Accelerated noncontrast-enhanced pulmonary vein MRA with distributed compressed sensing

Purpose:

To investigate the efficacy of distributed compressed sensing (CS) to accelerate free‐breathing, electrocardiogram (ECG)‐triggered noncontrast pulmonary vein (PV) magnetic resonance angiography (MRA).

Materials and Methods:

Fully sampled ECG‐triggered noncontrast PV MRA, using a spatially selective slab inversion preparation sequence, was acquired on seven healthy adult subjects (27 ± 17 years, range: 19–65 years, 4 women). The k‐space data were retrospectively randomly undersampled by factors of 2, 4, 6, 8, and 10 and then reconstructed using distributed CS and coil‐by‐coil CS methods. The reconstructed images were evaluated by two blinded readers in consensus for assessment of major PV branches as well as the presence of artifacts in left atrium (LA) and elsewhere. Diameters of right inferior and right superior PV branches were measured. Additionally, mean square errors (MSE) of the reconstructions were calculated.

Results:

Both CS methods resulted in image quality scores similar to the fully sampled reference images at undersampling factors up to 6‐fold for distributed CS and 4‐fold for coil‐by‐coil CS reconstructions. There was no difference in the presence of artifacts in LA and freedom from important artifacts elsewhere between the two techniques up to undersampling factors of 10 compared to the fully sampled reconstruction. For the PV diameters, no systematic variation between the reference and the reconstructions were observed for either technique. There were no significant differences in MSE between the two methods when compared at a given rate, but the difference was significant when compared across all rates.

Conclusion:

The sparsity of noncontrast PV MRA and the joint sparsity of different coil images allow imaging at high undersampling factors (up to 6‐fold) when distributed CS is used. J. Magn. Reson. Imaging 2011;33:1248–1255. © 2011 Wiley‐Liss, Inc.     

Model-based nonlinear inverse reconstruction for T2 mapping using highly undersampled spin-echo MRI

Purpose:

To develop a model‐based reconstruction technique for T2 mapping based on multi‐echo spin‐echo MRI sequences with highly undersampled Cartesian data encoding.

Materials and Methods:

The proposed technique relies on a nonlinear inverse reconstruction algorithm which directly estimates a T2 and spin‐density map from a train of undersampled spin echoes. The method is applicable to acquisitions with single receiver coils but benefits from multi‐element coil arrays. The algorithm is validated for trains of 16 spin echoes with a spacing of 10 to 12 ms using numerical simulations as well as human brain MRI at 3 Tesla (T).

Results:

When compared with a standard T2 fitting procedure using fully sampled T2‐weighted images, and depending on the available signal‐to‐noise ratio and number of coil elements, model‐based nonlinear inverse reconstructions for both simulated and in vivo MRI data yield accurate T2 estimates for undersampling factors of 5 to 10.

Conclusion:

This work describes a promising strategy for T2‐weighted MRI that simultaneously offers accurate T2 relaxation times and properly T2‐weighted images at arbitrary echo times. For a standard spin‐echo MRI sequence with Cartesian encoding, the method allows for a much higher degree of undersampling than obtainable by conventional parallel imaging. J. Magn. Reson. Imaging 2011;. © 2011 Wiley‐Liss, Inc.     

Compressed‐Sensing multispectral imaging of the postoperative spine

Purpose:

To apply compressed sensing (CS) to in vivo multispectral imaging (MSI), which uses additional encoding to avoid magnetic resonance imaging (MRI) artifacts near metal, and demonstrate the feasibility of CS‐MSI in postoperative spinal imaging.

Materials and Methods:

Thirteen subjects referred for spinal MRI were examined using T2‐weighted MSI. A CS undersampling factor was first determined using a structural similarity index as a metric for image quality. Next, these fully sampled datasets were retrospectively undersampled using a variable‐density random sampling scheme and reconstructed using an iterative soft‐thresholding method. The fully and undersampled images were compared using a 5‐point scale. Prospectively undersampled CS‐MSI data were also acquired from two subjects to ensure that the prospective random sampling did not affect the image quality.

Results:

A two‐fold outer reduction factor was deemed feasible for the spinal datasets. CS‐MSI images were shown to be equivalent or better than the original MSI images in all categories: nerve visualization: P = 0.00018; image artifact: P = 0.00031; image quality: P = 0.0030. No alteration of image quality and T2 contrast was observed from prospectively undersampled CS‐MSI.

Conclusion:

This study shows that the inherently sparse nature of MSI data allows modest undersampling followed by CS reconstruction with no loss of diagnostic quality. J. Magn. Reson. Imaging 2013;37:243–248. © 2012 Wiley Periodicals, Inc.     

Accelerating three‐dimensional molecular cardiovascular MR imaging using compressed sensing

Purpose:

To accelerate the acquisition of three‐dimensional (3D) high‐resolution cardiovascular molecular MRI by using Compressed Sensing (CS) reconstruction.

Materials and Methods:

Molecular MRI is an emerging technique for the early assessment of cardiovascular disease. This technique provides excellent soft tissue differentiation at a molecular and cellular level using target‐specific contrast agents (CAs). However, long scan times are required for 3D molecular MRI. Parallel imaging can be used to speed‐up these acquisitions, but hardware considerations limit the maximum acceleration factor. This limitation is important in small‐animal studies, where single‐coils are commonly used. Here we exploit the sparse nature of molecular MR images, which are characterized by localized and high‐contrast biological target‐enhancement, to accelerate data acquisition. CS was applied to detect: (a) venous thromboembolism and (b) coronary injury and aortic vessel wall in single‐ and multiple‐coils acquisitions, respectively.

Results:

Retrospective undersampling showed good overall image quality with accelerations up to four for thrombus and aortic images, and up to three for coronary artery images. For higher acceleration factors, features with high CA uptake were still well recovered while low affinity targets were less preserved with increased CS undersampling artifacts. Prospective undersampling was performed in an aortic image with acceleration of two, showing good contrast and well‐defined tissue boundaries in the contrast‐enhanced regions.

Conclusion:

We demonstrate the successful application of CS to preclinical molecular MR with target specific gadolinium‐based CAs using retrospective (accelerations up to four) and prospective (acceleration of two) undersampling. J. Magn. Reson. Imaging 2012; 36:1362–1371. © 2012 Wiley Periodicals, Inc.     

Accelerating cine-MR imaging in mouse hearts using compressed sensing

Purpose:

To combine global cardiac function imaging with compressed sensing (CS) in order to reduce scan time and to validate this technique in normal mouse hearts and in a murine model of chronic myocardial infarction.

Materials and Methods:

To determine the maximally achievable acceleration factor, fully acquired cine data, obtained in sham and chronically infarcted (MI) mouse hearts were 2–4‐fold undersampled retrospectively, followed by CS reconstruction and blinded image segmentation. Subsequently, dedicated CS sampling schemes were implemented at a preclinical 9.4 T magnetic resonance imaging (MRI) system, and 2‐ and 3‐fold undersampled cine data were acquired in normal mouse hearts with high temporal and spatial resolution.

Results:

The retrospective analysis demonstrated that an undersampling factor of three is feasible without impairing accuracy of cardiac functional parameters. Dedicated CS sampling schemes applied prospectively to normal mouse hearts yielded comparable left‐ventricular functional parameters, and intra‐ and interobserver variability between fully and 3‐fold undersampled data.

Conclusion:

This study introduces and validates an alternative means to speed up experimental cine‐MRI without the need for expensive hardware. J. Magn. Reson. Imaging 2011. © 2011 Wiley Periodicals, Inc.     

Computational simulations and experimental studies of 3D phase‐contrast imaging of fluid flow in carotid bifurcation geometries

Purpose:

To evaluate the use of computational fluid dynamic (CFD)‐based magnetic resonance imaging (MRI) simulations to predict the image appearance and velocity measurement of fluid flow in human carotid bifurcation geometries, and to compare the results with images from experimental MRI studies.

Materials and Methods:

Simulated particle paths were calculated from available CFD datasets of normal and moderately stenosed carotid bifurcation geometries. An MRI simulator based on the spin isochromat method was used to generate images corresponding to a 3D phase‐contrast sequence with velocity encoding in three orthogonal directions. The resulting images were compared qualitatively with experimental MRI scans of the corresponding physical models.

Results:

The simulations predicted the main features observed in experimental studies, such as the low image intensity in regions of complex flow and the position and bright appearance of the jet in the stenosed bifurcation. Simulated velocity images also agreed well with experimental results. The effects of sequence parameters such as repetition time (TR) and echo time (TE) were readily demonstrated by the simulations.

Conclusion:

CFD‐based MRI simulations can be used to predict the appearance of MRI images of regions of physiological flow, and may be useful in the development of improved pulse sequences for flow measurement. J. Magn. Reson. Imaging 2010;31:928–934. ©2010 Wiley‐Liss, Inc.     

Diffusion‐weighted imaging of human carotid artery using 2D single‐shot interleaved multislice inner volume diffusion‐weighted echo planar imaging (2D ss‐IMIV‐DWEPI) at 3T: Diffusion measurement in atherosclerotic plaque

Purpose:

To determine if 2D single‐shot interleaved multislice inner volume diffusion‐weighted echo planar imaging (ss‐IMIV‐DWEPI) can be used to obtain quantitative diffusion measurements that can assist in the identification of plaque components in the cervical carotid artery.

Materials and Methods:

The 2D ss‐DWEPI sequence was combined with interleaved multislice inner volume region localization to obtain diffusion weighted images with 1 mm in‐plane resolution and 2 mm slice thickness. Eleven subjects, six of whom have carotid plaque, were studied with this technique. The apparent diffusion coefficient (ADC) images were calculated using DW images with b = 10 s/mm² and b = 300 s/mm².

Results:

The mean ADC measurement in normal vessel wall of the 11 subjects was 1.28 ± 0.09 × 10^?3 mm²/s. Six of the 11 subjects had carotid plaque and ADC measurements in plaque ranged from 0.29 to 0.87 × 10^?3 mm²/s. Of the 11 common carotid artery walls studied (33 images), at least partial visualization of the wall was obtained in all ADC images, more than 50% visualization in 82% (27/33 images), and full visualization in 18% (6/33 images).

Conclusion:

2D ss‐IMIV‐DWEPI can perform diffusion‐weighted carotid magnetic resonance imaging (MRI) in vivo with reasonably high spatial resolution (1 × 1 × 2 mm³). ADC values of the carotid wall and plaque are consistent with similar values obtained from ex vivo endarterectomy specimens. The spread in ADC values obtained from plaque indicate that this technique could form a basis for plaque component identification in conjunction with other MRI/MRA techniques. J. Magn. Reson. Imaging 2009;30:1068–1077. © 2009 Wiley‐Liss, Inc.

     

Automated assessment of whole‐body adipose tissue depots from continuously moving bed MRI: A feasibility study

Purpose

To present an automated algorithm for segmentation of visceral, subcutaneous, and total volumes of adipose tissue depots (VAT, SAT, TAT) from whole‐body MRI data sets and to investigate the VAT segmentation accuracy and the reproducibility of all depot assessments.

Materials and Methods

Repeated measurements were performed on 24 volunteer subjects using a 1.5 Tesla clinical MRI scanner and a three‐dimensional (3D) multi‐gradient‐echo sequence (resolution: 2.1 × 2.1 × 8 mm³, acquisition time: 5 min 15 s). Fat and water images were reconstructed, and fully automated segmentation was performed. Manual segmentation of the VAT reference was performed by an experienced operator.

Results

Strong correlation (R = 0.999) was found between the automated and manual VAT assessments. The automated results underestimated VAT with 4.7 ± 4.4%. The accuracy was 88 ± 4.5% and 7.6 ± 5.7% for true positive and false positive fractions, respectively. Coefficients of variation from the repeated measurements were: 2.32 % ± 2.61%, 2.25% ± 2.10%, and 1.01% ± 0.74% for VAT, SAT, and TAT, respectively.

Conclusion

Automated and manual VAT results correlated strongly. The assessments of all depots were highly reproducible. The acquisition and postprocessing techniques presented are likely useful in obesity related studies. J. Magn. Reson. Imaging 2009;30:185–193. © 2009 Wiley‐Liss, Inc.     

Breathheld autocalibrated phase‐contrast imaging

Purpose:

To compare generalized autocalibrating partially parallel acquisitions (GRAPPA), modified sensitivity encoding (mSENSE), and SENSE in phase‐contrast magnetic resonance imaging (PC‐MRI) applications.

Materials and Methods:

Aliasing of the torso can occur in PC‐MRI applications. If the data are further undersampled for parallel imaging, SENSE can be problematic in correctly unaliasing signals due to coil sensitivity maps that do not match that of the aliased volume. Here, a method for estimating coil sensitivities in flow applications is described. Normal volunteers (n = 5) were scanned on a 1.5 T MRI scanner and underwent PC‐MRI scans using GRAPPA, mSENSE, SENSE, and conventional PC‐MRI acquisitions. Peak velocity and flow through the aorta and pulmonary artery were evaluated.

Results:

Bland–Altman statistics for flow in the aorta and pulmonary artery acquired with mSENSE and GRAPPA methods (R = 2 and R = 3 cases) have comparable mean differences to flow acquired with conventional PC‐MRI. GRAPPA and mSENSE PC‐MRI have more robust measurements than SENSE when there is aliasing artifact caused by insufficient coil sensitivity maps. For peak velocity, there are no considerable differences among the mSENSE, GRAPPA, and SENSE reconstructions and are comparable to conventional PC‐MRI.

Conclusion:

Flow measurements of images reconstructed with autocalibration techniques have comparable agreement with conventional PC‐MRI and provide robust measurements in the presence of wraparound. J. Magn. Reson. Imaging 2010;31:1004–1014. ©2010 Wiley‐Liss, Inc.     

Efficient flow suppressed MRI improves interscan reproducibility of carotid atherosclerosis plaque burden measurements

Purpose:

To determine if better flow suppression can meaningfully improve the reproducibility of measurements associated with carotid atherosclerotic disease, particularly for lumen and wall areas.

Materials and Methods:

Eighteen subjects with carotid artery stenosis identified by duplex ultrasound (11 with 16%–49% stenosis; 7 with 50%–79% stenosis) underwent two carotid magnetic resonance imaging (MRI) examinations on a 3T scanner with a 4‐channel phased array coil. High‐resolution intermediate‐weighted TSE (TR/TE = 4000/8.5 msec, 0.55 mm in‐plane resolution, 2 mm slice thickness, 16 slices, 3‐minute scan time) with two different flow‐suppression techniques (multislice double inversion recovery [mDIR] and motion‐sensitized driven‐equilibrium [MSDE]) were obtained separately. For each subject, bilateral arteries were reviewed. One radiologist blinded to timepoints, flow suppression techniques, and clinical information measured the arterial lumen area, wall area, and total vessel wall area.

Results:

Compared to mDIR, the MSDE technique had a smaller interscan standard deviation (SD) in lumen (SD: 3.6 vs. 5.2 mm², P = 0.02), wall area measurements (SD: 4.5 vs. 6.4 mm², P = 0.02), and a trend towards smaller SD in total vessel area measurement (SD: 4.4 vs. 4.9 mm², P = 0.07).

Conclusion:

The results from this study demonstrate that vessel wall imaging could quantify atherosclerotic plaque measurements more reliably with an improved blood suppression technique. This relationship between flow‐suppression efficiency and reproducibility of plaque measurements is important, as more reliable area measurements will be useful in clinical diagnosis and in serial MRI studies that monitor carotid atherosclerotic lesion progression and regression. J. Magn. Reson. Imaging 2010;32:452–458. © 2010 Wiley‐Liss, Inc.     

Scan‐rescan reproducibility of carotid atherosclerotic plaque morphology and tissue composition measurements using multicontrast MRI at 3T

Purpose:

To evaluate interscan reproducibility of both vessel morphology and tissue composition measurements of carotid atherosclerosis using a fast, optimized, 3T multicontrast protocol.

Materials and Methods:

A total of 20 patients with carotid stenosis >15% identified by duplex ultrasound were recruited for two independent 3T MRI (Philips) scans within one month. A multicontrast protocol including five MR sequences was applied: TOF, T1‐/T2‐/PD‐weighted and magnetization‐prepared rapid acquisition gradient‐echo (MP‐RAGE). Carotid artery morphology (wall volume, lumen volume, total vessel volume, normalized wall index, and mean/maximum wall thickness) and plaque component size (lipid rich/necrotic core, calcification, and hemorrhage) were measured over two time points.

Results:

After exclusion of images with poor image quality, 257 matched locations from 18 subjects were available for analysis. For the quantitative carotid morphology measurements, coefficient of variation (CV) ranged from 2% to 15% and intraclass correlation coefficient (ICC) ranged from 0.87 to 0.99. Except for maximum wall thickness (ICC = 0.87), all ICC were larger than 0.90. For the quantitative plaque composition measurements, the ICC of the volume and relative content of lipid rich/necrotic core and calcification were larger than 0.90 with CV ranging from 22% to 32%.

Conclusion:

The results from the multicontrast high‐resolution 3T MR study show high reliability for carotid morphology and plaque component measurements. 3T MRI is a reliable tool for longitudinal clinical trials, with shorter scan time compared to 1.5T. J. Magn. Reson. Imaging 2010;31:168–176. © 2009 Wiley‐Liss, Inc.     

An analysis algorithm for accurate determination of articular cartilage thickness of hip joint from MR images

Purpose:

To test the accuracy of the most widely used technique based on edge detection for thickness measurement of the hip joint cartilage in MR images, and to improve the measurement accuracy by developing a new measurement method based on a model of the MRI process.

Materials and Methods:

MRI was performed in 3 normal cadaver hips, 25 hips of normal volunteers, and 25 hips of patients with osteoarthritis. In general, thickness was defined as the distance between the two sides of the hip cartilage along the normal directions of the cartilage surfaces. In this article this is referred to as the “edge detection method.” A theoretical simulation analysis revealed that the edge detection method considerably underestimated the cartilage thickness of the hip joint. A new measurement method based on a model of the MR imaging process was accordingly proposed for correcting the measurement errors.

Results:

In the experiment using the cadaver hips, anatomical measurement of cartilage thickness was used as reference standard. For measurements at 35 sites, the proposed model‐based method gave results similar to those presented from anatomic section, while the edge detection method gave underestimation compared with the anatomic thickness. The underestimation biases for the edge detection method were consistent with the biases predicted by theoretical simulation. In the experiment using the hips of volunteers and patients, the edge detection result was an underestimation compared with the result generated by using the model‐based method.

Conclusion:

The edge detection method underestimated the hip cartilage thickness in MR images. The proposed model‐based method was more accurate than the edge detection method for thickness measurement of the hip cartilage. J. Magn. Reson. Imaging 2011;. © 2011 Wiley‐Liss, Inc.     

Model‐based Acceleration of Parameter mapping (MAP) for saturation prepared radially acquired data

A reconstruction technique called Model‐based Acceleration of Parameter mapping (MAP) is presented allowing for quantification of longitudinal relaxation time and proton density from radial single‐shot measurements after saturation recovery magnetization preparation. Using a mono‐exponential model in image space, an iterative fitting algorithm is used to reconstruct one well resolved and consistent image for each of the projections acquired during the saturation recovery relaxation process. The functionality of the algorithm is examined in numerical simulations, phantom experiments, and in‐vivo studies. MAP reconstructions of single‐shot acquisitions feature the same image quality and resolution as fully sampled reference images in phantom and in‐vivo studies. The longitudinal relaxation times obtained from the MAP reconstructions are in very good agreement with the reference values in numerical simulations as well as phantom and in‐vivo measurements. Compared to available contrast manipulation techniques, no averaging of projections acquired at different time points of the relaxation process is required in MAP imaging. The proposed technique offers new ways of extracting quantitative information from single‐shot measurements acquired after magnetization preparation. The reconstruction simultaneously yields images with high spatiotemporal resolution fully consistent with the acquired data as well as maps of the effective longitudinal relaxation parameter and the relative proton density. Magn Reson Med 70:1524–1534, 2013. © 2013 Wiley Periodicals, Inc.     

MRI measurements of carotid plaque in the atherosclerosis risk in communities (ARIC) study: Methods,reliability and descriptive statistics

Purpose:

To measure carotid plaque components using MRI and estimate reliability in the population‐based Atherosclerosis Risk in Communities (ARIC) study.

Materials and Methods:

Contrast‐enhanced high‐resolution (0.51 × 0.58 × 2 mm³) MRI images were acquired through internal (ICA) and common carotid arteries (CCA) of 2066 ARIC participants at four sites. Sixty‐one exams were repeated and 164 pairs had repeated interpretations. Plaque component thicknesses, areas and volumes over eight slices (1.6‐cm segment) were measured. Intraplaque hemorrhage was recorded. Reliability was evaluated by intraclass correlations and κ statistics.

Results:

There were 1769 successful MRI exams (mean age 71 years; 57% females; 81% white; 19% African‐Americans). Repeat scan reliability was highest for CCA lumen area (0.94) and maximum wall thickness (0.89), ICA lumen area (0.89) and maximum wall thickness (0.77) and total wall volume (0.79), and lowest for small structures—core volume (0.30) and mean cap thickness (0.38). Overall reliability was primarily related to reader variability rather than scan acquisition. K's for presence of core, calcification and hemorrhage were fair to good. White men had the thickest plaques (average maximum ICA wall thickness = 2.3 mm) and the most cores (34%).

Conclusion:

The most important limiting factor for MRI measurements of plaque components is reader variability. Measurement error depends largely on the analyzed structure's size. J. Magn. Reson. Imaging 2010; 31: 406–415. © 2010 Wiley‐Liss, Inc.     

Magnetic resonance imaging and model prediction for thermal ablation of tissue

Purpose

To monitor and predict tissue temperature distributions and lesion boundaries during thermal ablation by combining MRI and thermal modeling methods.

Materials and Methods

Radiofrequency (RF) ablation was conducted in the paraspinal muscles of rabbits with MRI monitoring. A gradient‐recalled echo (GRE) sequence via a 1.5T MRI system provided tissue temperature distribution from the phase images and lesion progression from changes in magnitude images. Post‐ablation GRE estimates of lesion size were compared with post‐ablation T2‐weighted turbo‐spin‐echo (TSE) images and hematoxylin and eosin (H&E)‐stained histological slices. A three‐dimensional (3D) thermal model was used to simulate and predict tissue temperature and lesion size dynamics.

Results

The lesion area estimated from repeated GRE images remained constant during the post‐heating period when the temperature of the lesion boundary was less than a critical temperature. The final lesion areas estimated from multi‐slice (M/S) GRE, TSE, and histological slices were not statistically different. The model‐simulated tissue temperature distribution and lesion area closely corresponded to the GRE‐based MR measurements throughout the imaging experiment.

Conclusion

For normal tissue in vivo, the dynamics of tissue temperature distribution and lesion size during RF thermal ablation can be 1) monitored with GRE phase and magnitude images, and 2) simulated for prediction with a thermal model. J. Magn. Reson. Imaging 2007;26:123–132. © 2007 Wiley‐Liss, Inc.     

Effect of flip angle on volume flow measurement with nontriggered phase‐contrast MR: In vivo evaluation in carotid and basilar arteries

Purpose

To evaluate the effect of flip angle on volume flow rate measurements obtained with nontriggered phase‐contrast magnetic resonance imaging (MRI) in vivo.

Materials and Methods

We prospectively measured volume flow rates of the bilateral internal carotid artery and the basilar artery with cine and nontriggered phase‐contrast MRI. For nontriggered phase‐contrast imaging, flip angles of 4, 15, 60, and 90° were used for 40 volunteers and of 8, 15, and 30° for 54 volunteers. Lumen boundaries were semiautomatically determined by pulsatility‐based segmentation using cine phase‐contrast MRI. Identical lumen boundaries were used for nontriggered phase‐contrast imaging.

Results

The ratio of volume flow rate obtained with nontriggered phase‐contrast imaging to that obtained with cine phase‐contrast imaging significantly increases with an increase in the flip angle. The mean ratios lie within a relatively narrow range of ±15% with a wide range of flip angles of 8–90°. As the flip angle increases, ghost artifacts become prominent and signal‐to‐noise and contrast‐to‐noise ratios increase.

Conclusion

Flip angles between 8 and 60° are most appropriate for nontriggered phase‐contrast MR measurements in the internal carotid and the basilar artery. J. Magn. Reson. Imaging 2009;29:1218–1223. © 2009 Wiley‐Liss, Inc.     

Automated quantification of muscle and fat in the thigh from water-, fat-, and nonsuppressed MR images

Purpose:

To introduce and validate an unsupervised muscle and fat quantification algorithm based on joint analysis of water‐suppressed (WS), fat‐suppressed (FS), and water and fat (nonsuppressed) volumetric magnetic resonance imaging (MRI) of the mid‐thigh region.

Materials and Methods:

We first segmented the subcutaneous fat by use of a parametric deformable model, then applied centroid clustering in the feature domain defined by the voxel intensities in WS and FS images to identify the intermuscular fat and muscle. In the final step we computed volumetric and area measures of fat and muscle. We applied this algorithm on datasets of water‐, fat‐, and nonsuppressed volumetric MR images acquired from 28 participants.

Results:

We validated our tissue composition analysis against fat and muscle area measurements obtained from semimanual analysis of single‐slice mid‐thigh computed tomography (CT) images of the same participants and found very good agreement between the two methods. Furthermore, we compared the proposed approach with a variant that uses nonsuppressed images only and observed that joint analysis of WS and FS images is more accurate than the nonsuppressed only variant.

Conclusion:

Our MRI algorithm produces accurate tissue quantification, is less labor‐intensive, and more reproducible than the original CT‐based workflow and can address interparticipant anatomic variability and intensity inhomogeneity effects. J. Magn. Reson. Imaging 2012;35:1152‐1161. © 2011 Wiley Periodicals, Inc.     

Reconstruction of 3D dynamic contrast-enhanced magnetic resonance imaging using nonlocal means

Purpose

To develop and test a nonlocal means‐based reconstruction algorithm for undersampled 3D dynamic contrast‐enhanced (DCE) magnetic resonance imaging (MRI) of tumors.

Materials and Methods

We propose a reconstruction technique that is based on the recently proposed nonlocal means (NLM) filter which can relax trade‐offs in spatial and temporal resolutions in dynamic imaging. Unlike the original application of NLM for image denoising, the MR reconstruction framework here can offer high‐quality images from undersampled k‐space data. The method is based on enforcing similarity constraints in terms of neighborhoods of pixels rather than individual pixels. The method was applied on undersampled 3D DCE imaging of breast and brain tumor datasets and the results were compared to sliding window reconstructions and to a compressed sensing method using total variation constraints on the images.

Results

Undersampling factors of up to five were obtained with the proposed approach while preserving the spatial and temporal characteristics. The NLM reconstruction method offered improved performance over the sliding window and the total variation constrained reconstruction techniques.

Conclusion

The reconstruction framework here can give high‐quality images from undersampled DCE MRI data and has the potential to improve the quality of DCE tumor imaging. J. Magn. Reson. Imaging 2010;32:1217–1227. © 2010 Wiley‐Liss, Inc.     

Machine learning study of several classifiers trained with texture analysis features to differentiate benign from malignant soft‐tissue tumors in T1‐MRI images

Purpose:

To study, from a machine learning perspective, the performance of several machine learning classifiers that use texture analysis features extracted from soft‐tissue tumors in nonenhanced T1‐MRI images to discriminate between malignant and benign tumors.

Materials and Methods:

Texture analysis features were extracted from the tumor regions from T1‐MRI images of clinically proven cases of 49 malignant and 86 benign soft‐tissue tumors. Three conventional machine learning classifiers were trained and tested. The best classifier was compared to the radiologists by means of the McNemar's statistical test.

Results:

The SVM classifier performs better than the neural network and the C4.5 decision tree based on the analysis of their receiver operating curves (ROC) and cost curves. The classification accuracy of the SVM, which was 93% (91% specificity; 94% sensitivity), was better than the radiologist classification accuracy of 90% (92% specificity; 81% sensitivity).

Conclusion:

Machine learning classifiers trained with texture analysis features are potentially valuable for detecting malignant tumors in T1‐MRI images. Analysis of the learning curves of the classifiers showed that a training data size smaller than 100 T1‐MRI images is sufficient to train a machine learning classifier that performs as well as expert radiologists. J. Magn. Reson. Imaging 2010;31:680–689. © 2010 Wiley‐Liss, Inc.     

Magnetic resonance imaging in real time: Advances using radial FLASH

Purpose

To develop technical advances for real‐time magnetic resonance imaging (MRI) that allow for improved image quality and high frame rates.

Materials and Methods

The approach is based on a combination of fast low‐angle shot (FLASH) MRI sequences with radial data sampling and view sharing of successive acquisitions. Gridding reconstructions provide images free from streaking or motion artifacts and with a flexible trade‐off between spatial and temporal resolution. Immediate image reconstruction and online display is accomplished with the use of an unmodified 3 T MRI system. For receive coils with a large number of elements this process is supported by a user‐selectable channel compression that is based on a principal component analysis and performed during initial preparation scans.

Results

In preliminary applications to healthy volunteers, real‐time radial FLASH MRI visualized continuous movements of the temporomandibular joint during voluntary opening and closing of the mouth at high spatial resolution (0.75 mm in‐plane) and monitored cardiac functions at high temporal resolution (20 images per second) during free breathing and without synchronization to the electrocardiogram.

Conclusion

Real‐time radial FLASH MRI emerges as a simple and versatile tool for a large range of clinical applications. J. Magn. Reson. Imaging 2010. © 2009 Wiley‐Liss, Inc