点击显示 收起
1 From the Laboratory of Functional and Molecular Imaging, Clinical Center (M.N.Y., Y.Y., J.H.D.), and Laboratory of Diagnostic Radiological Research (F.F.), National Institutes of Health, Bethesda, Md; Department of Psychiatry, Functional Neuro-Imaging Laboratory, Cornell University Medical College, New York, NY (J.A.F.); and Institute of Experimental Medicine and Biotechnology, National Research Council, Cosenza, Italy (F.F.). Received October 23, 2000; revision requested December 6; final revision received May 30, 2001; accepted June 20, 2001. Y.Y. supported by the Whittaker Foundation.
ABSTRACT |
---|
Top ABSTRACT INTRODUCTION Materials and Methods Results Discussion REFERENCES |
---|
Index terms: Brain, diffusion, 10.121413, 10,12144 Brain, MR, 10.121413, 10,12144 Brain, perfusion, 10.121413 Magnetic resonance (MR), functional imaging, 10.121413, 10,12144 • Magnetic resonance (MR), perfusion study, 10.121413
INTRODUCTION |
---|
Top ABSTRACT INTRODUCTION Materials and Methods Results Discussion REFERENCES |
---|
There are two ASL approaches: continuous and pulsed labeling. In continuous labeling (7,10), tagging is achieved by means of continuous inversion of inflowing arterial water in blood by using a long (3-second) adiabatic radio-frequency pulse. Pulsed labeling (8,9,11–13), on the other hand, involves use of a relatively short (10-msec) hyperbolic secant inversion (14) or a frequency-offset–corrected inversion (15) radio frequency for more efficient inversion of the arterial water (16). Although these techniques have become popular for experiments with functional MR imaging (17,18), they are still not used routinely in clinical examinations, primarily because of their characteristic low SNR. This problem is further compounded by additional loss in perfusion signal through T1 as the labeled arterial water protons migrate from the tagging zone to the imaging section. Since the T1 decay rate decreases at higher magnetic field strengths (>3.0 T), the associated perfusion loss is substantially lower owing to the longer T1 of blood (17).
Despite the apparent advantages of high-field-strength systems, the bulk of ASL studies performed to date have been performed at 1.5 T, with a limited number (reported in abstract form [18]) performed at high field strength. The latter may be due in part to the limited number of whole-body high-field-strength MR imaging systems. Given the expected improvement in perfusion SNR at high field strengths and the progressive increase in the number of 3.0-T MR imagers, there is a need to evaluate the performance of ASL perfusion techniques at higher field strengths, particularly in light of the recent approval of 3.0-T magnets for clinical use by the U.S. Food and Drug Administration.
The goals of this study were to (a) evaluate the feasibility and performance of pulsed ASL techniques for perfusion measurements at 3.0 T and (b) measure the perfusion SNR for gray matter and white matter with and without vascular suppression at 3.0 T and compare with the same data at 1.5 T by means of t test analysis. Additionally, the perfusion activation area after a simple motor task was measured at both field strengths, and the activation area was assessed and compared by using standard methods based on functional MR imaging. For this study, we used software and hardware from the same manufacturer.
Materials and Methods |
---|
Top ABSTRACT INTRODUCTION Materials and Methods Results Discussion REFERENCES |
---|
Perfusion-weighted (PW) MR images are acquired with the flow-alternating inversion-recovery technique by performing two experiments: In the first experiment, a selective inversion radio-frequency pulse (section-selective frequency-offset–corrected inversion in this study) is applied to (ideally) selectively invert only the spins in the imaging volume. As shown in Figure 1, this inversion radio frequency is followed by a delay period (inversion time) during which uninverted fresh blood flows into the imaging section, and some exchange with the tissue protons occurs. This exchange increases the bulk apparent relaxation rate in the voxel, which results in an image with signal intensity that is flow weighted. In the second experiment (interleaved with the first), image acquisition is performed with identical timings but with the inversion radio frequency rendered nonselective by switching the frequency-offset–corrected inversion gradient off. In this manner, all incoming blood spins into the section are inverted so that any subsequent exchange with tissue spins does not alter the relaxation rate; hence, a non–flow-weighted image is acquired. The PW images are obtained by subtracting the non–flow-weighted image from the flow-weighted image.
|
Multisection brain perfusion images (10 transverse 5-mm-thick sections) were acquired in six healthy volunteers (three men and three women; age range, 23–30 years; mean age, 26 years). In each subject, 1.5- or 3.0-T MR imaging was performed on a different day. The age range was limited to between 23 and 30 years to minimize age dependency on the perfusion signal. Since the total imaging volume was 50 mm thick (ie, 10 sections x 5 mm), an inversion width of the same size should (ideally) be used in the selective inversion radio-frequency pulse to minimize the transit time of blood to the imaging volume. Unfortunately, use of an inversion width equal to the width of the imaging slab causes interaction between the pulses, which results in contamination from static tissue. On the basis of findings in previous calibration experiments (16), we used a selective inversion width of 80 mm. Bipolar crusher gradients with an amplitude of 15 mT/m (2-msec duration and separation) were applied (Fig 1) to selectively eliminate contributions from large vessels (10,13). In some experiments, the crusher-gradient pulses were switched off to allow the acquisition of PW images with vascular signal intensity.
For absolute regional cerebral blood flow (CBF) quantification, PW images were acquired with a series of 11 inversion time values that ranged from 0.5 to 3.3 seconds. These data permitted the estimation of transit times, on the basis of subsequent signal versus inversion time curves for PW images, which were then used to quantify CBF, as previously described (13). Other imaging parameters included recovery delay, 2.5 seconds; echo time, 17 msec; field of view, 240 mm; 40 flow-weighted and non–flow-weighted raw images to calculate the PW images. To estimate the perfusion SNR, 120 flow-weighted and non–flow-weighted raw images were used to calculate the PW images. Inversion time for the latter measurement was set to 1.8 seconds with crusher gradients and 1.3 seconds without crusher gradients, because these values are known to provide optimal perfusion sensitivity under the two given crusher-gradient conditions (13,17). The spiral readout involved use of trapezoidal gradients to achieve minimal T2* weighting and to minimize the effects of blood oxygenation level dependence. The duration of this pulse was 22 msec, which is substantially shorter than that implemented in most current echo-planar sequences.
Given the widespread use of ASL techniques for functional MR imaging investigations, we also conducted experiments to compare the perfusion-related activation signals at both field strengths by using a simple motor task. Forty perfusion images were collected in 4 minutes (ie, 6 seconds for each flow-alternating inversion-recovery image) during which subjects (n = 4) alternated between rest (off state) and finger tapping (on state). There were four off and four on states, which each lasted 30 seconds. The finger-tapping frequency was paced at 1 Hz by using a computer-controlled digit that was regularly projected on a visible screen to indicate when the subject should finger tap with the thumb of the subject’s dominant hand (right in all cases). The other parameters (recovery delay, 1,575 msec; echo time, 7 msec; inversion time, 1,200 msec; field of view, 24 cm; matrix, 64 x 64; five sections acquired) combined to yield a PW image every 6 seconds. In each examination, the preceding experiment was repeated five times in the same subject to test for reproducibility.
Data Analyses
Both baseline perfusion and functional MR imaging data were analyzed offline with a workstation (SunSparc; Sun Microsystems, Palo Alto, Calif) by using functional MR imaging software (MEDX; Sensor Systems, Sterling, Va) and programs written in-house (Interactive Data Language ; Research Systems, Boulder, Colo). To obtain perfusion data, the nonselective raw images were subtracted from the corresponding selective flow-weighted images. The resultant PW images were averaged to provide a PW image with a relatively high SNR for each inversion time. For SNR estimates, regions of interest (ROIs) (12–15 pixels) were manually prescribed in maps of segmented gray matter and white matter by using the functional MR imaging display software. Segmentation of the gray matter and white matter was based on precalculated T1 maps by using nonselective inversion-recovery images for each subject. The T1-based ROIs for segmentation of gray matter and white matter were then transferred to the same spatial locations on the flow-alternating inversion-recovery PW images, and their mean signal intensity was calculated.
Noise was computed as the SD in a 12–15-pixel ROI that was free of signal intensity and placed outside the PW images in the upper left corner of the image. Images for which the measured noise was more than 5% of the noise estimated on a reference (signal intensity–free) image acquired with parameters identical to those used to acquire the PW images, but with the radio-frequency pulses switched off, were excluded. In this study, the number of excluded PW images ranged between zero and three. The mean perfusion SNR with and without vascular suppression was measured for both gray matter and white matter at both field strengths. A t test analysis was then performed to assess for significant differences between the mean gray matter perfusion SNR and the mean white matter perfusion SNR across the two magnetic fields. These analyses were conducted separately within each technique (ie, with and without vascular suppression).
To further compare the perfusion performance at both field strengths and to estimate absolute CBF, the normalized perfusion-related signal change was integrated in gray matter and white matter ROIs. The ROI approach provided adequate SNR to allow multiparameter fit of the wash-in and washout curves from which transit times were estimated by using previously described methods (13). All ROI prescriptions in this study were performed by one neuroradiologist (F.F.).
Automated image registration performed with the software sought to minimize the sampling coefficient of variation of ratios of voxel values between the first raw image and each successive raw image. Perfusion activation maps were obtained with a pixel-wise cross correlation of perfusion signal-to-time courses with a boxcar reference function that reflected the activation paradigm. For this study, only pixels with a significant activation (P < .05) were included.
Results |
---|
Top ABSTRACT INTRODUCTION Materials and Methods Results Discussion REFERENCES |
---|
|
|
For comparison with results in other groups, the perfusion signal was normalized by dividing it by the equilibrium magnetization (ie, the normalized perfusion-related signal change), since this ratio is independent of the total number of images. Gray matter normalized perfusion-related signal change was a mean 0.708% ± 0.109 at 3.0 T and 0.598% ± 0.079 at 1.5 T. The corresponding white matter normalized perfusion-related signal change at 3.0 T was 0.184% ± 0.063 of equilibrium magnetization and 0.120% ± 0.044 at 1.5 T.
Mean CBF values for the gray matter and white matter ROIs at 3.0 T in the healthy volunteers were 65.1 mL/100 g/min ± 6.1 and 26.3 mL/100 g/min ± 5.9, respectively, with a mean gray matter to white matter CBF ratio of 2.5. Similar results were obtained at 1.5 T. They are summarized in the Table because they have already been reported (10,13).
|
|
|
Discussion |
---|
Top ABSTRACT INTRODUCTION Materials and Methods Results Discussion REFERENCES |
---|
Compared with findings on perfusion images acquired at 1.5 T, the improvement in perfusion signal intensity afforded by the increased field strength is readily apparent from our results. The mean increase in perfusion SNR in gray matter at 3.0 T was substantially greater than that observed for white matter. This discrepancy may be due to the differences in transit time between gray matter and white matter, which was shown in a recent study (19–21) to be approximately 0.5 second longer in white matter than in gray matter. Given the exponential loss of perfusion signal with transit time (10,13), the perfusion signal increase in white matter that results from the increased field strength is likely to be severely offset by transit time losses. In our implementation, we attempted to minimize this problem by using the highly selective frequency-offset–corrected inversion pulse (15,16) in place of the conventional hyperbolic secant pulse (14).
SNR values reported for both field strengths depend primarily on the measurement parameters used and the hardware resources that are available. For example, the perfusion SNR will be different for different imaging section thickness, as will other measurement parameters, such as echo time, repetition time, or both. Furthermore, the use of multiple receive (phased-array) coils will likely enhance the perfusion signal. Because T2* losses are lower at 1.5 T, it may be possible to use a longer spiral readout, which would result in the use of a smaller receiver bandwidth and consequently reduce receiver fold-back noise into the MR images. While this approach may be suitable for single-section ASL studies, it is not appropriate for multisection investigations because it can cause big intersection inversion time differences and quantitative problems. Additionally, the shorter readout duration is desired especially at higher field strengths (3.0 T in this study) to minimize the effects of both T2* and blood oxygenation level dependence on the PW images.
In most functional MR imaging investigations, the simultaneous measurement of blood oxygenation level dependence and perfusion-based activation can provide complementary information. Therefore, a more complete investigation would incorporate measurements of blood oxygenation level dependence obtained with functional MR imaging. However, the measurements of blood oxygenation level dependence were precluded primarily because they have been performed previously (22,23).
In performing the perfusion comparison, we attempted to control for possible variations between the two imagers that may bias the results. For example, we controlled for experimental parameters, manufacturer hardware, and processing software. Owing to the possibility of spatial mismatch in section positioning between the two field strengths, normalization of the data to a standardized space may have been necessary. The latter requirement was minimized by using standard in-house equipment at both field strengths to secure the volunteer’s head in the same spatial position within the standard head coil. Further consistency in spatial localization was ensured by using high-spatial-resolution T1-weighted images as a guide. Additionally, the latter task was performed in all subjects by the same observer (F.F.).
Findings in our study demonstrate that high-quality ASL perfusion images can be acquired at 3.0 T with significant improvement in SNR over measurements performed at 1.5 T. ASL approaches to MR imaging of CBF offer the potential for completely noninvasive quantitative imaging of an important physiologic and diagnostic quantity. For clinical perfusion studies, however, approaches that involve use of dynamic contrast agents still remain the main choice primarily because of better SNR performance. Inadequate SNR continues to be the main drawback in the clinical application of ASL-based techniques. For example, despite the increased sensitivity observed in the current 3.0-T study, absolute quantification of perfusion was possible only with measurements performed over relatively large ROIs from data acquired during 45 minutes. However, the situation is completely different for functional MR imaging applications for which the improved perfusion sensitivity at 3.0 T clearly results in more precise mapping of brain function. Future studies are needed that focus on further improvements in ASL perfusion sensitivity, particularly in conjunction with the use of multiple receiver coils.
REFERENCES |
---|
Top ABSTRACT INTRODUCTION Materials and Methods Results Discussion REFERENCES |
---|