| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Duke University Medical Center, Box 3808, Durham, NC
27710.
2 Duke University School of Medicine, Durham, NC.
3 Department of Biomedical Engineering, Duke University, Durham, NC.
4 Novartis Pharmaceuticals, Inc., East Hanover, NJ.
Received April 29, 2004;
accepted after revision September 4, 2005.
Address correspondence to J. M. Provenzale.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Ten patients with biopsy-proven high-grade gliomas underwent dynamic contrast-enhanced MRI using T1-weighted fast spoiled gradient-echo technique (TR/TE, 8.3/1.5) during IV infusion of 0.1 mmol/kg of MR contrast medium. This sequence was followed within 5 minutes by dynamic susceptibility contrast (DSC) imaging (1,500/80) during IV infusion of 0.2 mmol/kg of MR contrast medium. Dynamic contrast-enhanced analysis was performed using the maximum-signal-intensity algorithm, and DSC analysis was performed using the negative enhancement integral program. For each tumor, we performed two comparisons: first, the average dynamic contrast-enhanced and rCBV values within a region of interest drawn around the entire contrast-enhancing tumor on a single image through the center of the lesion and, second, the highest dynamic contrast-enhanced and highest rCBV values within each tumor. Statistical analyses of the first comparison were performed using Pearson's correlation coefficient, R2 correlation coefficient, and Spearman's rank correlation and for the second comparison using Kendall's tau correlation.
RESULTS. The mean signal intensity values ranged between 3.48 and 7.16 SDs above baseline values (mean, 4.89 SDs). The mean rCBV values ranged between 57.9% and 122.7% of the normal lentiform nucleus (mean, 76.6%). The Pearson's correlation coefficient was 0.867, the R2 correlation coefficient was 0.752, and the Spearman's rank correlation was 0.794 (p = 0.001). Dynamic contrast-enhanced values from the region of highest signal intensity ranged between 7.7 and 48.6 SDs above baseline values (mean, 17.3 SDs). The highest rCBV values ranged between 105% and 400% of the normal lentiform nucleus (mean, 292%). The correlation was estimated to be 0.7778 and was statistically significant at the 0.01 level of statistical significance (p = 0.0035).
CONCLUSION. We found a high correlation between degree of contrast enhancement on dynamic contrast-enhanced images and rCBV values in whole tumors and in regions having the highest degree of contrast enhancement in this small study. Our findings, which suggest that relative permeability and rCBV values may be correlated in high-grade glial neoplasms, deserve further study in a larger patient population.
Keywords: brain • gliomas • oncologic imaging • perfusion-weighted MRI
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
It is well established that vessel growth in tumors is strongly influenced by a number of angiogenesis factors [2]. One of the most important angiogenesis factors under investigation is vascular endothelial growth factor (VEGF), which acts as both a potent tumor angiogenesis factor and a potent permeability factor by acting directly on the endothelium [3]. Because VEGF affects both tumor vascularity and vascular permeability, it is reasonable to ask whether these two features of tumors are correlated on hemodynamic MRI studies. Measuring the rate of contrast enhancement provides a method of measuring leakage of contrast material across an interrupted blood-brain barrier and thus is an indirect means of measuring permeability. One MRI technique that has been successfully used for this purpose is T1-weighted dynamic contrast-enhanced imaging [4]. However, because VEGF and other angiogenesis factors also produce an increase in the number of vessels, these factors would also be expected to increase cerebral blood volume within tumors. In fact, this feature has been shown in numerous trials. Specifically, investigators have shown that high-grade brain tumors have substantially higher relative cerebral blood volume (rCBV) as measured on T2*-weighted dynamic susceptibility contrast (DSC) MRI than low-grade tumors, which parallels histologic features of both tumor types [5]. The hypothesis that higher rCBV correlates with higher lesion grade has been verified by a number of DSC MRI-based studies [6-9].
Based on the previous discussion, it seems reasonable that vessel permeability within tumors and tumoral rCBV should be correlated. However, few imaging studies have addressed this possibility. Our hypothesis was that the rate of contrast enhancement as measured on dynamic contrast-enhanced imaging would correlate with rCBV as measured on DSC imaging in high-grade neoplasms.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The standard imaging protocol consisted of a dynamic contrast-enhanced T1-weighted pulse sequence followed within 10 minutes by a DSC T2*-weighted pulse sequence. The T1-weighted 3D fast spoiled gradient-echo sequence was performed during bolus infusion of 0.1 mmol/kg of gadopentetate dimeglumine (Magnevist, Berlex Laboratories) at 4 mL/s with TR/TE, 11.9/5.3. Imaging parameters were as follows: 10 slices, 7.5-mm slice thickness, zero gap, 256 x 128 matrix, 24-cm field of view, 1 excitation, and imaging time of 9 minutes 20 seconds. This sequence allows calculation of pixel intensity, which we used as a measure of relative permeability.
The T2*-weighted sequence used 1,500/80 and was performed during a bolus infusion of 0.2 mmol/kg of gadopentetate dimeglumine at 4 mL/s. Imaging parameters were as follows: 8 slices, 40 images per slice, 7.5-mm slice thickness, 2.5-mm gap, 256 x 128 matrix, 24-cm field of view, 1 excitation, and imaging time of 1 minute. This pulse sequence allowed calculation of rCBV. Infusion of a half dose (i.e., 0.05 mmol/kg) of contrast material before performance of T2*-weighted imaging has been advocated to minimize T1 effects on rCBV measurements [10]. Therefore, initial performance of dynamic contrast-enhanced imaging (even using a standard dose of 0.1 mmol/kg rather than a half dose) was not thought to interfere with the DSC portion of the study. Imaging for the study was performed on a total of three 1.5-T MR scanners (all Signa, GE Healthcare). All patients provided written informed consent for the study, and the institutional review board at our hospital approved the study.
Dynamic Contrast-Enhanced Data Analysis
Tumors on dynamic contrast-enhanced images were analyzed for highest signal
intensity change during contrast material administration on a workstation
(Advantage Windows, GE Healthcare) using the Maximal Pixel Intensity program
operating on FuncTool software (GE Healthcare). This program gives a measure
of degree of contrast enhancement, which is a relative measure of
permeability, within the lesion. In each case, we visually assessed the
resultant time-signal intensity curves to confirm that a rapid rise in signal
was seen soon after infusion of the contrast bolus. All analyses were
performed on the single axial slice that showed the greatest tumor diameter.
The single signal intensity image showing peak enhancement was first chosen
and saved on the FuncTool workstation as a DICOM image and translated into a
text image using the ImageJ program (version 1.28u, National Institutes of
Health [NIH]). These text images provide a matrix consisting of numbers that
represents signal intensity values for each pixel.
The method for drawing regions of interest (ROIs) was as follows. First, we drew an ROI around the enhancing portion of the tumor on the DICOM image and then flipped and superimposed that ROI onto the same position within the brain in the normal hemisphere. Then, the mean value and SD of signal intensities were calculated for the pixels within the ROI drawn in the normal hemisphere, which served as the basis for comparison with tumor values. Next, the signal intensity values within each pixel in the ROI surrounding the tumor were recorded and expressed as a function of the number of SDs above the mean value in the ROI that had been placed in the normal hemisphere. We then color-coded the images using 3 SDs as the lower threshold and assigning different colors for the following three ranges of values: 3-5 SDs, 5-7 SDs, and > 7 SDs above normal background tissue (Figs. 1A, 1B, 1C, 1D, and 1E).
|
|
|
|
|
We validated the technique for using mean signal intensity values by measuring the effect of various locations of ROIs on mean normal values in the normal hemisphere. Using three ROIs that differed slightly in size but were each placed at five different locations in the normal hemisphere by a single observer, the mean signal intensity value for each ROI was recorded. The mean signal intensities of the three ROIs were 55.67 ± 5.94, 55.05 ± 6.10, and 55.74 ± 6.61. Therefore, various placements of the ROI within the normal hemisphere appear to have little effect on the resultant mean values in normal tissue.
To validate the reliability of our ImageJ software program, we tested the degree of interobserver variability by having eight observers independently measure mean signal intensity and enhancing tumor size in the patient population reported in this study. This analysis showed that the interobserver variability for use of our program is very low. After using the ImageJ program for our original analysis, we developed a more user-friendly version of the program that reduced the time for analysis from approximately 10 minutes to approximately 20 seconds. The new version of the program is otherwise the same as the first program and yields the same results. We found a very low degree of interobserver variability for both the measurement of tumor size and the measurement of the degree of contrast enhancement using the more recent version of our software program.
Highest Pixel Intensity Analysis on Dynamic Contrast-Enhanced Images
In addition to measuring mean values of signal intensity in the entire
enhancing tumor on a single image through the center of the lesion, a second
observer measured the signal intensity values in the region having the highest
signal intensity values anywhere in the tumor. Because the original dynamic
contrast-enhanced and cerebral blood volume (CBV) images for one patient were
unavailable at the time the new technique was developed, this analysis was
performed on nine of the 10 patients enrolled in the study. For this analysis,
we developed another software program that allows the mean values in a cluster
of 6 pixels to be displayed as the computer mouse is moved over the tumor.
Using this program, we were able to find the 6-pixel ROI with the highest
signal intensity within the tumor. This technique allowed us to specifically
study the heterogeneity of the lesion, rather than the average for the entire
tumor on a single image through the center of the lesion, which may include
areas of necrosis or other nontumor hemodynamic abnormalities. We subsequently
compared the highest values within the tumor with the highest rCBV values in
the tumor. The highest relative permeability values anywhere in the tumor were
compared against the highest rCBV values anywhere in the tumor. The locations
for these measurements always differed from one another in the same tumor.
DSC Data Analysis
The rCBV maps were generated on an Advantage Windows workstation (GE
Healthcare) by a single observer using the negative enhancement integral
program available in the FuncTool software program. The same observer analyzed
all rCBV maps. Briefly, the rCBV maps constructed from DSC MR images were
based on the change in signal intensity (S) from baseline
(S0) observed during the passage of contrast material.
Subsequently, this measurement was converted to a change in T2*
relaxation rate by using the following formula:
 |
The same observer analyzed all rCBV maps. The CBV mask image that corresponded to the entire enhancing tumor on a single image through the center of the lesion on the dynamic contrast-enhanced maps was superimposed on the DSC image after ensuring that the image slice location and the slice orientation between the two images were similar. Resizing of the images was also performed as necessary to 256 x 256 pixels. DSC data were normalized in the following manner. We did not use the same ROI reflected on the exact contralateral location (as for dynamic contrast-enhanced image analysis) because the two ROIs might differ with regard to the proportions of gray matter and white matter, which could result in inaccurate estimations of ratios of rCBV within tumor to normal brain tissue. Instead, an ROI was placed around the entire lentiform nucleus on a single image through the center of this structure in the hemisphere contralateral to the tumor after establishing that the lentiform nucleus had a normal appearance on all images. The mean rCBV value of the lentiform nucleus was recorded and used as a means of normalizing rCBV values found within tumors. Using the same technique outlined earlier for the dynamic contrast-enhanced data, all the pixels in the ROI in the tumor on the rCBV map were expressed as a percentage of the mean value in the lentiform nucleus in that patient. The mean percentage across pixels was then averaged for that patient and compared with the mean signal intensity value for that patient's tumor on the dynamic contrast-enhanced analysis (Figs. 1A, 1B, 1C, 1D, and 1E).
|
|
Statistical Analysis
Dynamic contrast-enhanced signal intensity measurements and rCBV
measurements within the entire tumor on a single image through the center of
the lesion were compared for each patient using Pearson's correlation
coefficient, R2 correlation coefficient, and Spearman's
rank correlation. The Spearman's rank correlation was calculated in addition
to the Pearson's correlation coefficient to verify the robustness of the
statistical inference. Highest dynamic contrast-enhanced measurements and
highest rCBV measurements were compared for each patient using Kendall's tau
correlation.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Dynamic contrast-enhanced values from the region of highest signal intensity ranged between 7.7 and 48.6 SDs above baseline values (mean, 17.3 SDs). The maximum rCBV values ranged between 105% and 400% of the normal lentiform nucleus (mean, 292%). A plot of signal intensity data and rCBV data is shown in Figure 3. The correlation was estimated to be 0.7778 and was statistically significant at the 0.01 level of statistical significance (p = 0.0035).
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
A number of imaging studies focusing on either vascular permeability or cerebral blood volume in brain tumors have been published. However, imaging studies correlating permeability and blood volume within brain tumors are relatively rare. In one study, both microvascular permeability and an assessment of the fractional blood volume were measured in human brain tumors using a dynamic contrast-enhanced technique [15]. The authors of that study found very good correlation between microvascular permeability and tumor grade and only moderate correlation between fractional blood volume and tumor grade. Correlation between microvascular permeability and fractional blood volume was not reported. In one small study of high-grade gliomas in humans, investigators compared endothelial permeability and rCBV from the same dynamic contrast-enhanced sequence [16]. Permeability was assessed by calculating the endothelial permeability surface area product using an iterative estimation that decomposes tissue residue function into intravascular and extravascular components and then generating tumor blood volume maps from T1-weighted images corrected for contamination by contrast agent leakage. Those investigators found a very good correlation between histograms of voxel values of permeability surface area product and those of T1-weighted blood volume maps corrected for contrast material leakage, and they found a good correlation between distributions on the T1-weighted rCBV map and T2*-weighted rCBV maps in the same patients. However, the investigators in that very promising study did not directly correlate a measure of permeability (endothelial surface area product) and rCBV values derived from T2*-weighted images, as in our study.
In another study, investigators used a modified pharmacokinetic model to map T1-weighted dynamic contrast-enhanced data to quantify blood volume according to various compartments such as blood, cells, and interstitium [17]. Using this technique, the tumor fractional vascular volume was found to correlate well with tumor grade. Finally, in another study, investigators assessed both the rCBV and the volume transfer constant (Ktrans) in typical meningiomas and atypical meningiomas using dynamic contrast-enhanced perfusion MRI [18]. That study showed that the two tumor types could be distinguished on the basis of Ktrans but not rCBV but that the two hemodynamic parameters were not correlated in that study population. In our study, we did not assess Ktrans but instead measured an indirect expression of Ktrans—that is, the degree of contrast enhancement using the maximum signal intensity algorithm. Although our technique of measuring signal intensity does not specifically measure permeability as such (as is available by measuring Ktrans), it is more straightforward, less computationally complex, and substantially easier to reproduce than the Ktrans approach. Nonetheless, we are currently undertaking a prospective comparison of measurements of Ktrans and measurements obtained using our technique for measuring the degree of contrast enhancement.
Because many of the imaging studies in humans in the medical literature reporting rCBV in tumors have used T2*-weighted (DSC) imaging, we used that technique for our measurements of rCBV. Some limitations of this technique, such as the ability to obtain only relative values, have been noted [15]. However, the technique is easy to use and software programs for rCBV analysis are available on many commercially available workstations. Furthermore, a large number of tumor imaging studies that use this technique have been published against which results of new studies can be compared [5-9]. In our study, we used the normal contralateral lentiform nucleus as our internal reference. This structure, by virtue of its gray matter composition, is expected to have high rCBV; furthermore, it is a well-demarcated structure that is easily identifiable, which should assist in studies that attempt to reproduce our findings. We did not use the entire contralateral hemisphere as a comparison reference for the DSC analysis because the entire hemisphere is an inhomogeneous structure composed of variable amounts of gray matter, white matter, and blood vessels. The resultant mean rCBV values might substantially differ from slice to slice and from patient to patient. As expected, a range of rCBV ratios was seen within the entire tumor on a single image through the center of the lesion, with a mean for all tumors that was approximately 75% of rCBV values in the lentiform nucleus. These values indicate that the rCBV within tumors was generally close to that of gray matter, which has a much higher rCBV than does white matter.
We performed an initial assessment of tumor heterogeneity by examining regions of highest relative permeability and highest rCBV within tumors. That assessment showed that in our small group of high-grade brain tumors, regions containing very high relative permeability and rCBV values exist. These sites, which are likely to represent the location of high degrees of angiogenesis (and therefore the highest grade of tumor), could potentially serve as targets for biopsy if the region of tumor having the highest grade is sought for staging purposes.
In our study, we used a well-established form of dynamic contrast-enhanced imaging using commercially available software. Our findings of high rates of relative permeability within tumors are consistent with many previous reports [15, 16]. Relative permeability values within the entire contrast-enhancing tumor on a single image through the center of the lesion were always at least approximately 3.5 SDs above that of normal brain tissue, with a mean value of approximately 5 SDs above that of normal tissue. These findings, which reflect the interrupted blood-brain barrier that is characteristic of neoangiogenesis in brain tumors, may prove to be a fruitful target for monitoring antiangiogenesis therapy. We found a high degree of correlation of rCBV and relative permeability values in our patients in assessment of both the entire contrast-enhancing tumor on a single image through the center of the lesion and small regions showing the highest relative permeability and highest rCBV values. These findings may reflect the dual effects of angiogenesis promoters on blood volume and the leakiness of the blood-brain barrier.
We recognize that abnormal blood vessels exist within the ROIs drawn around our tumors during dynamic contrast-enhanced signal intensity measurements. Therefore, the signal intensity from contrast material within these vessels is expected to contribute to the overall signal intensity of the lesion. To some extent, then, our signal intensity measurements also reflect blood volume and, thus, relative permeability and blood volume are not wholly separable. However, our data suggest that the combination of imaging techniques that we used may appropriately reflect both the increased vascularity within tumors and the fact that new blood vessels lack the tight endothelial junctions that contribute to the blood-brain barrier. It is also possible that neither of these techniques adequately distinguishes the effects of increased permeability from those of hypervascularity.
We recognize that our small number of patients precludes application of our findings to brain tumor patients as a whole. Larger studies comparing these two hemodynamic parameters are needed before our results can be deemed reliable. Further studies designed to correlate our findings with histologic samples and studies to determine whether both hemodynamic parameters decrease proportionately in response to antiangiogenesis therapy are warranted.
Simultaneous mapping of blood-brain barrier permeability and rCBV values has been reported using CT [19]. However, correlation of permeability and rCBV values was not performed in that study. The reported theoretic advantages of CT techniques include the fact that the relationship between CT density and CT contrast concentration is linear (as opposed to the nonlinear relationship between MR contrast material concentration and MR signal intensity), making quantification easier, and the absence of susceptibility effects, which complicate the analysis of data obtained using T2*-weighted DSC techniques. These susceptibility effects are avoided using T1-weighted dynamic contrast-enhanced techniques.
In summary, our study showed a significant correlation between two hemodynamic parameters that are thought to reflect different, but related, fundamental physiologic processes within brain neoplasms. This correlation was seen both in the entire contrast-enhancing tumor on a single image through the center of the lesion and in small regions of the tumor containing the sites of highest relative permeability and highest rCBV. If verified by future studies that also provide correlation with histologic data, these imaging techniques may prove to be fruitful methods with which to monitor tumor progression or response to antiangiogenesis therapies and to determine optimal biopsy sites. The results of this small study need to be confirmed by a larger study before a definitive determination can be made whether dynamic contrast-enhanced and DSC techniques each provide unique or overlapping information. Optimally such a study would compare data obtained using each technique against histologic data or outcome data in patients. Such information is needed before conclusions about an optimal choice of imaging technique can be made.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
