Quantification of Rodent Cerebral Blood Flow (CBF) in Normal and High Flow States Using Pulsed Arterial Spin Labeling Magnetic Resonance Imaging

Transcription

1 JOURNAL OF MAGNETIC RESONANCE IMAGING 26: (2007) Original Research Quantification of Rodent Cerebral Blood Flow (CBF) in Normal and High Flow States Using Pulsed Arterial Spin Labeling Magnetic Resonance Imaging Susanne Wegener, MD, 1,3 * Wen-Chau Wu, PhD, 1 Joanna E. Perthen, PhD, 1 and Eric C. Wong, MD, PhD 1,2 Purpose: To implement a pulsed arterial spin labeling (ASL) technique in rats that accounts for cerebral blood flow (CBF) quantification errors due to arterial transit times (dt) the time that tagged blood takes to reach the imaging slice and outflow of the tag. Materials and Methods: Wistar rats were subjected to air or 5% CO 2, and flow-sensitive alternating inversion-recovery (FAIR) perfusion images were acquired. For CBF calculation, we applied the double-subtraction strategy (Buxton et al., Magn Reson Med 1998;40: ), in which data collected at two inversion times (TIs) are combined. Results: The ASL signal fell off more rapidly than expected from TI one second onward, due to outflow effects. Inversion times for CBF calculation were therefore chosen to be larger than the longest transit times, but short enough to avoid systematic errors caused by outflow of tagged blood. Using our method, we observed a marked regional variability in CBF and dt, and a region dependent response to hypercapnia. Conclusion: Even when flow is accelerated, CBF can be accurately determined using pulsed ASL, as long as dt and outflow of the tag are accounted for. Key words: arterial spin labeling; cerebral blood flow; hypercapnia; magnetic resonance imaging; arterial transit times J. Magn. Reson. Imaging 2007;26: Wiley-Liss, Inc. 1 Department of Radiology, University of California San Diego, San Diego, California, USA. 2 Department of Psychiatry, University of California San Diego, San Diego, California, USA. 3 Berlin NeuroImaging Center, Berlin, Germany. Contract grant sponsor: Deutsche Forschungsgemeinschaft, 6th Framework Program of the European Commission (Marie Curie Outgoing International Fellowship); Contract grant sponsor: National Institutes of Health (NIH); Contract grant number: R01 EB *Address reprint requests to: S.W., MD, Center for Functional MRI, Department of Radiology, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA. swegener@ucsd.edu Received December 8, 2006; Accepted May 7, DOI /jmri Published online in Wiley InterScience ( ARTERIAL SPIN LABELING (ASL) (2) has evolved into an important alternative to contrast-enhanced bolus tracking techniques for imaging of cerebral perfusion. It does not rely on the application of a contrast agent, but directly assesses intrinsic blood magnetization that has been modulated through a radio frequency (RF) pulse, either applied continuously (continuous: CASL) or intermittently (pulsed: PASL) (3,4). The labeled arterial blood then travels into the imaging plane, where a tag image is acquired. In a control acquisition, the inflowing arterial blood is not tagged, so that after subtraction of tag and control images, static tissue magnetization cancels out, and the remaining signal difference is proportional to local perfusion. Quantitative values for cerebral blood flow (CBF) have been successfully obtained using ASL, and have been validated against invasive autoradiographic (5) or microsphere-based (6) techniques to measure blood flow. Although there has been good agreement between these methods for baseline blood flow, discrepancies have been noted in states of altered perfusion (6,7). One major source of error is the arterial transit time (dt), which is the time that tagged blood takes to reach the imaging slice (8). In continuous ASL approaches, a postlabeling delay is typically inserted to account for transit delays (9). In contrast, although dt correction methods have been developed for human pulsed ASL applications (10), these have not been implemented for rodent imaging. One reason could be that the arterial transit time is rather short in rodent PASL experiments, because baseline CBF is high and the gap between the labeling and imaging plane can be kept small. Some groups have therefore not considered dt in their CBF quantification approaches (11,12). However, dt is not a constant, but a region-dependent variable that can introduce unpredictable errors into the CBF estimation (1,13 15). Aside from arterial transit delay considerations, certain assumptions implicit in most CBF quantification models might not be valid in altered flow states. Among them are the assumptions that all of the tagged blood water that enters the imaging slice exchanges with tis Wiley-Liss, Inc. 855

2 856 Wegener et al. sue water, and that there is no significant outflow of tagged water. When blood flow increases, the blood/ tissue water extraction fraction decreases, leading to an underestimation of CBF (16,17). Further attenuation of the ASL signal with elevated flow rates might arise from significant amounts of tagged blood leaving the imaging slice before image acquisition (outflow). It is not known if and how such changes in the distribution and flow of blood water might affect the CBF quantification with PASL. In this study, we demonstrate the changes in arterial transit times and the ASL signal course in normal and high flow conditions (breathing air vs. 5% CO 2 )inthe rat brain. We present a robust PASL method based on the double subtraction strategy introduced by Buxton et al (1) in 1998, which is insensitive to these sources of CBF quantification error. While CASL methods have been previously modified to incorporate such CBF quantification issues (14), this, to our knowledge, represents the first reported implementation of an equivalent PASL technique for rodent imaging. Using this approach, we show that changes in CBF and arterial transit time with hypercapnia (5% CO 2 ), a method commonly used to calibrate functional blood oxygenation level dependent (BOLD) experiments (18), are region-dependent and might resemble differential responses of cerebral vasculature to changes in perfusion. MATERIALS AND METHODS Animals MRI data were obtained on five adult male Wistar rats. Animals were anesthetized with 3% isoflurane in medical air (21% O 2 ), and maintained on spontaneous inhalation anesthesia with 1.5% isoflurane through a nose cone. Body temperature was monitored and maintained at 37 C, and respiratory rates were monitored. For induction of hypercapnia, the isoflurane concentration was kept constant while the inhaled gas composition was changed to 5% CO 2 in medical air. A twominute adjustment was allowed after switching the gas before starting the data acquisition. For head restraint, ear bars were used. An Eppendorf tube containing saline was taped to the rat head to obtain a reference signal for the calculation of CBF (see below). Animals quickly recovered after the MRI scans. In six additional Wistar rats, anesthesia was induced and maintained as above, the tail artery was catheterized, and arterial blood gases were obtained for the air and 5% CO 2 condition. MRI Experiments All experiments were done on a 3T GE Signa Excite whole-body system with a body transmit coil and a custom built single-loop receive-only head coil of 3 cm diameter, which was passively decoupled during transmit. Anatomical images (T2 weighted) were acquired using a fast spin-echo sequence with the same slice prescription as the ASL experiment. For perfusion imaging, we used multislice flow-sensitive alternating inversion recovery (FAIR), a pulsed ASL method developed by Kwong et al (19) and further advanced by Kim (20) and Kim and Tsekos (21), in which a global inversion pulse is used to generate the tag and a slice-selective inversion pulse to create the control image. A multishot gradient echo spiral readout was used for data acquisition. An in-plane saturation (PreSat) was applied immediately after the preparatory inversion pulse and was used in conjunction with two nonselective adiabatic inversion pulses to provide background suppression and thereby stabilize the signal (22). These pulses also served to avoid signal negation when the inversion time fell between the null points of tissues with different T1s (20,23). Because of the imperfection in slice profile, we set both the slice-selective ( control ) FAIR inversion pulse and the PreSat pulse 5 mm thicker than the imaging region at both sides, which created a 5-mm gap between the tagging and imaging regions. We found that despite the background suppression, the pulsation of the carotids below the brain resulted in substantial artifact in the slice of interest. Therefore, a spatial saturation pulse was applied in the coronal plane at the level of the arteries, significantly reducing the artifact. To avoid modulation of the inflowing blood signal for the subsequent acquisition by this pulse, the image acquisition was immediately followed by a global saturation (PostSat) to reset the blood signal. Imaging parameters were: slice thickness 2mm (three slices), gap 1 mm, FOV 4 4 cm, matrix size 64 64, flip angle 90, number of interleaves 8, number of repetitions 10, and TE 4.3 msec. To obtain the full ASL signal curve, images were acquired at the following inversion times (TI): 0.3, 0.4, 0.6, 0.8, 1.0, 1.25, 1.5, 1.75, 2.0, and 3.0 seconds with TR TI 3.4 seconds, under anesthesia with air and 5% Co 2 in three animals. After reviewing the data, ASL images were acquired only up to a TI of 1.25 seconds in the remaining two animals. An equilibrium magnetization reference scan was acquired as follows: no tag or saturation pulses, TR 15 seconds, TE 4.3 msec, one repetition. Although the tag created by the body coil was very homogenous, the signal received with the small diameter receive-only surface coil showed a pronounced top-to-bottom gradient. To approximate the coil sensitivity profile, an image with minimum contrast between tissue types and the saline reference was acquired using the same spiral sequence as above: no tag or saturation pulses, TR 800 msec, TE 4.3 msec, three repetitions (the first two were discarded for approach to steady state). ASL signal images were acquired with a temporal resolution of 32 seconds (TR 4 seconds 8 interleaves) for TI 600 msec, and 35.2 seconds (TR 4.4 seconds 8 interleaves) for TI 1 second. The whole imaging protocol including M 0 reference and minimum contrast image took 15 minutes. Data Analysis Images were analyzed using AFNI software (24) and Matlab routines.

3 CBF Quantification Using ASL 857 Regions of interest (ROI) were drawn on anatomical images and transferred to the perfusion datasets. Statistical analyses were performed using SPSS Differences between the two gas conditions (air, 5% CO 2 ) were compared in different ROIs with a paired t-test. The effects of ROIs and the two gas conditions on CBF and dt values were also tested in a two-way analysis of variance (ANOVA) analysis. A P value 0.05 was considered significant. CBF Quantification and Fitting of Inflow Curves According to the general kinetic model for quantitative perfusion imaging with ASL (1), CBF (f) can be calculated from ASL data using Eq. [1]: M f2 T BGS M A TI dt exp TI/T 1, (1) where M is the difference in magnetization between tag and control images, T is the inversion efficiency of the tag pulse, BGS is the efficiency of the background suppression pulses, M A is the magnetization of arterial blood, dt is the arterial transit time and exp( TI/T1) describes the T1 decay of arterial blood at TI. Although the tag will decay with the T1 of tissue after exchanging with tissue water, it is likely to remain in the vascular space for the majority of our PASL experiment for shorter values of TI (25). We therefore make the simplifying assumption that the rate of decay of tagged blood magnetization occurs at the T1 of blood rather than the T1 of tissue. Equation [1] is based on the assumptions that clearance of tagged blood by venous flow during an ASL experiment is negligible (25). The inversion efficiency was found to have an average value above 0.98 in our experimental setting, and T was therefore set to 1. In Eq. [1], the factor 2 is based on the assumption that blood is fully relaxed as tagging is applied. By contrast, the PostSat included in our protocol resets the blood signal after each acquisition, so that M A M 0A 1 e TR TI /T1), where M 0A is the equilibrium magnetization of arterial blood. The period of TR-TI left for relaxation resulted in M A 0.95M 0A for our experiments. In common with other groups (22,26), attenuation of the ASL signal due to background suppression pulses was found to be about 20% in preliminary experiments, and BGS was therefore set to 0.8. From Eq. [1] it can be derived that with a minimum number of two ASL acquisitions at different TIs, CBF quantification can be achieved that is independent of transit time effects (1), if the first TI (TI1) is longer than the longest transit time of the tags and the second (TI2) is applied before the end of the tag bolus reaches the imaging slice and before outflow of the tag can occur: f M 1exp TI 1 /T 1 M 2 exp TI 2 /T 1 BGS 1.9M 0A TI 1 TI 2 This approach was termed double subtraction strategy by Buxton et al (1), because one subtraction is (2) applied between tag and control images, and the other between ASL signals measured at two TIs. Equation [2] essentially states that after correction for T1 decay, the slope of the curve between two ASL signals is proportional to CBF. Since it is difficult to measure M 0A directly, a tube filled with saline was attached to the rat head at each scan, and M 0A was derived from this measurement. The ratio between the proton density signal of arterial blood and saline at 3T was measured to be 0.71 by obtaining a slice covering the sagittal sinus in a human subject with a saline tube attached to the forehead. In addition, inhomogeneities in the coil sensitivity profile were accounted for by acquisition of a minimum contrast image, which was smoothed to approximate a coil sensitivity map and applied to both the saline reference image and ASL images before calculations. For quantitative CBF maps, two TIs were chosen according to the theory above (see results), that were appropriate under both gas conditions. For the analysis of inflow curves, the ASL signal acquired at different TIs was normalized to the magnetization of fully relaxed arterial blood. Data points were fitted according to Eq. [1]. Based on initial results, an additional fitting regime was applied to data with CBF 100 ml/100 g/minute, incorporating extraction fraction of water (E) and outflow time (T out, measured from the time blood enters a voxel) as variables: M 0 TI dt M f2 M 0A TI dt exp TI/T 1 A dt TI T out dt M A 1 1 E TI T out dt TI dt TI T out dt Constraints were applied to T out based on literature values: CBF 100 ml/100 g/minute: T out 0.15s (27). Parameters E and T out were chosen as those producing the lowest root-mean-square residual error between the experimental and fitted data. RESULTS ASL Signal Dynamics in the Rat Brain in Air and 5% CO 2 Inflow curves were generated from ASL acquisitions at multiple TIs, to analyze the temporal characteristics of the ASL signal under isoflurane anesthesia with air or 5% CO 2. Arterial blood gases were analyzed in six separate rats to determine the level of hypercapnia in our Table 1 Arterial Blood Gas Parameters in Rats Subjected to Air or 5% CO 2 in 1.5% Isoflurane Anesthesia ph pco 2 (mm Hg) po 2 (mm Hg) SO 2 (%) Air % CO a a a Denotes statistical significance in paired t-test. (3)

4 858 Figure 1. ASL signal maps acquired at variable inversion times (TI) from 0.3 to 3.0 seconds for normoxia (21% O2) and hypercapnia (5% CO2). The color bar depicts the ASL signal intensity of the images in arbitrary units. setup (Table 1). The pco2 was significantly higher ( mm Hg) and the ph lower when rats were breathing 5% CO2 in air. In Fig. 1, images of the ASL signal at different TIs are presented for the two experimental conditions. The ASL signal increased toward the one-second to two-second time points and declined at a TI of three seconds for all Wegener et al. datasets. Overall signal was larger in 5% CO2 as compared to air, in line with acceleration of blood flow to brain tissue with CO2. For the analysis of the ASL signal curves within different ROIs, data were categorized according to their CBF into 1) CBF 100 ml/100 g/minute ( low to normal ) and 2) CBF ⱖ 100 ml/100 g/minute ( normal to increased ). Typical signal courses for ROIs at these different CBF levels are shown in Fig. 2. Note that there is no sharp bend in the curve marking the end of the tag as can be observed in a more spatially confined tagging scheme (1). In our setup with the whole body coil, the FAIR tagging pulse tagged blood in the whole animal, so that there is no end to the tag. Typically, the measured signal was well modeled by Eq. [1] only if CBF was in the lower range ( 100 ml/100 g/minute). When blood flow was above 100 ml/100 g/minute (Fig. 2b), the ASL signal increased earlier, peaked earlier and showed a steeper decline afterward. Fitting of these data points was not successful with Eq. [1], and typically resulted in negative values for dt (see intersection of the dashed curve with the x-axis). The two mechanisms for decay of the ASL signal after tagged blood is delivered to the tissue are T1 decay and venous outflow. In the absence of outflow, the signal will decay with a time constant that is intermediate between that of blood and tissue, depending on the exchange of water between those two compartments. However, in our data with CBF 100 ml/100 g/minute, we measured a decay rate of approximately one second, which is shorter than the T1 of either blood or tissue at 3T. We therefore conclude that there must be venous outflow of tagged magnetization contributing to this decay. In an effort to better understand the ASL time course at later time points, we included the extraction fraction (E) and Tout as variables (see Eq. [3]). For most ASL approaches, it is assumed that all of the tagged blood water diffuses into the tissue, although it has been shown that even in normal flow states, E is less than one (28,29). We defined Tout as the time from arrival of the tag in a tissue voxel until its exit. When including Figure 2. ASL signal course from a dataset acquired with CBF 100 ml/100 g/minute (a) and one with CBF ⱖ 100 ml/100 g/minute (b). The ASL signal intensity (dm), normalized to fully relaxed blood magnetization (M0b), is plotted against inversion times (TI). Dashed line: fitting with Eq. [1]; solid line in (b): fitting with Eq. [3], including a term for extraction fraction (E) and outflow time (Tout). E and Tout were 0.6 and 900 msec in this example. The red dashed and dotted line in (b) indicates fitting of data points acquired at TI one second only using Eq. [1].

5 CBF Quantification Using ASL 859 minimize this effect for the calculation of arterial transit times (dt) and CBF by including only values obtained at TI one second. Arterial Transit Times in the Rat Brain in Air and 5% CO 2 Figure 3. Regions of interest for analysis of dt and CBF, superimposed on a T2-weighted image. Corresponding ROIs from both hemispheres were averaged. a: Back slice (IA 4.5 mm), 1: sensory/auditory cortex (Par1/Te1), 2: thalamus, 3: hippocampus (b) middle slice (IA 7.5 mm), 4: sensory cortex (Par1/Par2), 5: globus pallidus/caudate putamen (c) front slice (IA 10.5 mm), 6: sensory cortex (Par1), 7: caudate putamen. T out and E (with constraints derived from previously published data, see Materials and Methods), the fitting of ASL data points was considerably improved in higher flow conditions (Fig. 2b, solid line). An average decrease in E to and a T out of seconds were estimated for CBF 100 ml/100 g/minute for best fit of the data. Values for arterial transit times derived from this fitting approach closely matched the values obtained when using Eq. [1] and including only data points from TIs one second. While the outflow effects are of interest for the study of water exchange, they are not the focus of this study, and we chose to From the inflow data sets, arterial transit times were calculated for seven bilateral ROIs, defined on anatomical images (Fig. 3). The mean dt over all ROIs was msec in air and msec in 5% CO 2. In room air, the shortest dts were observed in the thalamus, and longest in the globus pallidus/caudate putamen (Fig. 4a). In all analyzed ROIs, dt was shortened in hypercapnia, which was expected since hypercapnia induces an increase in CBF. The degree to which hypercapnia influenced arterial transit times varied significantly between ROIs, the largest decrease being induced in the sensory/auditory cortex and globus pallidus/caudate putamen (56% and 60% decrease), and least in the thalamus and caudate putamen on slice 3 (21% and 22% decrease). In two-way ANOVA analysis, the effect of the gas condition (air vs. 5% CO 2 )ondt values was significant (P 0.001), while the effect of ROI was not (P 0.24). Arterial transit times were below 400 msec for air, and below 300 msec for 5% CO 2 for all ROIs considered. Aside from regional differences in transit delays between ROIs in these multislice data, note that blood travels a longer distance to distal than to proximal slices (2 mm slice thickness and 1 mm gap between slices), and dts typically increased from slice 1 to 3. CBF Changes in the Rat Brain in Air and 5% CO 2 Quantification of CBF with the double subtraction method using Eq. [2] requires ASL measurements at two different TIs. For the measurements not to be influenced by transit delay or outflow effects, we chose the TIs to be 600 msec (above the longest dt measured), and one second (before outflow effects were significant in high CBF data) for both experimental conditions. By measuring the ASL signal at shorter TIs, we intended to avoid two effects that will be more apparent with higher Figure 4. Mean calculated arterial transit times (dt in msec) (a) and CBF values (in ml/100 g/minute) (b) for seven ROIs under normoxia (black) and hypercapnia (gray). Error bars represent SD. * denotes significance (P 0.05) in paired t- test analysis. Sensory cortex (2) or (45) refers to the slice number.

6 860 Wegener et al. CBF: the outflow of tagged blood through the veins and the effect of water exchange between blood and tissue. CBF values (in ml/100 g/minute) were obtained within the same ROIs as for the dt analysis (Fig. 4b). Average CBF values across all ROIs were ml/100 g/minute for air and ml/100 g/minute for 5% CO 2. In all ROIs except for the hippocampus, hypercapnia resulted in a significant increase in CBF. The largest CBF increases were found in the globus pallidus/caudate putamen as well as the thalamus ( %), followed by moderate increases in the cortex ( %). The variance of CBF values within ROIs was found to be significant in two-way ANOVA (P 0.028), as was the effect of the gas condition (P 0.001). DISCUSSION The dynamic changes of the PASL signal with time observed in this study offer valuable insights into the blood flow response of the rat brain. At a given field strength, the signal will be influenced by flow velocity changes, routes of blood supply, vessel diameter, and extraction and distribution of tagged blood water in the brain. Using FAIR-ASL, we noted that in a baseline perfusion state, the ASL signal reached a maximum at about 1.5 seconds after an average dt of msec and declined thereafter. There was a striking difference, though, in the ASL signal dynamics when rats were breathing 5% CO 2. Arterial transit times were significantly shorter (128.6 msec on average), the maximum signal was reached earlier, and the signal decreased more rapidly than predicted by decay of the tag with the T1 of either blood (1.6 seconds) or gray matter (1.3 seconds). We therefore attribute this steep decline of the signal curve with hypercapnia to outflow of tagged water from the imaging slice. This was corroborated by the fact that fitting of all ASL data (acquired at TIs from 0.3 to three seconds) with a CBF 100 ml/100 g/minute was only successful if additional variables: 1) extraction fraction (E), and 2) outflow (T out ) time were included. Although there is a wide range of published values for E and T out (16,17,27), and the uncertainty in these estimates is high, it should be noted that in our study E is comparatively low and T out is quite large. Establishing more accurate estimates of these parameters would be an interesting goal for future studies. However, our data support previous observations that E is below one even in normal flow states in the rat, and that for establishing a pulsed ASL protocol to measure brain perfusion, both transit delay and outflow effects are relevant entities that should be incorporated into the CBF measurement. Of note, the shift of tagged water molecules from arterial to venous blood vessels containing more deoxygenated blood comes along with an increase in T2* effects, which might add to the observed ASL signal decrease at longer TIs. These contributions, however, are very small at the short echo times used in this study, and were calculated to be 4% at most; assuming all of the tagged blood shifted to the venous side at longer TIs. We found that the two most important sources of potential errors in CBF quantitation are transit delay effects and venous outflow, and these were both addressed using the double subtraction strategy (1). This approach uses two ASL acquisitions at different TIs, to subtract out dt effects and obtain a measurement that is dependent only on local perfusion. The timing of the measurements is critical: the earlier TI acquisition has to occur after the longest dt and the latter before the end of the tag. We set the first TI to 600 msec to be certain not to be biased by transit delays. Since we were using FAIR with a whole body transmit coil, the tagging pulse affected the whole animal, therefore, we did not have to consider an end of the tag in our setup. However, we found that outflow effects in fast flow states profoundly affected the calculation of CBF, even at relatively low TIs such as 1.5 seconds. To overcome this problem and at the same time avoid errors introduced by a change in the water extraction fraction with higher flow rates, we chose to set the longer TI to a time where outflow effects were not evident in our data, which was at a TI of one second. After these adjustments had been made, all MRI parameters could be kept constant for both normal and high flow conditions. ROI analysis revealed that, in line with a previous report using continuous ASL, CBF and dt are heterogeneous throughout the rat brain (14). The baseline CBF values (air) obtained using our PASL method were in good agreement with previous studies of CBF in isoflurane anesthetized rats using ASL (30,31) or autoradiography (32). In principle, dt should be proportional to 1/CBF in the absence of changes in arterial caliber, and a proportionality between the two parameters has been shown previously (15). In high flow states, there can be an uncoupling of the two measures: CBF can increase further without additional dt decreases (33). In our data, there was a clear shift toward lower mean dt values in the high flow state. However, we observed regional variations in transit delays that were not associated with the same degree of CBF change, which can be attributed to differences in capillary anatomy within the analyzed ROIs. The distribution of dt was rather broad when animals were breathing air (Fig. 5) and slightly narrower in the hypercapnic condition. It seems, therefore, that there is a tendency for regional differences in arterial transit times to be less distinct in high flow states. This phenomenon could be attributed to diminished pulsations and flow reversals in the capillaries when blood flow is increased. With distinct local responses to flow changes, the distribution of dt could be a valuable parameter of vascular physiology. Reports about regional analyses of the hemodynamic response to hypercapnia in the rat are sparse (12,34). It can be expected to be nonuniform across the brain, since baseline CBF and metabolism vary in different brain regions (35,36). The issue is complicated by the fact that anesthesia itself exerts a region dependent flow change, as has been demonstrated for isoflurane (31). With hypercapnia, the average increase from normocapnia CBF values was 80% in our data. Mean dt values decreased by 39%. Previous studies have reported tran-

7 CBF Quantification Using ASL 861 Figure 5. Individual CBF values (in ml/100 g/minute) for all ROIs in all five rats plotted against the arterial transit times (dt in msec). On the left, the distribution of values is shown in a histogram for air (black) and 5% CO 2 (gray). sit delay decreases by 13% with a pco 2 of 70 mmhg in tissue voxels (15), and 48% with a pco 2 of 86 mmhg (37). In a study by Sicard et al. (31), administration of 5% CO 2 to spontaneously breathing, isoflurane-anesthetized animals resulted in a 25% increase of CBF with an average pco 2 of 50 mm/hg. A more pronounced increase in CBF was found in other studies with isoflurane (15), halothane (37,38), and -chloralose (12). With a mean pco 2 of 54.1 mm/hg, our data are consistent with a 5% CBF change per mmhg CO 2 increase with invasive methods (39). However, the induced dt changes were rather large in our study. These differences may be explained by the fact that with the smaller gap between tag and imaging slab, smaller diameter vessels were assessed that might show a more pronounced fractional increase in CBF with hypercapnia (40). The thalamus and the globus pallidus/caudate putamen showed a relatively large CBF increase with 5% CO 2, cortical areas displayed a large to moderate CBF change, while in the hippocampus, CBF was only very slightly and nonsignificantly increased. The hippocampus, especially the cornu ammonis area 1 (CA1) subregion, is known for its susceptibility to ischemia (41) and low capillary density (42), with a low basal blood flow rate (14). The attenuated increase in hippocampal CBF in response to hypercapnia is in line with previous observations using BOLD MRI (34), and might be an indicator for a lower basal vasoreactivity as one possible factor determining higher susceptibility to ischemic damage (43). In conclusion, CBF quantification with ASL techniques is vulnerable to transit delay and outflow effects, especially when blood flow is elevated. Cerebrovascular disease, such as arteriosclerosis and stroke, and even modulations in blood flow to the healthy brain, such as in functional activation, can complicate accurate CBF measurements. In normal as well as increased flow states, we observed a large regional variability in arterial transit times (dt) in the rat brain. In addition, in high flow states, ASL signal curves strongly suggested significant outflow of tagged blood, which might lead to an underestimation of CBF. We developed a pulsed ASL protocol that allows CBF quantification in rodents that is insensitive to arterial transit time or outflow effects. Using this method, we found local variations in CBF over the rat brain, as well as distinct responses to a hypercapnic challenge. Accurate and locally specific CBF analyses will facilitate the understanding of how blood flow is maintained under physiological and pathological conditions. ACKNOWLEDGMENTS We thank Harrieth Wagner, Department of Physiology, University of California San Diego (UCSD), for analysis of the arterial blood gas samples. REFERENCES 1. Buxton RB, Frank LR, Wong EC, Siewert B, Warach S, Edelman RR. A general kinetic model for quantitative perfusion imaging with arterial spin labeling. Magn Reson Med 1998;40: Detre JA, Zhang W, Roberts DA, et al. Tissue specific perfusion imaging using arterial spin labeling. NMR Biomed 1994;7: Calamante F, Thomas DL, Pell GS, Wiersma J, Turner R. Measuring cerebral blood flow using magnetic resonance imaging techniques. J Cereb Blood Flow Metab 1999;19: Barbier EL, Lamalle L, Decorps M. Methodology of brain perfusion imaging. J Magn Reson Imaging 2001;13: Ewing JR, Wei L, Knight RA, et al. Direct comparison of local cerebral blood flow rates measured by MRI arterial spin-tagging and quantitative autoradiography in a rat model of experimental cerebral ischemia. J Cereb Blood Flow Metab 2003;23: Walsh EG, Minematsu K, Leppo J, Moore SC. Radioactive microsphere validation of a volume localized continuous saturation perfusion measurement. Magn Reson Med 1994;31: Ewing JR, Cao Y, Knight RA, Fenstermacher JD. Arterial spin labeling: validity testing and comparison studies. J Magn Reson Imaging 2005;22: Detre JA, Alsop DC, Vives LR, Maccotta L, Teener JW, Raps EC. Noninvasive MRI evaluation of cerebral blood flow in cerebrovascular disease. Neurology 1998;50: Alsop DC, Detre JA. Reduced transit-time sensitivity in noninvasive magnetic resonance imaging of human cerebral blood flow. J Cereb Blood Flow Metab 1996;16: Wong EC, Buxton RB, Frank LR. Quantitative imaging of perfusion using a single subtraction (QUIPSS and QUIPSS II). Magn Reson Med 1998;39:

8 862 Wegener et al. 11. Schepers J, Veldhuis WB, Pauw RJ, et al. Comparison of FAIR perfusion kinetics with DSC-MRI and functional histology in a model of transient ischemia. Magn Reson Med 2004;51: Tsekos NV, Zhang F, Merkle H, Nagayama M, Iadecola C, Kim SG. Quantitative measurements of cerebral blood flow in rats using the FAIR technique: correlation with previous iodoantipyrine autoradiographic studies. Magn Reson Med 1998;39: Gonzalez-At JB, Alsop DC, Detre JA. Cerebral perfusion and arterial transit time changes during task activation determined with continuous arterial spin labeling. Magn Reson Med 2000;43: Thomas DL, Lythgoe MF, van der Weerd L, Ordidge RJ, Gadian DG. Regional variation of cerebral blood flow and arterial transit time in the normal and hypoperfused rat brain measured using continuous arterial spin labeling MRI. J Cereb Blood Flow Metab 2006;26: Barbier EL, Silva AC, Kim SG, Koretsky AP. Perfusion imaging using dynamic arterial spin labeling (DASL). Magn Reson Med 2001;45: Silva AC, Zhang W, Williams DS, Koretsky AP. Estimation of water extraction fractions in rat brain using magnetic resonance measurement of perfusion with arterial spin labeling. Magn Reson Med 1997;37: Zaharchuk G, Bogdanov AA Jr, Marota JJ, et al. Continuous assessment of perfusion by tagging including volume and water extraction (CAPTIVE): a steady-state contrast agent technique for measuring blood flow, relative blood volume fraction, and the water extraction fraction. Magn Reson Med 1998;40: Davis TL, Kwong KK, Weisskoff RM, Rosen BR. Calibrated functional MRI: mapping the dynamics of oxidative metabolism. Proc Natl Acad Sci USA 1998;95: Kwong KK, Belliveau JW, Chesler DA, et al. Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc Natl Acad Sci USA 1992;89: Kim SG. Quantification of relative cerebral blood flow change by flow-sensitive alternating inversion recovery (FAIR) technique: application to functional mapping. Magn Reson Med 1995;34: Kim SG, Tsekos NV. Perfusion imaging by a flow-sensitive alternating inversion recovery (FAIR) technique: application to functional brain imaging. Magn Reson Med 1997;37: Garcia DM, Duhamel G, Alsop DC. Efficiency of inversion pulses for background suppressed arterial spin labeling. Magn Reson Med 2005;54: Mai VM, Hagspiel KD, Christopher JM, et al. Perfusion imaging of the human lung using flow-sensitive alternating inversion recovery with an extra radiofrequency pulse (FAIRER). Magn Reson Imaging 1999;17: Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res 1996;29: Buxton RB. Quantifying CBF with arterial spin labeling. J Magn Reson Imaging 2005;22: Ye FQ, Frank JA, Weinberger DR, McLaughlin AC. Noise reduction in 3D perfusion imaging by attenuating the static signal in arterial spin tagging (ASSIST). Magn Reson Med 2000;44: Ewing JR, Cao Y, Fenstermacher J. Single-coil arterial spin-tagging for estimating cerebral blood flow as viewed from the capillary: relative contributions of intra- and extravascular signal. Magn Reson Med 2001;46: Eichling JO, Raichle ME, Grubb RL Jr, Ter-Pogossian MM. Evidence of the limitations of water as a freely diffusible tracer in brain of the rhesus monkey. Circ Res 1974;35: Go KG, Lammertsma AA, Paans AM, Vaalburg W, Woldring MG. Extraction of water labeled with oxygen 15 during single-capillary transit. Influence of blood pressure, osmolarity, and blood-brain barrier damage. Arch Neurol 1981;38: Liu ZM, Schmidt KF, Sicard KM, Duong TQ. Imaging oxygen consumption in forepaw somatosensory stimulation in rats under isoflurane anesthesia. Magn Reson Med 2004;52: Sicard K, Shen Q, Brevard ME, et al. Regional cerebral blood flow and BOLD responses in conscious and anesthetized rats under basal and hypercapnic conditions: implications for functional MRI studies. J Cereb Blood Flow Metab 2003;23: Young WL, Barkai AI, Prohovnik I, Nelson H, Durkin M. Effect of PaCO 2 on cerebral blood flow distribution during halothane compared with isoflurane anaesthesia in the rat. Br J Anaesth 1991; 67: Barbier EL, Silva AC, Kim HJ, Williams DS, Koretsky AP. Perfusion analysis using dynamic arterial spin labeling (DASL). Magn Reson Med 1999;41: Kannurpatti SS, Biswal BB, Hudetz AG. Regional dynamics of the fmri-bold signal response to hypoxia-hypercapnia in the rat brain. J Magn Reson Imaging 2003;17: Nehls DG, Park CK, MacCormack AG, McCulloch J. The effects of N-methyl-D-aspartate receptor blockade with MK-801 upon the relationship between cerebral blood flow and glucose utilisation. Brain Res 1990;511: Shimoji K, Ravasi L, Schmidt K, et al. Measurement of cerebral glucose metabolic rates in the anesthetized rat by dynamic scanning with 18F-FDG, the ATLAS small animal PET scanner, and arterial blood sampling. J Nucl Med 2004;45: Zhang W, Williams DS, Detre JA, Koretsky AP. Measurement of brain perfusion by volume-localized NMR spectroscopy using inversion of arterial water spins: accounting for transit time and cross-relaxation. Magn Reson Med 1992;25: Williams DS, Detre JA, Leigh JS, Koretsky AP. Magnetic resonance imaging of perfusion using spin inversion of arterial water. Proc Natl Acad Sci USA 1992;89: Morii S, Ngai AC, Ko KR, Winn HR. A venous outflow method for continuously monitoring cerebral blood flow in the rat. Am J Physiol 1986;250(Pt 2):H304 H Yoder E. Brain03 and BrainPET03. Presented at: XXIst International Symposium on Cerebral Blood Flow, Metabolism, and Function and the VI: International Conference on Quantification of Brain Function With PET; June 29 July 3, 2003; Calgary, Alberta, Canada. 41. Kirino T, Tamura A, Sano K. Selective vulnerability of the hippocampus to ischemia reversible and irreversible types of ischemic cell damage. Prog Brain Res 1985;63: Cavaglia M, Dombrowski SM, Drazba J, Vasanji A, Bokesch PM, Janigro D. Regional variation in brain capillary density and vascular response to ischemia. Brain Res 2001;910: Olah L, Franke C, Schwindt W, Hoehn M. CO(2) reactivity measured by perfusion MRI during transient focal cerebral ischemia in rats. Stroke 2000;31: