Muscle shape consistency and muscle volume prediction of thigh muscles

The present study investigated the applicability of a muscle volume prediction method using only the muscle length (L_M), the maximum anatomical cross‐sectional area (ACSA_max), and a muscle‐specific shape factor (p) on the quadriceps vastii. L_M, ACSA_max, muscle volume, and p were obtained from magnetic resonance images of the vastus intermedius (VI), lateralis (VL), and medialis (VM) of female (n = 20) and male (n = 17) volleyball athletes. The average p was used to predict muscle volumes (V_p) using the equation V_p = p × ACSA_max × L_M. Although there were significant differences in the muscle dimensions between male and female athletes, p was similar and on average 0.582, 0.658, 0.543 for the VI, VL, and VM, respectively. The position of ACSA_max showed low variability and was at 57%, 60%, and 81% of the thigh length for VI, VL, and VM. Further, there were no significant differences between measured and predicted muscle volumes with root mean square differences of 5–8%. These results suggest that the muscle shape of the quadriceps vastii is independent of muscle dimensions or sex and that the prediction method could be sensitive enough to detect changes in muscle volume related to degeneration, atrophy, or hypertrophy.     

A gradient-based method for segmenting FDG-PET images: methodology and validation

Purpose A new gradient-based method for segmenting FDG-PET images is described and validated. Methods The proposed method relies on the watershed transform and hierarchical cluster analysis. To allow a better estimation of the gradient intensity, iteratively reconstructed images were first denoised and deblurred with an edge-preserving filter and a constrained iterative deconvolution algorithm. Validation was first performed on computer-generated 3D phantoms containing spheres, then on a real cylindrical Lucite phantom containing spheres of different volumes ranging from 2.1 to 92.9 ml. Moreover, laryngeal tumours from seven patients were segmented on PET images acquired before laryngectomy by the gradient-based method and the thresholding method based on the source-to-background ratio developed by Daisne (Radiother Oncol 2003;69:247–50). For the spheres, the calculated volumes and radii were compared with the known values; for laryngeal tumours, the volumes were compared with the macroscopic specimens. Volume mismatches were also analysed. Results On computer-generated phantoms, the deconvolution algorithm decreased the mis-estimate of volumes and radii. For the Lucite phantom, the gradient-based method led to a slight underestimation of sphere volumes (by 10–20%), corresponding to negligible radius differences (0.5–1.1 mm); for laryngeal tumours, the segmented volumes by the gradient-based method agreed with those delineated on the macroscopic specimens, whereas the threshold-based method overestimated the true volume by 68% (p = 0.014). Lastly, macroscopic laryngeal specimens were totally encompassed by neither the threshold-based nor the gradient-based volumes. Conclusion The gradient-based segmentation method applied on denoised and deblurred images proved to be more accurate than the source-to-background ratio method. The first two authors (Xavier Geets and John A. Lee) have equally contributed to this paper.     

Evaluation of adipose tissue volume quantification with IDEAL fat-water separation

Purpose:

To validate i terative d ecomposition of water and fat with e cho a symmetry and l east‐squares estimation (IDEAL) for adipose tissue volume quantification. IDEAL allows MRI images to be produced only from adipose‐containing tissues; hence, quantifying adipose tissue should be simpler and more accurate than with current methods.

Materials and Methods:

Ten healthy controls were imaged with 1.5 Tesla (T) Spin Echo (SE), 3.0T T1‐weighted spoiled gradient echo (SPGR), and 3.0T IDEAL‐SPGR. Images were acquired from the abdomen, pelvis, mid‐thigh, and mid‐calf. Mean subcutaneous and visceral adipose tissue volumes were compared between the three acquisitions for each subject.

Results:

There were no significant differences (P > 0.05) between the three acquisitions for subcutaneous adipose tissue volumes. However, there was a significant difference (P = 0.0002) for visceral adipose tissue volumes in the abdomen. Post hoc analysis showed significantly lower visceral adipose tissue volumes measured by IDEAL versus 1.5T (P < 0.0001) and 3.0T SPGR (P < 0.002). The lower volumes given by IDEAL are due to its ability to differentiate true visceral adipose tissue from other bright structures like blood vessels and bowel content that are mistaken for adipose tissue in non‐fat suppressed images.

Conclusion:

IDEAL measurements of adipose tissue are equivalent to established 1.5T measurement techniques for subcutaneous depots and have improved accuracy for visceral depots, which are more metabolically relevant. J. Magn. Reson. Imaging 2011;. © 2011 Wiley‐Liss, Inc.     

A simple approach for three‐dimensional mapping of baseline cerebrospinal fluid volume fraction

A simple method of measuring baseline cerebrospinal fluid volume fraction (V_CSF) in three‐dimensional is proposed that used the characteristic of cerebrospinal fluid with very long T₂. It is based on the fitting of monoexponential decay of only cerebrospinal fluid signal, using a nonselective T₂ preparation scheme. Three‐dimensional gradient‐ and spin‐echo acquisition also improves signal‐to‐noise ratio efficiency and brain coverage. Both V_CSF and T_2,CSF are fitted voxel by voxel and analyzed in different cortical areas across subjects. V_CSF is largely regionally dependent (occipital: 8.9 ± 1.7%, temporal: 11.4 ± 2.4%, and frontal: 21.4 ± 6.9%). Measured T_2,CSF was 1573 ± 146 msec within cortical lobes as compared with 2062 ± 37 msec from ventricle regions. Different parameter set were compared, and the robustness of the new method is demonstrated. Conversely, when comparing with the proposed approach, large overestimation of segmentation based method using T₁‐weighted images is found, and the underlying causes are suggested. Magn Reson Med, 2011. © 2010 Wiley‐Liss, Inc.     

Informed RESTORE: A method for robust estimation of diffusion tensor from low redundancy datasets in the presence of physiological noise artifacts

Physiological noise artifacts, especially those originating from cardiac pulsation and subject motion, are common in clinical Diffusion tensor‐MRI acquisitions. Previous works show that signal perturbations produced by artifacts can be severe and neglecting to account for their contribution can result in erroneous diffusion tensor values. The Robust Estimation of Tensors by Outlier Rejection (RESTORE) method has been shown to be an effective strategy for improving tensor estimation on a voxel‐by‐voxel basis in the presence of artifactual data points in diffusion‐weighted images. In this article, we address potential instabilities that may arise when using RESTORE and propose practical constraints to improve its usability. Moreover, we introduce a method, called informed RESTORE designed to remove physiological noise artifacts in datasets acquired with low redundancy (less than 30–40 diffusion‐weighted image volumes)—a condition in which the original RESTORE algorithm may converge to an incorrect solution. This new method is based on the notion that physiological noise is more likely to result in signal dropouts than signal increases. Results from both Monte Carlo simulation and clinical diffusion data indicate that informed RESTORE performs very well in removing physiological noise artifacts for low redundancy diffusion‐weighted image datasets. Magn Reson Med, 2012. © 2012 Wiley Periodicals, Inc.     

Tracking the effects of crusher gradients on gradient‐echo BOLD signal in space and time during rat sensory stimulation

A unique method to map the effect of crusher gradients in space and time on the gradient echo blood oxygen level dependent (BOLD) signal is introduced. Using the Radial Correlation Contrast (RCC) analysis method, amplitude‐RCC maps at different time segments and different gradient strengths were obtained. The ratio of amplitude‐RCC cluster volumes, with and without crusher gradients, showed a temporal dependency with stronger volume reduction for stimulation‐onset versus stimulation‐decline. Aside from signal‐to‐noise ratio reduction in diffusion weighted images, the average temporal patterns were equal. Comparison of the data with and without crushers showed a stronger reduction in local coherence for stimulation‐onset times. We hypothesize that the stimulation decline was weighted by extravascular effects originating in expanded veins due to their larger volume and long range susceptibility which couples neighboring voxels. The ratio of amplitude‐RCC with and without crushers calculated for each voxel at each time segment yielded a spatial–temporal mapping of the crusher effect. These maps suggest that early stimulation‐onset (～9 s) is weighted by flow; later a dynamic steady‐state between intra‐ and extravascular effects is obtained. Stimulation‐decline was dominated by extravascular effects, and at late stimulation decline as well as at early stimulation onset, clusters were small and localized to expected site of neuronal activity. Magn Reson Med 60:548–554, 2008. © 2008 Wiley‐Liss, Inc.     

Impact of outliers on diffusion tensor and Q-ball imaging: clinical implications and correction strategies

Purpose:

To measure the impact of corrupted images often found to occur in diffusion‐weighted magnetic resonance imaging (DW‐MRI). To propose a robust method for the correction of outliers, applicable to diffusion tensor imaging (DTI) and q‐ball imaging (QBI).

Materials and Methods:

Monte Carlo simulations were carried out to measure the impact of outliers on DTI and QBI reconstruction in a single voxel. Methods to correct outliers based on q‐space interpolation and direction removal were then implemented and validated in real image data.

Results:

Corruption in a single voxel led to clear variations in DTI and QBI metrics. In real data, the method of q‐space interpolation was successful in identifying corrupted voxels and restoring them to values consistent with those of uncorrupted images.

Conclusion:

For images containing few gradient directions, where outlier removal was either impossible due to limited volumes or resulted in large changes in DTI/QBI metrics, q‐space interpolation proved to be the method of choice for image restoration. A simple decision support system is proposed to assist clinicians in the correction of their corrupted DW data. J. Magn. Reson. Imaging 2011;33:1491–1502. © 2011 Wiley‐Liss, Inc.     

Regional and temporal variations in tissue sodium concentration during the acute stroke phase

A technique for noninvasively quantifying the concentration of sodium (²³Na) ions was applied to the study of ischemic stroke. ²³Na‐magnetic resonance imaging techniques have shown considerable potential for measuring subtle changes in ischemic tissue, although studies to date have suffered primarily from poor signal/noise ratio. In this study, accurate quantification of tissue sodium concentration (TSC) was achieved in ²³Na images with voxel sizes of 1.2 μL acquired in 10 min. The evolution of TSC was investigated from 0.5 to 8 h in focal cortical and subcortical ischemic tissue following permanent middle cerebral artery occlusion in the rat (n = 5). Infarct volumes determined from TSC measurements correlated significantly with histology (P = 0.0006). A delayed linear model was fitted to the TSC time course data in each voxel, which revealed that the TSC increase was more immediate (0.2 ± 0.1 h delay time) in subcortical ischemic tissue, whereas it was delayed by 1.6 ± 0.5 h in ischemic cortex (P = 0.0002). No significant differences (P = 0.5) were measured between TSC slope rates in cortical (10.2 ± 1.1 mM/h) and subcortical (9.7 ± 1.1 mM/h) ischemic tissue. The data suggest that any TSC increase measured in ischemic tissue indicates infarction (core) and regions exhibiting a delay to TSC increase indicate potentially salvageable tissue (penumbra). Magn Reson Med, 2012. © 2011 Wiley Periodicals, Inc.     

Direct saturation MRI: Theory and application to imaging brain iron

When applying RF saturation to tissue, MRI signal reductions occur due to magnetization transfer (MT) and direct saturation (DS) effects on water protons. It is shown that the direct effects, often considered a nuisance, can be used to distinguish gray matter (GM) regions with different iron content. DS effects were selected by reducing the magnitude and duration of RF irradiation to minimize confounding MT effects. Contrary to MT saturation spectra, direct water saturation spectra are characterized by a symmetric Lorentzian‐shaped frequency dependence that can be described by an exact analytical solution of the Bloch equations. The effect of increased transverse relaxation, e.g., due to the presence of iron, will broaden this saturation spectrum. As a first application, DS ratio (DSR) images were acquired to visualize GM structures in the human brain. Similar to T₂^*‐weighted images, the quality of DSR images was affected by local field inhomogeneity, but this could be easily corrected for by centering the saturation spectrum on a voxel‐by‐voxel basis. The results show that, contrary to commonly used T₂^*‐weighted and absolute R₂ images, the DSR images visualize all GM structures, including cortex. A direct correlation between DSR and iron content was confirmed for these structures. Magn Reson Med, 2009. © 2009 Wiley‐Liss, Inc.     

Magnetic resonance imaging of athlete's heart: myocardial mass, left ventricular function, and cross-sectional area of the coronary arteries

To evaluate left ventricular myocardial mass and function as well as ostial coronary artery cross-sectional area in endurance athletes, an athlete group of 12 highly trained rowers and a control group of 12 sedentary healthy subjects underwent MR examination. An ECG-gated breath-hold cine gradient-echo sequence was used to calculate myocardial mass, end-diastolic and end-systolic volumes, stroke volume, and cardiac output, all related to body surface area, as well as ejection fraction. A 3D fat-saturated ECG- and respiratory-triggered navigator echo sequence was used to evaluate coronary arteries: left main (LM), left anterior descending (LAD), left circumflex (LCx), and right coronary artery (RCA). Cross-sectional area was calculated and divided for body surface area. Myocardial mass was found significantly larger in athlete group than in control group (p = 0.0078), the same being for end-diastolic volume (p = 0.0078), stroke volume (p = 0.0055), LM (p = 0.0066) and LAD (p = 0.0129). No significant difference was found for all the remaining parameters. Significant correlation with myocardial mass was found for LM (p < 0.001) and LAD (p = 0.0340), not for LCx and RCA. Magnetic resonance imaging is a useful tool in evaluating the myocardial hypertrophy and function of athlete's heart. Magnetic resonance angiography is a valuable noninvasive method to visualize the correlated cross-sectional area increase of the left coronary artery system. Received: 25 March 1999; Revised: 31 August 1999; Accepted: 1 September 1999     

Reproducibility of intracranial volume measurement by unsupervised multispectral brain segmentation

To assess the inter-study variability of a recently published unsupervised segmentation method (Magn. Reson. Med. 1997;37:84–93), 14 brain MR studies were performed in five normal subjects. Standard deviations for absolute and fractional volumes of intracranial compartments, which reflect the experimental variability, were smaller than 16.5 ml and 1.1%, respectively. By comparing the experimental component of the variability with the variability observed in our reference database, an estimate of the biological variability of the intracranial fractional volumes in the database population was obtained.     

Comparison of 3 Tesla proton MR spectroscopy, MR perfusion and MR diffusion for distinguishing glioma recurrence from posttreatment effects

Purpose:

To compare 3 Tesla (3T) multi‐voxel and single‐voxel proton MR spectroscopy (MRS), dynamic susceptibility contrast perfusion MRI (DSC), and diffusion‐weighted MRI (DWI) for distinguishing recurrent glioma from postradiation injury.

Materials and Methods:

We reviewed all 3T MRS, DSC and DWI studies performed for suspicion of malignant glioma recurrence between October 2006 and December 2008. Maximum Cho/NAA and Cho/Cr peak‐area and peak‐height ratios were recorded for both multi‐voxel and single‐voxel MRS. Maximum cerebral blood volume (CBV) and minimum apparent diffusion coefficient (ADC) were normalized to white matter. Histopathology and clinical‐radiologic follow‐up served as reference standards. Receiver operating characteristic curves for each parameter were compared.

Results:

Forty lesions were classified as glioma recurrence (n = 30) or posttreatment effect (n = 10). Diagnostic performance was similar for CBV ratio (AUC = 0.917, P < 0.001), multi‐voxel Cho/Cr peak‐area (AUC = 0.913, P = 0.002), and multi‐voxel Cho/NAA peak‐height (AUC = 0.913, P = 0.002), while ADC ratio (AUC = 0.726, P = 0.035) did not appear to perform as well. Single‐voxel MRS parameters did not reliably distinguish tumor recurrence from posttreatment effects.

Conclusion:

A 3T DSC and multi‐voxel MRS Cho/Cr peak‐area and Cho/NAA peak‐height appear to outperform DWI for distinguishing glioma recurrence from posttreatment effects. Single‐voxel MRS parameters do not appear to distinguish glioma recurrence from posttreatment effects reliably, and therefore should not be used in place of multi‐voxel MRS. J. Magn. Reson. Imaging 2012;35:56‐63. © 2011 Wiley Periodicals, Inc.     

A novel MRI-compatible brain ventricle phantom for validation of segmentation and volumetry methods

Purpose:

To create a standardized, MRI‐compatible, life‐sized phantom of the brain ventricles to evaluate ventricle segmentation methods using T₁‐weighted MRI. An objective phantom is needed to test the many different segmentation programs currently used to measure ventricle volumes in patients with Alzheimer's disease.

Materials and Methods:

A ventricle model was constructed from polycarbonate using a digital mesh of the ventricles created from the 3 Tesla (T) MRI of a subject with Alzheimer's disease. The ventricle was placed in a brain mold and surrounded with material composed of 2% agar in water, 0.01% NaCl and 0.0375 mM gadopentetate dimeglumine to match the signal intensity properties of brain tissue in 3T T₁‐weighted MRI. The 3T T₁‐weighted images of the phantom were acquired and ventricle segmentation software was used to measure ventricle volume.

Results:

The images acquired of the phantom successfully replicated in vivo signal intensity differences between the ventricle and surrounding tissue in T₁‐weighted images and were robust to segmentation. The ventricle volume was quantified to 99% accuracy at 1‐mm voxel size.

Conclusion:

The phantom represents a simple, realistic and objective method to test the accuracy of lateral ventricle segmentation methods and we project it can be extended to other anatomical structures. J. Magn. Reson. Imaging 2012;36:476–482. © 2012 Wiley Periodicals, Inc.     

Measurement of cardiac ventricular volumes using multidetector row computed tomography: comparison of two- and three-dimensional methods

This study compared a three-dimensional volumetric threshold-based method to a two-dimensional Simpson’s rule based short-axis multiplanar method for measuring right (RV) and left ventricular (LV) volumes, stroke volumes, and ejection fraction using electrocardiography-gated multidetector computed tomography (MDCT) data sets. End-diastolic volume (EDV) and end-systolic volume (ESV) of RV and LV were measured independently and blindly by two observers from contrast-enhanced MDCT images using commercial software in 18 patients. For RV and LV the three-dimensionally calculated EDV and ESV values were smaller than those provided by two-dimensional short axis (10%, 5%, 15% and 26% differences respectively). Agreement between the two methods was found for LV (EDV/ESV: r=0.974/0.910, ICC=0.905/0.890) but not for RV (r=0.882/0.930, ICC=0.663/0.544). Measurement errors were significant only for EDV of LV using the two-dimensional method. Similar reproducibility was found for LV measurements, but the three-dimensional method provided greater reproducibility for RV measurements than the two-dimensional. The threshold value supported three-dimensional method provides reproducible cardiac ventricular volume measurements, comparable to those obtained using the short-axis Simpson based method.     

Regional ventilation mapping of the rat lung using hyperpolarized (129) Xe magnetic resonance imaging

Lung ventilation was mapped in seven healthy male Sprague‐Dawley rats (433 ± 24 g) using hyperpolarized ¹²⁹Xe magnetic resonance imaging (MRI) at 3.0 T, and validated with hyperpolarized ³He MRI under similar ventilator conditions. Ventilation maps were obtained using flip angle variation for offset of RF and relaxation (FAVOR) which is a multiple breath imaging technique that extracts the fractional ventilation parameter, r, on a pixel‐by‐pixel basis from the dynamic signal enhancement. r is defined as the fractional refreshment of gas per breath. Under the ventilator conditions used in this work, whole‐lung measurements of fractional ventilation obtained using hyperpolarized ¹²⁹Xe were not significantly different from those obtained using hyperpolarized ³He (p = 0.8125 by a Wilcoxon matched pairs test). Fractional ventilation gradients calculated in the superior/inferior (S/I) and anterior/posterior (A/P) directions obtained using hyperpolarized ¹²⁹Xe were not significantly different from those obtained using hyperpolarized ³He (p = 0.9375 and p = 0.1563, for the S/I and A/P directions, respectively). Following baseline fractional ventilation measurements, one representative rat was challenged with methacholine and fractional ventilation measurements were performed over a time course of 10 min. A reduction and subsequent recovery in whole‐lung r values were detected using the FAVOR method. Magn Reson Med, 2012. © 2012 Wiley Periodicals, Inc.     

Increased gray matter volume of left pars opercularis in male orchestral musicians correlate positively with years of musical performance

Purpose

To compare manual volumetry of gray matter (GM) / white matter (WM) of Broca's area subparts: pars opercularis (POP) and pars triangularis (PTR) in both hemispheres between musicians and nonmusician, as it has been shown that these regions are crucial for musical abilities. A previous voxel‐based morphometric (VBM) study conducted in our laboratory reported increased GM density in Broca's area of left hemisphere in male orchestral musicians. Functional segregation of POP/PTR justified separate volumetric analysis of these parts.

Materials and Methods

We used the same cohort for the VBM study. Manual morphometry (stereology) was used to compare volumes between 26/26 right‐handed orchestral musicians/nonmusicians.

Results

As expected, musicians showed significantly increased GM volume in the Broca's area, specifically in the left POP. No significant results were detected in right POP, left/right PTR GM volumes, and WM volumes for all regions. Results were positively correlated with years of musical performance (r = 0.7, P = 0.0001).

Conclusion

This result corroborates the VBM study and is in line with the hypothesis of critical involvement of POP in hearing‐action integration being an integral component of frontoparietotemporal mirror neuron network. We hypothesize that increased size of musicians' left POP represent use‐dependent structural adaptation in response to intensive audiomotor skill acquisition. J. Magn. Reson. Imaging 2011;33:24–32. © 2010 Wiley‐Liss, Inc.     

Multidimensionally encoded magnetic resonance imaging

Magnetic resonance imaging (MRI) typically achieves spatial encoding by measuring the projection of a q‐dimensional object over q‐dimensional spatial bases created by linear spatial encoding magnetic fields (SEMs). Recently, imaging strategies using nonlinear SEMs have demonstrated potential advantages for reconstructing images with higher spatiotemporal resolution and reducing peripheral nerve stimulation. In practice, nonlinear SEMs and linear SEMs can be used jointly to further improve the image reconstruction performance. Here, we propose the multidimensionally encoded (MDE) MRI to map a q‐dimensional object onto a p‐dimensional encoding space where p > q. MDE MRI is a theoretical framework linking imaging strategies using linear and nonlinear SEMs. Using a system of eight surface SEM coils with an eight‐channel radiofrequency coil array, we demonstrate the five‐dimensional MDE MRI for a two‐dimensional object as a further generalization of PatLoc imaging and O‐space imaging. We also present a method of optimizing spatial bases in MDE MRI. Results show that MDE MRI with a higher dimensional encoding space can reconstruct images more efficiently and with a smaller reconstruction error when the k‐space sampling distribution and the number of samples are controlled. Magn Reson Med, 2013. © 2012 Wiley Periodicals, Inc.     

Iterative Probabilistic Voxel Labeling: Automated Segmentation for Analysis of The Cancer Imaging Archive Glioblastoma Images

BACKGROUND AND PURPOSE:Robust, automated segmentation algorithms are required for quantitative analysis of large imaging datasets. We developed an automated method that identifies and labels brain tumor–associated pathology by using an iterative probabilistic voxel labeling using k-nearest neighbor and Gaussian mixture model classification. Our purpose was to develop a segmentation method which could be applied to a variety of imaging from The Cancer Imaging Archive.MATERIALS AND METHODS:Images from 2 sets of 15 randomly selected subjects with glioblastoma from The Cancer Imaging Archive were processed by using the automated algorithm. The algorithm-defined tumor volumes were compared with those segmented by trained operators by using the Dice similarity coefficient.RESULTS:Compared with operator volumes, algorithm-generated segmentations yielded mean Dice similarities of 0.92 ± 0.03 for contrast-enhancing volumes and 0.84 ± 0.09 for FLAIR hyperintensity volumes. These values compared favorably with the means of Dice similarity coefficients between the operator-defined segmentations: 0.92 ± 0.03 for contrast-enhancing volumes and 0.92 ± 0.05 for FLAIR hyperintensity volumes. Robust segmentations can be achieved when only postcontrast T1WI and FLAIR images are available.CONCLUSIONS:Iterative probabilistic voxel labeling defined tumor volumes that were highly consistent with operator-defined volumes. Application of this algorithm could facilitate quantitative assessment of neuroimaging from patients with glioblastoma for both research and clinical indications.

Glioblastoma is the most common primary brain tumor and remains one of the deadliest human cancers.¹ During the past 50 years, improvement with regard to patient outcomes has been marginal.² A major barrier in therapeutic development is attributable to the misconception that glioblastoma constitutes a single disease. Molecular profiling has revealed that glioblastoma comprises multiple subtypes characterized by distinct molecular pathways.³ To improve the clinical outcome of patients with glioblastoma, technologies must be developed to distinguish these subtypes.There are compelling reasons that MR imaging may serve as a tool for dissecting the variability of glioblastoma. First, radiographic data are available for every patient because the clinical management of glioblastoma tumors is largely driven by the interpretation of MR images. Second, available data suggest that the radiographic appearance of glioblastoma is related to its physiologic state.^4,5 To better define this relationship, imaging archives with corresponding genomic profiling, such as The Cancer Imaging Archive (TCIA), have been launched (http://cancerimagingarchive.net/).Much of the early work correlating MR imaging appearances of glioblastoma tumors with genomic profiling was performed by using manually delineated tumor volumes or qualitative assessments provided by trained clinicians.^4,5 These approaches are limited by the inherent variability of subjective interpretation, and significant interrater discrepancies have been reported.^6,7 Additionally, manual segmentation is time-consuming for large datasets. This limitation is particularly apparent when multiple radiographic features require segmentation. To address these deficiencies, effort has been devoted to developing automated algorithms for segmenting tumor volumes.^8–12 These algorithms include clustering,^13,14 discriminative strategies,¹⁵ and generative approaches.^11,16,17 The success of these methods has been limited by widely differing MR imaging protocols for image acquisition and quality¹⁸ and the significant overlap between the radiographic appearance of glioblastoma tumors and normal cerebrum on MR imaging. Although many of these methods can generate high-quality volumes from a training set, segmentation algorithms may fail when applied to images acquired by using different protocols.We hypothesized that a probabilistic approach by using subject-specific classifiers would reliably discriminate glioblastoma from the surrounding cerebrum. In this algorithm, termed iterative probabilistic voxel labeling (IPVL), sparse, high-specificity, preliminary volumes were created for each subject by using a combination of region-growing and K-means-based tissue segmentation. Sampling of these preliminary volumes trained k-nearest neighbor (KNN) and Gaussian mixture model (GMM) classifiers by using voxel intensity and spatial coordinates. Voxel labels are assigned probabilistically by iteratively trained classifiers. Finally, each voxel is labeled as contrast-enhancing tumor volume (CEV), FLAIR hyperintensity volume (FHV), gray matter, white matter, CSF, and blood vessel (BV). Most important, our algorithm reliably segments images from the TCIA that were acquired by a variety of scanners and protocols.     

Susceptibility weighted imaging at ultra high magnetic field strengths: Theoretical considerations and experimental results

We present numerical simulations and experimental results for susceptibility weighted imaging (SWI) at 7 T. Magnitude, phase, and SWI contrast were simulated for different voxel geometries and imaging parameters, resulting in an echo time of 14 msec for optimum contrast between veins and surrounding tissue. Slice thickness of twice the in‐plane voxel size or more resulted in optimum vessel visibility. Phantom and in vivo data are in very good agreement with the simulations and the delineation of vessels at 7 T was superior compared to lower field strengths. The phase of the complex data reveals anatomical details that are complementary to the corresponding magnitude images. Susceptibility weighted imaging at very high field strengths is a promising technique because of its high sensitivity to tissue susceptibility, its low specific absorption rate, and the phase's negligible sensitivity to B₁ inhomogeneities. Magn Reson Med 60:1155–1168, 2008. © Wiley‐Liss, Inc.     

Comparison of dynamic contrast-enhanced MRI with WHO tumor grading for gliomas

Assessment of vascular proliferation as an important grading criterion has been employed in both the histologic and the radiologic characterization of gliomas with encouraging results. Perfusion in gliomas can be measured by dynamic contrast-enhanced magnetic resonance imaging (dMRI). The goal of this study was to develop a model for simultaneously quantifying the fractional volumes of different tissue compartments of gliomas by dMRI. A modified method for evaluating dynamic contrast-enhanced MR images is presented which simultaneously determines the fractional vascular, interstitial, and cellular volumes of gliomas. This method differs from techniques used in other studies in that it is based on a three-compartment model: a single blood compartment and two interstitial ones. The fractional volume maps are compared with the WHO glioma grading. The results show the method to be feasible. Using cerebral blood volume (CBV), dMRI grading showed a correspondence with WHO grading in 83% of the cases (20/24 gliomas WHO grades II-IV). The use of interstitial volume maps can also be helpful, for instance, in differentiating gliomas from other brain tumors. As a supplement to conventional MRI, dynamic MR techniques thus provide a useful tool for improving in vivo glioma characterization.