| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, Ehime University School of Medicine, Shitsukawa,
Shigenobu-cho, Onsen-gun, Ehime, 791-0295 Japan.
2 Department of Medical Engineering, Division of Allied Health Sciences, Osaka
University Medical School, 1-7 Yamadaoka, Suita, Osaka, 565-0871 Japan.
3 Department of Neurological Surgery, Ehime University School of Medicine,
Ehime, 791-0295 Japan.
Received April 6, 2001;
accepted after revision December 27, 2001.
Adress correspondence to K. Kikuchi.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Seventeen patients with chronic cerebral artery occlusion or stenosis diagnosed by digital subtraction angiography were examined. Dynamic susceptibility contrast-enhanced MR imaging data were acquired using a multishot echoplanar sequence. Our procedure for quantification of mean transit time was based on the indicator dilution theory and deconvolution analysis.
RESULTS. The increased mean transit time values obtained with dynamic susceptibility contrast-enhanced MR imaging correlated well (r=-0.789, p < 0.0001) with decreased cerebral perfusion reserve determined by performing dynamic 133Xe SPECT before and after administration of acetazolamide. The mean transit time values in the regions with severely decreased perfusion reserve were significantly higher than those in the regions with normal or moderately decreased perfusion reserve (p < 0.0001 and p = 0.0004, respectively).
CONCLUSION. Mean transit time images generated from dynamic susceptibility contrast-enhanced MR imaging data could be used to evaluate the extent of cerebral perfusion reserve impairment in patients with occlusive cerebrovascular disease.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Values for cerebral blood flow, cerebral blood volume, and mean transit time can also be obtained using dynamic susceptibility contrast-enhanced MR imaging [9,10,11,12]. Compared with SPECT or PET, quantification of cerebral perfusion using dynamic susceptibility contrast-enhanced MR imaging has advantages such as improved spatial resolution, absence of radiation exposure, and means of analyzing both morphologic and functional information with data obtained during a single imaging session. The clinical usefulness of dynamic susceptibility contrast-enhanced MR imaging for depicting acute cerebral infarction [13,14,15] and brain tumors [16, 17] has been reported. Initially, only relative (with respect to the normal side of the brain) rather than absolute values of perfusion parameters could be obtained. Recently, however, the dynamic susceptibility contrast-enhanced MR imaging procedure has been improved so that absolute values of the cerebral blood volume and cerebral blood flow can be measured [18, 19]. To quantify cerebral blood flow and cerebral blood volume from dynamic susceptibility contrast-enhanced MR imaging data requires an empiric normalization constant be determined in advance to facilitate correspondence to the PET data [20,21,22]. This conversion constant needs to be determined at every institution and for every image sequence. However, it is possible to determine a quantitative value for the mean transit time without assuming a normalization constant.
In our study, we compared the mean transit time values obtained using dynamic susceptibility contrast-enhanced MR imaging with the cerebral perfusion reserve determined using dynamic 133Xe SPECT.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
On the basis of digital subtraction angiograms, occlusion of the internal carotid artery was diagnosed in three of the 17 patients, and stenosis of the middle cerebral artery was diagnosed in one patient. Stenosis of the internal carotid artery was diagnosed in all patients, and stenosis of the middle cerebral artery was diagnosed in three patients. In seven patients, occlusion and stenosis were found bilaterally.
Dynamic Susceptibility Contrast-Enhanced MR Imaging Protocol
Dynamic susceptibility contrast-enhanced MR imaging studies were performed
on a 1.5-T MR unit (Gyroscan ACS-NT; Philips Medical System, Best, The
Netherlands) using a multishot gradient-echo echoplanar pulse sequence with
parameters of TR range/TE, 252-280/30; flip angle, 40°; and echo factor,
9. The acquisition matrix was 128 x 60, with a field of view of 23
x 17 cm and a slice thickness of 6 mm. Ten milliliters of gadopentetate
dimeglumine was injected into an antecubital vein with a 22-gauge cannula at a
speed of 3 mL/sec by an MR-compatible power injector (Nemoto, Tokyo, Japan),
followed by a 20-mL injection of saline. We obtained 50 dynamic images with a
1.32- to 1.44-sec time resolution in five slices parallel to the anterior
commissure—posterior commissure line. Because TR varied somewhat
(252-280) depending on the imaging plane, we had a range for the time
resolution (1.32-1.44 sec).
Quantification of Mean Transit Time
All dynamic susceptibility contrast-enhanced MR imaging data were
transferred to a postprocessing workstation (Octane; SGI, Mountain View, CA),
and quantitative mean transit time images were generated using our proprietary
software. The arterial input function was obtained automatically from the
internal carotid artery or middle cerebral artery contralateral to the
occluded or severely stenosing side by fuzzy C-means clustering, a
process of classifying multidimensional data in such a way that samples in a
cluster are more similar to one another than to those belonging to a different
cluster [23]. For
determination of arterial input function, we chose the pixels showing the
earliest and largest decreases in signal intensity after contrast injection
among the clustered data. Fuzzy C-means clustering allows automatic
and accurate determination of arterial input function. We described the
details of fuzzy C-means clustering in a previous article
[24].
The tissue residue function was calculated through deconvolution analysis
on single value decomposition between tissue concentration—time curve
and the arterial input function, which is described in detail by Ostergaard et
al. [18]. The initial height
of the deconvoluted concentration—time curve (h[t]) is
proportional to the cerebral blood flow. Therefore, the cerebral blood flow
value was obtained with equation 1:
 | (1) |
Cerebral blood volume was determined from the area under the tissue
concentration—time curve, as described in our previous article
[12]. The cerebral blood
volume was obtained with equation 2:
 | (2) |
) = concentration of contrast agent in the
volume of interest, d = delta (difference), and
CAIF() = concentration of contrast agent in the
arterial input function.
According to the indicator dilution theory, mean transit time is defined by
the central volume principle in equation 3:
 | (3) |
Substituting equations 1 and 2 into equation 3 yielded equation 4:
 | (4) |
Equation 4 indicates that it is not necessary to use the K value for calculating mean transit time. Thus, the mean transit time value can be obtained without normalization.
Because the gradient-echo sequence is sensitive to large vessels, the large vessels on the brain surface tend to be prominently depicted. We objectively eliminated the superficial vessels using the sophisticated automatic threshold selection methods reported by Otsu [25]. This automatic threshold selection method assumes that an image consists of two classes. These two classes are pixels with and pixels without superficial vessels. The optimal threshold value for discriminating between the two classes was determined by maximizing the variance between them [12].
SPECT Protocol
Dynamic 133Xe SPECT studies were performed using a 2000-Hz
scanner (Hitachi Medical, Tokyo, Japan). Patients inhaled 1.85-3.70 GBq of
133Xe gas, and cerebral blood flow was measured according to the
method described by Kanno and Lassen
[26] and by Celsis et al.
[27] using dynamic SPECT data
(16 scans with a duration of 20 sec each). The arterial input curve of
133Xe was estimated from the end-expiratory 133Xe
concentration curve observed by a single sodium iodine detector. The cerebral
blood flow measurement with SPECT was performed with the patient at rest
(unenhanced cerebral blood flow). Fifteen minutes after IV administration of 1
g of acetazolamide, blood flow (enhanced cerebral blood flow) was measured in
the same manner [7]. Cerebral
perfusion reserve was evaluated by the percentage of increase in cerebral
blood flow after administration of acetazolamide using the following formula:
percentage of cerebral blood flow increase = 100 x (enhanced cerebral
blood flow — unenhanced cerebral blood flow) / unenhanced cerebral blood
flow.
Data Analysis
We analyzed four tomographic planes of basal ganglia through the centrum
semiovale using 133Xe SPECT. On each imaging plane, we placed
regions of interest on images of the cerebral cortex that approximately
corresponded to the territory of the middle cerebral artery
(Fig. 1). On 133Xe
SPECT and mean transit time MR images of each patient, regions of interest
were drawn on four slices, such as those shown in
Figure 1. Mean values of mean
transit time in the middle cerebral artery territories were calculated from
the corresponding regions of interest in each hemisphere. We did not register
the mean transit time images and the 133Xe SPECT images because the
spatial resolution of the 133Xe SPECT images was quite inferior to
that of the mean transit time images. However, we tried as much as possible to
draw the region of interest in the same middle cerebral artery territories on
both 133Xe SPECT and MR images. We analyzed 34 hemispheres in 17
patients.
|
The mean values of cerebral blood flow and percentage of increase in
cerebral blood flow obtained from the control subjects (three men and four
women; age range, 48-69 years; mean age, 58.9 years) in our hospital were 50.9
± 3.9 mL/100 g per minute (mean ± SD) and 40.8% ± 12.7%,
respectively [28]. We defined
a percentage of increased cerebral blood flow that exceeded 15% (mean, -2 SDs)
as a normal perfusion reserve. A percentage of increase of 0 or less was
considered to indicate a severely decreased perfusion reserve in which a
stealing effect [7] was
observed. Thus, we classified the perfusion reserve findings into three grades
depending on the percentage of increase in cerebral blood flow: severely
decreased perfusion reserve (percentage of cerebral blood flow increase, 
0%), moderately decreased perfusion reserve (percentage of cerebral blood flow
increase, >0% but 15%), and normal perfusion reserve (percentage of
cerebral blood flow increase >15%). The mean transit time values in the
three grades of perfusion reserve were compared using the Kruskal-Wallis rank
test and then using the Fisher's protected least significant difference test.
The mean transit time values and percentage of cerebral blood flow increase
were also analyzed using the chi-square test. A p value of less than
0.05 was considered significant.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
Regression analysis revealed that the mean transit time values significantly correlated with the cerebral blood flow values (r = -0.567, p = 0.0003) and the percentage of cerebral blood flow increase (r = -0.789, p < 0.0001) (Fig. 3A,3B).
|
|
As for the cerebral perfusion reserve, 24 regions of the 34 hemispheres we studied showed normal perfusion reserve, four regions showed a moderate decrease, and six regions showed a severely decreased perfusion reserve. The cerebral blood flow values obtained with 133Xe SPECT were 33.3 ± 4.8 mL/100 g per minute in the territory of normal perfusion reserve, 32.5 ± 8.0 mL/100 g per minute in territories of moderately decreased perfusion reserve, and 27.8 ± 6.1 mL/100 g per minute in territories of severely decreased perfusion reserve. The mean transit time values were 6.15 ± 1.06 sec for the territories of normal perfusion reserve and 6.54 ± 0.89 sec for the territories of moderately decreased perfusion reserve. For the territories of severely decreased perfusion reserve, the mean transit time value was 9.62 ± 1.84 sec. The coefficient of variation of the mean transit time was 0.35 ± 0.08. The mean transit time values in the territories of severely decreased perfusion reserve were significantly higher than those in the territories of moderately decreased or normal perfusion reserve (p = 0.0004 and p < 0.0001, respectively) (Fig. 4).
|
Given that cerebral perfusion reserve is considered to be severely decreased in a territory in which the mean transit time exceeds 8 sec, we found a significant difference in the percentage of cerebral blood flow increase when using the chi-square test to compare the territory with severely decreased perfusion reserve with the territories with moderately decreased or normal perfusion reserve (p < 0.01, Table 1).
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
In this study of patients with chronic ischemia, we quantitatively evaluated the mean transit time value measured by dynamic susceptibility contrast-enhanced MR imaging in comparison with the cerebral blood flow and percentage of increase in cerebral blood flow measured by dynamic 133Xe SPECT. Although the quantitative mean transit time values correlated with both cerebral blood flow and the percentage of cerebral blood flow increase, transit time values correlated more closely with the percentage of cerebral blood flow increase (Fig. 3A,3B). The explanation for this finding is that, although mean transit time seems to be a good indicator of cerebral blood flow, mean transit time is also influenced by cerebral blood volume. In the compensatory condition of vasoreactivity, in which the cerebral blood flow is decreased and the cerebral blood volume is increased, the mean transit time does not increase significantly. However, once the limit of vasoreactivity (maximal dilatation) has been reached, the mean transit time begins to rise. Therefore, the mean transit time should be more closely correlated with the percentage of cerebral blood flow increase or cerebral perfusion reserve than with cerebral blood flow.
Territories in which the percentage of increase in cerebral blood flow was less than or equal to 0% showed substantially higher (slower) mean transit times than those in which the percentage of increase in cerebral blood flow was more than 0%. In the regions in which the mean transit time exceeded 8 sec, cerebral perfusion reserve had markedly deteriorated (Table 1). This finding is consistent with that of Gibbs et al. [1], who reported that the value of cerebral blood flow divided by cerebral blood volume (the reciprocal of mean transit time) could be used as an index of hemodynamic reserve in patients with carotid occlusive disease. They reported that the mean value of cerebral blood flow divided by cerebral blood volume of unilateral internal carotid artery occlusion was 7.8 mL/100 g per minute and that of bilateral occlusion was 6.4 mL/100 g per minute. If the brain tissue density was assumed to be 1.04 g/mL [29], the blood flow and blood volume values became 7.7 sec and 9.4 sec, respectively, when converted into mean transit times.
Powers and Raichle [2] reported that the measurement of cerebral blood flow alone is not adequate to evaluate hemodynamic compromise. In one patient, the cerebral blood flow value of the right middle cerebral artery territory was 37 mL/100 g per minute. If the determination were based on the cerebral blood flow value alone, the patient's cerebral perfusion reserve would not have been considered particularly deteriorated. However, because the percentage of cerebral blood flow increase in this region was 0%, the cerebral perfusion reserve was considered impaired. In addition, the mean transit time in this region was 8.52 sec, and the cerebral perfusion reserve appeared to be considerably deteriorated. Therefore, mean transit time value could be used to evaluate the cerebral perfusion reserve, even if the cerebral blood flow value was maintained.
As shown in Figure 4, the regions with severely decreased perfusion reserve showed significantly higher mean transit times than those with moderately decreased or normal perfusion reserves. The level of reserve is thought to be a result of autoregulatory vasodilatation caused by reduced cerebral perfusion pressure, indicating that mean transit time can show the cerebral perfusion reserve status. Our results indicate that the mean transit time can indicate regions of decreased cerebral perfusion pressure and may indicate areas of misery perfusion. Therefore, dynamic susceptibility contrast-enhanced MR imaging can provide important clinical information for evaluating the extent of cerebral perfusion reserve impariment [29].
Several mathematic approaches can be used in evaluating cerebral hemodynamics using bolus passages of intravascular contrast agents. However, in applying the indicator dilution theory to MR techniques for assessment of tissue perfusion, one may encounter problems using a model-dependent deconvolution technique. Weisskoff et al. [30] concluded that the application of a simple analytic residue model to the measurement of tissue perfusion would introduce an error. They used mathematic models to show that absolute flow could not be measured with a model-dependent deconvolution technique without an accompanying understanding of the topology of the vasculature. The simulations performed by Ostergaard et al. [18] also showed that assuming a simple monoexponential residue function would introduce large systematic errors when comparing the flows in two regions with different residue functions. However, they indicated that a nonparametric deconvolution technique based on singular value decomposition allowed the estimation of flow to be relatively independent of the underlying vascular structure and volume. In our study, we chose an algebraic method based on singular value decomposition as a deconvolution approach to evaluate cerebral hemodynamics.
In the perfusion protocol described by most MR equipment manufacturers, mean transit time is obtained by fitting a gamma-variate function to the concentration—time curve obtained from each pixel and is calculated as the full width at half maximum without an arterial input function. However, the shape of the concentration—time curve is affected by many factors, such as contrast medium dose, injection rate, and patient condition (heart rate and cardiac output) [31]. Although qualitative evaluation may be possible by taking the ratio of perfusion parameters between the diseased and normal sides, quantitative evaluation presents a potential problem. However, the mean transit time value calculated by deconvolution using arterial input function is not influenced by these factors. For this reason, we developed methods for calculating the mean transit time quantitatively.
The prerequisite for quantification of measurements obtained with dynamic susceptibility contrast-enhanced MR imaging is the determination of arterial input function. Partial volume effects of arterial input function are critical because they may result in an overestimation of cerebral blood flow and cerebral blood volume [11]. To reduce partial volume effects and to standardize the determination of arterial input function, we used fuzzy C-means clustering [12, 24], which allows the automatic and accurate extraction of arterial pixels.
Another potential problem in determining arterial input function is the dispersion of the bolus on its way from the region in which the arterial input function is to be measured to the brain parenchyma. Stenosis and occlusion of the middle cerebral artery may cause this dispersion. In our study, three patients had middle cerebral artery stenosis, and one patient had middle cerebral artery occlusion. In the patient with middle cerebral artery occlusion, collateral vessels were confirmed with digital subtraction angiography. None of these patients had a cortical infarction area in the middle cerebral artery territories. The mean transit time values for one of the patients with middle cerebral artery stenosis and the one patient with middle cerebral artery occlusion was 8.55 and 8.52 sec, respectively. The corresponding 133Xe SPECT data showed severely decreased perfusion reserve (-7% and 0%, respectively). The mean transit time values for the two other patients with middle cerebral artery stenosis were within the normal range (6.59 and 4.58 sec, respectively), and the values of cerebral perfusion reserve were also normal (45% and 34%, respectively). Evaluation of the cerebral perfusion reserve with mean transit time and percentage of cerebral blood flow increase showed good agreement in these patients. The dispersion of the bolus caused by the middle cerebral artery stenosis or occlusion did not have a serious influence on the results of our study.
To measure absolute cerebral blood flow using dynamic susceptibility contrast-enhanced MR imaging, an empiric normalization constant has been introduced. Ostergaard et al. [20, 21] reported that the K values in equations 1 and 2 were 1.09 for pigs and 0.87 for human volunteers in quantification using a spin-echo echoplanar pulse sequence to correspond with PET data. Simonsen et al. [22] reported that the K values were 2.96 for cerebral blood volume and 2.53 for cerebral blood flow using gradient-echo echoplanar pulse sequence. It is necessary to predetermine the empiric normalization constant K used at the institution or for the image sequence, which is troublesome for quantification in a clinical setting. Here, however, we have shown that the quantitative mean transit time can be obtained without determining the normalization constant K. Simonsen et al. also reported that the mean transit time values derived by gradient-echo type echoplanar pulse sequences were not significantly different from those obtained by spin-echo echoplanar pulse sequences. This finding suggests that when evaluating mean transit time, we do not need to consider the difference caused by the use of a different imaging sequence. For these reasons, the mean transit time images derived by our procedure from dynamic susceptibility contrast-enhanced MR imaging data are considered both quantitative and clinically useful.
A limitation of our method is that the whole brain cannot be evaluated. We carried out dynamic susceptibility contrast-enhanced MR imaging for five slices and evaluated the brain hemodynamics from four slices. Because we gave priority to time resolution, a whole-brain evaluation could not be done. Using a single-shot echoplanar sequence (instead of a multishot approach) increases the number of slices, and one can evaluate the hemodynamics of the whole brain. However, distortion of the images and inferior spatial resolution occur with whole-brain analyses. The distortion of images may make it difficult to obtain adequate arterial input function. Thus, we used a multishot echoplanar sequence for dynamic susceptibility contrast-enhanced MR imaging to avoid as many of these problems as possible.
Another MR imaging—based perfusion method that can be used to measure cerebral perfusion is arterial spin-labeling. Several arterial spin-labeling techniques have been introduced, such as echoplanar imaging and signal targeting with alternating radiofrequency [32] and flow-sensitive alternating inversion recovery [33]. The basic principle of these techniques involves saturating the blood flowing to the brain in the neck region with a saturation pulse and then, after a delayed inversion time, acquiring the image. The advantages that arterial spin-labeling has over dynamic susceptibility contrast-enhanced MR imaging are that repeated measurements can be obtained easily over time and that arterial spin-labeling is a noninvasive procedure requiring no contrast medium. However, the signal change of the labeled blood is quite small, and the influence of the magnetization transfer effect cannot be disregarded in arterial spin-labeling. Additionally, when multiple slices are acquired at different inversion times, the images have different cerebral blood flow weights because of differences in blood transit time. With arterial spinlabeling, the quantification of cerebral blood flow requires more time, because the tissue T1 relaxation time and possibly the transit time have to be measured [29].
In conclusion, the procedure we described can be used to assess a patient's cerebral hemodynamic status with higher spatial resolution than that of established radionuclide techniques. Cerebral perfusion reserve impairment can be evaluated with mean transit time measured using dynamic susceptibility contrast-enhanced MR imaging.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  N.-J. Rim, H.S. Kim, Y.S. Shin, and S.Y. Kim Which CT Perfusion Parameter Best Reflects Cerebrovascular Reserve?: Correlation of Acetazolamide-Challenged CT Perfusion with Single-Photon Emission CT in Moyamoya Patients AJNR Am. J. Neuroradiol., October 1, 2008; 29(9): 1658 - 1663. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K.H. Kang, H.S. Kim, and S.Y. Kim Quantitative Cerebrovascular Reserve Measured by Acetazolamide-Challenged Dynamic CT Perfusion in Ischemic Adult Moyamoya Disease: Initial Experience with Angiographic Correlation AJNR Am. J. Neuroradiol., September 1, 2008; 29(8): 1487 - 1493. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Ohno, H. Hatabu, K. Murase, T. Higashino, M. Nogami, T. Yoshikawa, and K. Sugimura Primary Pulmonary Hypertension: 3D Dynamic Perfusion MRI for Quantitative Analysis of Regional Pulmonary Perfusion Am. J. Roentgenol., January 1, 2007; 188(1): 48 - 56. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
O. Togao, F. Mihara, T. Yoshiura, A. Tanaka, T. Noguchi, Y. Kuwabara, K. Kaneko, T. Matsushima, and H. Honda Cerebral Hemodynamics in Moyamoya Disease: Correlation between Perfusion-Weighted MR Imaging and Cerebral Angiography. AJNR Am. J. Neuroradiol., February 1, 2006; 27(2): 391 - 397. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Cronqvist, R. Wirestam, B. Ramgren, L. Brandt, B. Romner, O. Nilsson, H. Saveland, S. Holtas, and E.-M. Larsson Endovascular Treatment of Intracerebral Arteriovenous Malformations: Procedural Safety, Complications, and Results Evaluated by MR Imaging, Including Diffusion and Perfusion Imaging AJNR Am. J. Neuroradiol., January 1, 2006; 27(1): 162 - 176. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Ohno, T. Higashino, D. Takenaka, K. Sugimoto, T. Yoshikawa, H. Kawai, M. Fujii, H. Hatabu, and K. Sugimura MR Angiography with Sensitivity Encoding (SENSE) for Suspected Pulmonary Embolism: Comparison with MDCT and Ventilation-Perfusion Scintigraphy Am. J. Roentgenol., July 1, 2004; 183(1): 91 - 98. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Yamamoto, M. Watanabe, T. Uematsu, K. Takasawa, M. Nukata, and N. Kinoshita Correlation of Angiographic Circulation Time and Cerebrovascular Reserve by Acetazolamide-Challenged Single Photon Emission CT AJNR Am. J. Neuroradiol., February 1, 2004; 25(2): 242 - 247. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE |
= moderately decreased perfusion reserved 
= normal perfusion
reserve. Relationship between mean transit time and cerebral blood flow is
shown. We found significant correlation between mean transit time and cerebral
blood flow (r = -0.567, p = 0.0003).
