for 3D imaging of lung function in human subjects.

Hyperpolarized Helium-3 MRI of exercise-induced bronchoconstriction during challenge and therapy.

PURPOSE: To investigate the utility of hyperpolarized He-3 MRI for detecting regional lung ventilated volume (VV) changes in response to exercise challenge and leukotriene inhibitor montelukast, human subjects with exercise induced bronchoconstriction (EIB) were recruited. This condition is described by airway constriction following exercise leading to reduced forced expiratory volume in 1 second (FEV1) coinciding with ventilation defects on hyperpolarized He-3 MRI. MATERIALS AND METHODS: Thirteen EIB subjects underwent spirometry and He-3 MRI at baseline, postexercise, and postrecovery at multiple visits. On one visit montelukast was given and on two visits placebo was given. Regional VV was calculated in the apical/basilar dimension, in the anterior/posterior dimension, and for the entire lung volume. The whole lung VV was used as an end-point and compared with spirometry. RESULTS: Postchallenge FEV1 dropped with placebo but not with treatment, while postchallenge VV dropped more with placebo than treatment. Sources of variability for VV included region (anterior/posterior), scan, and treatment. VV correlated with FEV1/ forced vital capacity (FVC) and forced expiratory flow between 25 and 75% of FVC and showed gravitational dependence after exercise challenge. CONCLUSION: A paradigm testing the response of ventilation to montelukast revealed both a whole-lung and regional response to exercise challenge and therapy in EIB subjects.

Markers of vascular perturbation correlate with airway structural change in asthma.

RATIONALE: Air trapping and ventilation defects on imaging are characteristics of asthma. Airway wall thickening occurs in asthma and is associated with increased bronchial vascularity and vascular permeability. Vascular endothelial cell products have not been explored as a surrogate to mark structural airway changes in asthma. OBJECTIVES: Determine whether reporters of vascular endothelial cell perturbation correlate with airway imaging metrics in patients with asthma of varying severity. METHODS: Plasma from Severe Asthma Research Program subjects was analyzed by ELISAs for soluble von Willebrand factor mature protein (VWF:Ag) and propeptide (VWFpp), P-selectin, and platelet factor 4. Additional subjects were analyzed over 48 hours after whole-lung antigen challenge. We calculated ventilation defect volume by hyperpolarized helium-3 magnetic resonance imaging and areas of low signal density by multidetector computed tomography (less than -856 Hounsfield units [HU] at functional residual capacity and -950 HU at total lung capacity [TLC]). MEASUREMENTS AND MAIN RESULTS: VWFpp and VWFpp/Ag ratio correlated with and predicted greater percentage defect volume on hyperpolarized helium-3 magnetic resonance imaging. P-selectin correlated with and predicted greater area of low density on chest multidetector computed tomography less than -950 HU at TLC. Platelet factor 4 did not correlate. Following whole-lung antigen challenge, variation in VWFpp, VWFpp/Ag, and P-selectin among time-points was less than that among subjects, indicating stability and repeatability of the measurements. CONCLUSIONS: Plasma VWFpp and P-selectin may be useful as surrogates of functional and structural defects that are evident on imaging. The results raise important questions about why VWFpp and P-selectin are associated specifically with different imaging abnormalities.

Exercise-induced bronchoconstriction: reproducibility of hyperpolarized 3He MR imaging.

PURPOSE: To quantitatively evaluate interday, interreader, and intersite agreement of readers of hyperpolarized helium 3 (HPHe) MR images in patients with exercise-induced bronchoconstriction. MATERIALS AND METHODS: This HIPAA-compliant, institutional review board approved study included 13 patients with exercise-induced bronchoconstriction. On two separate days, HPHe MR imaging of the lungs was performed at baseline, immediately after a 10-minute exercise challenge (postchallenge), and 45 minutes after exercise (recovery). Patients were imaged at two sites, six at site A and seven at site B. Images were analyzed independently by multiple readers at each site. Lung volume, ventilation defect volume, ventilated volume, and the number of defects were measured quantitatively, and the location of defects was evaluated qualitatively at site A. Interday and interreader agreement were evaluated by using the intraclass correlation coefficient (ICC), and intersite agreement was evaluated by using a modified Bland-Altman analysis. RESULTS: The ICC between days for ventilation defect volume, ventilated volume, and number of defects was at least 0.74 at both sites. The ICC for lung volume was greater at site B (0.83-0.86) than at site A (0.60-0.65). Defects seen in the same location in the lung on both days included 19.7% of those seen on baseline images and 29.2% and 18.6% of defects on postchallenge and recovery images, respectively. Interreader ICC for each measurement was at least 0.82 for each site. Analysis of intersite agreement showed biases of 612 mL for lung volume, -60.7 mL for ventilation defect volume, 2.91% for ventilated volume, and -6.56 for number of defects. CONCLUSION: The reported measures of reproducibility of HPHe MR imaging may help in the design and interpretation of single- and multicenter studies of patients with exercise-induced bronchoconstriction.