Quantification of blood flow and volume in arterioles and venules of the rat cerebral cortex using functional micro-ultrasound

NeuroImage 63 (2012) 1030–1037

Contents lists available at SciVerse ScienceDirect

NeuroImage journal homepage: www.elsevier.com/locate/ynimg

Quantification of blood flow and volume in arterioles and venules of the rat cerebral cortex using functional micro-ultrasound☆ Martijn E. van Raaij, Liis Lindvere, Adrienne Dorr, Jianfei He, Bhupinder Sahota, F. Stuart Foster, Bojana Stefanovic ⁎ Imaging Research, Sunnybrook Research Institute, Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, Toronto, ON, Canada M4N 3M5

a r t i c l e

i n f o

Article history: Accepted 23 July 2012 Available online 31 July 2012 Keywords: Functional micro-ultrasound imaging (fMUS) Cerebral blood volume (CBV) Cerebral blood flow (CBF) Functional neuroimaging Cerebral hemodynamics Rat

a b s t r a c t Relative cerebral blood volume (rCBV), relative cerebral blood flow (rCBF), and blood flow speed are key parameters that characterize cerebral hemodynamics. We used contrast-enhanced functional micro-ultrasound (fMUS) imaging employing a disruption–replenishment imaging sequence to quantify these hemodynamic parameters in the anesthetized rat brain. The method has a spatial resolution of about 100 μm in-plane and around 600 μm through-plane, which is comparable to fMRI, and it has a superior temporal resolution of 40 ms per frame. We found no significant difference in rCBV of cortical and subcortical gray matter (0.89 ± 0.08 and 0.61 ± 0.09 times the brain-average value, respectively). The rCBV was significantly higher in the vascular regions on the pial surface (3.89 ± 0.71) and in the area of major vessels in the subcortical gray matter (2.02 ± 0.31). Parametric images of rCBV, rCBF, and blood flow speed demonstrate spatial heterogeneity of these parameters on the 100 μm scale. Segmentation of the cortex in arteriolar and venular-dominated regions identified through color Doppler imaging showed that rCBV is higher and flow speed is lower in venules than in arterioles. Finally, we show that the dependence of rCBV on rCBF was significantly different in cortical versus subcortical gray matter: the exponent α in the power law relation rCBV = s · rCBF α was 0.37 ± 0.13 in cortical and 0.75 ± 0.16 in subcortical gray matter. This work demonstrates that functional micro-ultrasound imaging affords quantification of hemodynamic parameters in the anesthetized rodent brain. This modality is a promising tool for neuroscientists studying these parameters in rodent models of diseases with a cerebrovascular component, such as stroke, neurodegeneration, and venous collagenosis. It is of particular import for studying conditions that selectively affect arteriolar versus venular compartments. © 2012 Elsevier Inc. All rights reserved.

Introduction The brain depends on an adequate supply of oxygen and glucose by the circulation. Changes in neuronal activity induce changes in the local blood supply, a phenomenon known as neurovascular coupling (Attwell et al., 2010). Neurovascular coupling is compromised in many neurological and neurodegenerative disorders (D'Esposito et al., 2003; Iadecola, 2004). Most imaging methods for studying of brain function rely on neurovascular coupling: hemodynamic parameters such as cerebral blood volume (CBV, in mL of blood per 100 g of tissue) and cerebral blood flow (CBF, in mL of blood per 100 g of tissue per minute) are used to infer the location(s) of neuronal activation, or identify regions

Abbreviations: fMUS, functional micro-ultrasound imaging; rCBV, relative cerebral blood volume; rCBF, relative cerebral blood flow. ☆ Disclosure: F. Stuart Foster is a consultant to VisualSonics Inc., Toronto, Canada. ⁎ Corresponding author at: Imaging Research, Sunnybrook Research Institute, S-640, 2075 Bayview Avenue, Toronto, ON, Canada M4N 3M5. Fax: +1 416 480 5714. E-mail address: [email protected] (B. Stefanovic). 1053-8119/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.neuroimage.2012.07.054

with an abnormal blood supply. While the mechanisms of neurovascular coupling are being actively studied (Iadecola and Nedergaard, 2007), there is an active research in the development and optimization of methodologies to study regional differences in neurovascular coupling across the whole brain (Sloan et al., 2010). Current techniques for in vivo imaging of brain blood flow include MRI, CT, PET, and optical imaging such as laser speckle contrast imaging and two photon fluorescence microscopy. Despite their respective strengths (Calamante et al., 1999; Coles, 2006; Devor et al., 2012), these methods are challenged to provide concomitant imaging of blood flow and volume at high temporal resolution while still allowing whole brain coverage. Functional micro-ultrasound imaging (fMUS, (van Raaij et al., 2011)) is a recent adaptation of high-frequency ultrasound imaging (Foster et al., 2002) that allows quantitative measurement of changes in CBV and CBF. The contrast mechanism used to measure hemodynamic parameters may be either power Doppler imaging of moving red blood cells (Mace et al., 2011) or nonlinear imaging of microbubble contrast agents (van Raaij et al., 2011). The term ‘functional’ is used here in the wider sense of measuring hemodynamic parameters of a steady-state situation, though it can also be used in

M.E. van Raaij et al. / NeuroImage 63 (2012) 1030–1037

a narrower definition as referring to a hemodynamic response to a stimulus. An important strength of high frequency ultrasound imaging is the availability of multiple contrast mechanisms. In addition to standard backscattered ultrasound images (B-mode), there are Doppler modes which measure the presence and speed of moving particles (e.g., red blood cells) in the organ of interest. The development of ultrasound contrast agents (Becher and Burns, 2000; Ferrara et al., 2007) led to another highly sensitive vascular contrast mechanism. The microbubbles are gas-filled lipid-shell particles with a diameter of 1 to 10 μm which are strictly intravascular, biologically safe (Mulvagh et al., 2008), and detectable using nonlinear imaging techniques that are very sensitive to the microbubbles while being minimally affected by signal from tissue (Powers et al., 2009). The present work quantifies changes in CBF and CBV using microbubbles via the so-called disruption–replenishment (also: destruction–reperfusion) technique. Most of the microbubbles in the imaging plane are disrupted by a high-intensity burst of ultrasound. After the burst pulse, microbubbles from surrounding vessels gradually replenish the imaging plane. The kinetics of this replenishment process confers information on hemodynamic parameters. (The perfluorocarbon gas that is released upon disruption as shell-less bubbles circulates briefly and freely through the vasculature, eventually dissolves in the plasma, and is released by the lungs.) We use a commercially available micro-ultrasound system to image these parameters with 100 μm spatial and 40 ms temporal resolution (van Raaij et al., 2011). Additionally, we investigate hemodynamic differences in cortical arterioles and venules using color Doppler imaging, which detects whether flow is towards or away from the ultrasound transducer. We thus establish the use of disruption–replenishment contrast-enhanced high-frequency ultrasound imaging and color Doppler imaging in the quantification and characterization of the spatial heterogeneity of hemodynamic parameters in the rat brain. Methods Animal preparation Male Sprague–Dawley rats (Charles River Laboratories Inc., Saint-Constant, Quebec, Canada) (N = 13) were anesthetized with isoflurane, tracheotomized, mechanically ventilated and placed on a heating pad as described in Lindvere et al. (2010) and van Raaij et al. (2011). The femoral artery, femoral vein and tail vein were cannulated for blood gas sampling/blood pressure monitoring, delivery of α-chloralose during imaging, and delivery of ultrasound contrast agent respectively. We monitored the breathing rate, heart rate, arterial blood oxygen saturation, body temperature, and blood gases, and made adjustments when required to ensure a normal physiological state during both the surgery and the data acquisition. A cranial window (6.5 mm × 3 mm) was opened over the forelimb representation of the primary somatosensory cortex (S1FL), leaving the dura intact. The cranial window was covered with agarose to prevent dehydration of the cortical surface. All experimental protocols reported on in this study have been approved by the Animal Care Committee of Sunnybrook Health Sciences Centre. Functional ultrasound image acquisition We used a Vevo 2100 microultrasound system (VisualSonics Inc., Toronto, ON, Canada) with a linear array transducer with a center frequency of 21 MHz (MS‐250, VisualSonics) as described in detail in (van Raaij et al., 2011). After acquisition of B-mode and color Doppler images, a contrast agent (Vevo MicroMarker, untargeted, VisualSonics Inc.) was infused at a concentration of 6 · 108 bubbles/mL using an infusion pump (New Era Pump Systems Inc., NE-1000) at a constant rate of 40 μl/min for 7 min. After a 2.5 minute stabilization period, imaging for

1031

the disruption–replenishment measurement was performed in a ‘nonlinear contrast’ (NLC) mode. This mode employs an amplitude modulation-type ultrasound pulse sequence designed to isolate backscattered ultrasound from nonlinear scatterers (i.e., microbubbles) and reject signal from primarily linearly scattering media like tissue or red blood cells (Goertz et al., 2005; Needles et al., 2010). Furthermore, the nonlinear signal has been shown to originate from microbubbles in the 1.8 μm range, very close to the volume peak of the microbubble distribution (Sprague et al., 2010). Moreover, for the low concentrations of bubbles employed here, the signal intensity in this imaging mode is linearly dependent on bubble concentration. Therefore, the signal is effectively a relative measure of local plasma volume (Lampaskis and Averkiou, 2010), and, assuming constant hematocrit (Herman et al., 2009), of blood volume. Using the standard definition, the mechanical index was about 0.4 (van Raaij et al., 2011). The disruption pulse was a 10 cycle pulse at max power, with peak derated rarefractional pressure of 3.7 MPa, played out at the system's rate of 21 MHz, with transmit focus at 11 mm. As in prior experiments, we have found no evidence for extravasation of bubbles in these experiments. Up to five repetitions of the disruption–replenishment sequence were performed per 90-second acquisition (“cineloop”), and three cineloops were recorded within the 7-minute infusion, yielding a maximum of 15 replenishment curves per subject. The field of view was 8 mm wide and 10 mm deep measured from the surface of the cortex. For the near-sagittal sections imaged here, the slice thickness is the extent in the lateral-to-medial direction. The slice thickness depth profile is hourglass-shaped by the acoustic lens on the transducer. The slice thickness is greatest (~ 2 mm) at the cortical surface and reduces to ~ 600 μm at a depth of 10 mm. The axial resolution is proportional to the physical length of the pulse at half maximum, while the lateral resolution is proportional to wavelength and f-number (focal length/diameter). The pixel resolution in the images is rather arbitrary: the system oversamples the physical resolution by a factor of about 4 axially (top-to-bottom in the current images) and about 2 laterally (left-to-right). “Ultrasound angiograms” were created by time-averaging the replenishment cineloops. Disruption–replenishment analysis and model fitting The average replenishment cineloop for each subject was binned in 4 by 4 pixel bins to improve signal-to-noise ratio and reduce computation time, and a mono-exponential replenishment model was fitted to each bin: y(t) = A ⋅ (1 − exp(− β ⋅ t)), where y is the signal intensity in each voxel, A is the equilibrium amplitude and β is a rate constant (Wei et al., 1998). The equilibrium signal amplitude A is taken to be a measure of CBV; the rate constant β is taken to be proportional to the blood inflow speed; and the product Aβ (the slope of the tangent to the replenishment curve at the start of the replenishment process) is a measure of blood flow (Wei et al., 1998). To reduce the influence of subject-to-subject variability in signal intensity, CBV and CBF were normalized to the “whole-slice” average CBV and CBF values; the resulting ratios are labeled rCBV and rCBF (r for relative) in the Results section. The parametric maps were then upsampled to the original image resolution. Two parenchymal (cortical gray matter, subcortical gray matter) and two primarily vascular (pial vessels, deep vessels) anatomical regions of interest were manually drawn on the ultrasound angiograms. The mono-exponential model was applied to region-average replenishment curves to obtain region-specific hemodynamic parameters. For every subject, color Doppler images were aligned to the nonlinear contrast images by cross-correlating the corresponding B-mode images allowing for in-plane translations. In cases where the cross-correlation algorithm did not produce satisfactory results, the alignment was adjusted manually. The aligned color Doppler images were used to identify regions dominated by arteriolar flow (blood flowing perpendicularly

1032

M.E. van Raaij et al. / NeuroImage 63 (2012) 1030–1037

down from the cortical surface) versus venular flow (blood flowing perpendicularly up towards the cortical surface) in the cortex, and the model was fitted to those regions as well. Statistical analysis The effect of the region of interest on the mean rCBV, rCBF, and rate constant estimates was assessed with a linear mixed-effects model treating the subject as the random effect and region of interest as the fixed effect. For the arteriolar and venular regions, significance was assessed with a two-sample paired t-test. Shapiro–Wilk tests confirmed normality of the data distribution in each group. When a difference between two values is stated to be significant, p b 0.05. Results Spatial mapping of rCBV and rCBF Fig. 1 shows an ‘ultrasound angiogram’ obtained by averaging a microbubble-based nonlinear contrast image over time, with a number of typical individual voxel microbubble replenishment curves. The microbubbles are strictly intravascular in the presence of an intact blood–brain barrier. Signal from bubbles in capillaries is detected, but not resolved: it adds to the signal level in a given voxel. The discrete nature of the microbubble contrast agent used to generate the images causes a strongly spiked appearance of the individual voxel time-intensity curves. Even though the traces shown in Fig. 1 result from 4 by 4 voxel bins, the traces remain quite noisy, and the mono-exponential model fit errors may be large as a result. A volume of about 25 μl is needed to produce reasonably good fits to the mono-exponential model. Parametric maps of relative CBV, blood inflow speed, and relative CBF for the same subject reveal the spatial heterogeneity of these parameters throughout the depth of the rodent brain (Fig. 2). The rCBV map (Fig. 2B) reveals a higher blood volume in the pial and subcortical gray regions than in the cortical region. The flow speed map (Fig. 2C) indicates a higher rate constant of replenishment (i.e., a faster flow speed of the blood) in the subcortical gray matter, and slowest refilling in the pial vessels near the cortical surface. The

rCBF map (Fig. 2D) shows that blood flow is higher in the subcortical gray matter than in the cortex. The signal-to-noise ratio of the time-intensity traces and the accuracy of the replenishment model fits is improved dramatically when larger anatomical regions (of consistent tissue type) are defined (Fig. 3). As a side note, for the region-averaged time-intensity traces of the pial region, in some subjects the replenishment process appears to have two time constants, which is not adequately captured by the mono-exponential fit (see the ‘pial’ trace in Fig. 3). Regional variations in rCBV and rCBF In the discussion below, statistical significance is defined as p b 0.0125 (upon correction for multiple comparisons given four regions of interest). Disruption–replenishment measurements in 12 rats (Fig. 4) showed that rCBV is highest in the pial region (3.89 ± 0.71 times the brain-average CBV; uncertainties quoted are the 95% confidence intervals on the mean). Cortical and subcortical gray rCBV values are not significantly different (0.89 ± 0.08 and 0.61 ± 0.09 respectively, p = 0.43), while the major vessel rCBV was 2.02 ± 0.31 times the brain-average. Variability between rats in the pial region was greater than in the other regions. The inflow speed as measured by the rate constant was lower for the vascular regions (pial and major vessel: 0.50 ± 0.07 s −1 and 0.80 ± 0.11 s −1, respectively) than for the cortical and subcortical parenchymal regions (0.92 ± 0.13 s −1 and 1.27 ± 0.23 s −1). Despite slower inflow, normalized rCBF was highest in the vascular regions (pial region 2.36 ± 0.57, deep vessels 1.85 ± 0.20 times brain-average CBF), and lower in cortical (0.99 ± 0.17) and subcortical gray regions (0.91 ± 0.15). The horizontal bars under the scatter plots in Fig. 4 indicate that the linked groups were statistically significantly different. Arteriovenular segmentation based on color Doppler images Overlays of color Doppler-based arteriovenular segmentation maps on the anatomical maps indicate which regions of the brain are primarily irrigated by venules versus arterioles. The sample maps shown in

Fig. 1. Ultrasound angiogram showing the vascular morphology throughout the entire depth of rat brain in a near-sagittal imaging plane (superior is to the top of the image, anterior to the right). The time-intensity curves are the replenishment curves (thin lines) with model fits (thick lines) for the indicated voxels (white dots). The vertical scale bars all show the same number of arbitrary intensity units. All traces are 10 s long.

M.E. van Raaij et al. / NeuroImage 63 (2012) 1030–1037

1033

Fig. 2. Parametric maps of hemodynamic parameters in the rat brain. Ultrasound angiogram of Fig. 1 (A) with the corresponding parametric maps of relative cerebral blood volume (B), the flow rate as measured by the rate constant β in s−1 (C) and the relative cerebral blood flow (D). The parametric maps have been upsampled to the original image resolution and values larger than 5× the whole-slice average (for rCBV and rCBF) or 5 s−1 (for flow rate) have been clipped for clarity of presentation. The scale bar in (B) applies to all three maps.

Fig. 5 show that while the resolution of the color Doppler images is lower than that of the nonlinear contrast-based anatomical image, it is clear which of the larger vessels are arterioles and which are venules. The sign of the color Doppler signal indicates whether flow occurs towards or away from the transducer. In the cortex (outlined in white in Fig. 5), most vessels lie in the axial direction of the ultrasound beam and thus can be identified unambiguously as arterioles (colored red in Fig. 5) and venules (blue). Note that this coloring scheme follows the biological convention of false-coloring arteries red and venules blue, and that this convention is the opposite of the color scheme commonly used in ultrasound imaging where flow away from the transducer is colored blue. Many acquisition parameters influence the color Doppler image. We note here that at the arrow in Fig. 5A, a vessel appears to change flow direction: this is an aliasing artifact due to the low pulse repetition frequency (PRF) used for this acquisition that allows good discrimination of slow flow but limits the maximum flow speed that can be measured. The impact of other acquisition parameters will be considered in the Discussion section. Hemodynamic parameters in cortical arterioles and venules

rCBV) pairs for each subject in the cortical gray, subcortical gray, cortical arterioles, and cortical venules regions (Fig. 7). Power-law regression lines after Grubb et al. (1974) were fitted using the Origin (OriginLab, Northampton, MA) nonlinear least-squares curve fitting tool employing the Levenberg–Marquardt algorithm (resulting parameters are summarized in Table 1). The model equation was rCBV = s · rCBF α, where s is the scaling factor and α is the exponent; both parameters were unconstrained and initialized to 1, and the data points were unweighted. Note that we did not systematically vary rCBF to obtain these data: the rCBF domains shown in Fig. 7 represent the range of normalized CBF values present in our study population. Also note that these regressions should be interpreted with caution since the standard assumption in nonlinear regression that X and Y variables are independent is not met in this case. The arteriolar and venular regression lines were found to be statistically significantly different according to an F-ratio test comparing a power law fit to the combined arteriole and venule datasets to the fits from the separate datasets (F = 5.1, p = 0.016) (Motulsky and Christopoulos, 2003, chapter 27). The same applies to the cortical versus subcortical gray regressions (F = 23.7, p = 5e − 6). Note that the scatter present

As shown in Fig. 6, for all subjects, rCBV was higher in cortical regions dominated by venules than in regions primarily supplied by arterioles (1.10 ± 0.10 versus 0.90 ± 0.11 times the brain-average CBV respectively). The rate constant was higher in the arterioles (0.95 ± 0.13 s −1) than in the venules (0.85 ± 0.13 s −1). Cerebral blood flow was significantly higher in venules than in arterioles, at 1.12 ± 0.16 versus 1.02 ± 0.20 (Fig. 6). Power-law relation between rCBF and rCBV To assess the relation between normalized cerebral blood flow and cerebral blood volume in our population, we plotted the (rCBF,

Fig. 3. Anatomical region of interest definition (left) and region-average replenishment curves (right) for a sample subject. The CBV (as represented by the plateau intensity) is lowest in the cortex; slightly higher in the subcortical gray parenchymal region, and highest in the two vascular regions. Solid lines, region-average signal intensity; dashed lines, mono-exponential fit to the data. The R2 values for the fits were: 0.97 (pial vessels), 0.84 (major vessels), 0.99 (subcortical gray), and 0.99 (cortical gray).

Fig. 4. rCBV, flow rate, and rCBF for anatomically defined regions of interest in 12 rats. The red cross indicates the mean; and the red bars, the 95% confidence interval on the mean. The green horizontal lines below the plot indicate that the linked groups are statistically significantly different.

1034

M.E. van Raaij et al. / NeuroImage 63 (2012) 1030–1037

Fig. 5. Arteriovenular segmentation based on color Doppler ultrasound imaging for three representative subjects. Color Doppler images have been ‘binarized’ so as to show only direction of flow, not speed. In the cortical regions (white outlines) most of the flow is in the axial direction of the ultrasound beam (top-to-bottom in these images) and the vessel type can thus be assigned without ambiguity (see the text). Regions primarily irrigated by arterioles are colored red (flow down into the brain); venular regions are colored blue (flow towards the cortical surface). The arrow in (A) indicates an aliasing artifact as described in the text.

in the cortical venular data results in a very low R 2 value for the fit in that region. Discussion This work shows that high-frequency ultrasound imaging allows for real-time quantification of hemodynamic parameters at the 100 μm spatial scale, throughout the rat brain, as well as differentiation between cortical penetrating arterioles and venules. Compared to other modalities that can image blood flow in the living brain, fMUS has a spatial resolution comparable to fMRI and higher than CT and PET, and has a higher frame rate. Two-photon fluorescence microscopy has higher spatial resolution but a much lower penetration depth (of approximately 1 mm). An obvious disadvantage of fMUS is the intrinsic two-dimensionality of the images: 3D imaging can be achieved by translating the transducer in the third dimension, but at the expense of temporal resolution. Also, like 2PM but unlike MRI and PET, fMUS requires the opening of a cranial window to have sufficient signal for high-resolution images. fMUS gives relative values for CBV and CBF, like most other imaging modalities. The main findings reported in this study are (1) that steady state rCBV, rCBF, and flow speed vary across the rat brain, (2) that rCBV and rCBF differ in cortical regions dominated by arteriolar versus venular flow, and (3) that the steady-state power-law relationship

Fig. 6. Hemodynamic parameters in cortical arterioles and cortical venules. The red cross indicates the mean; and the red bars, the 95% confidence interval on the mean. The gray lines link the corresponding measurements in each subject. Differences between arterioles and venules are statistically significant for each parameter.

between rCBV and rCBF differs in subcortical versus cortical gray matter in the rat brain. These three findings are discussed in the paragraphs below. Steady state hemodynamic parameters vary across the brain In the MRI study by Kim and Kim (2005), it was found that arterial CBV in cortex and caudate putamen in rats were 1.1±0.5 and 1.3± 0.6 mL/100 g (via MOTIVE MRI) and 1.0±0.3 mL/100 g for both regions (via ASL MRI). A study using synchrotron-based quantitative CT found CBV in the parietal cortex and in the caudate putamen to be 2.1± 0.38 mL/100 g and 1.92±0.32 mL/100 g respectively (Adam et al., 2003). Our relative CBV values (not specific to arteries, but including both arterial and venous CBV in the region) for cortical and subcortical gray matter were 0.89±0.08 and 0.61±0.09, and were not significantly

Fig. 7. Dependence of relative CBV on relative CBF. Each data point represents one subject (N=12). Plots show rCBV–rCBF dependence in cortical gray matter (circles), subcortical gray matter (triangles), cortical arterioles, and cortical venules as labeled; regions are defined as shown in Figs. 3 and 5. Lines represent power-law fits as described in the text; shaded areas represent 95% confidence intervals on the regressions.

M.E. van Raaij et al. / NeuroImage 63 (2012) 1030–1037 Table 1 Parameters of power law regression curves to the regional rCBV versus rCBF using the equation rCBV=s·rCBFα, as shown in Fig. 7. Values are given as parameter estimate±standard error. Region of interest

s

α

R2

Cortical gray matter Subcortical gray matter Cortical arterioles Cortical venules

0.91 ± 0.03 0.66 ± 0.02 0.89 ± 0.03 1.07 ± 0.04

0.37 ± 0.13 0.75 ± 0.16 0.46 ± 0.13 0.33 ± 0.17

0.45 0.75 0.56 0.29

different. We did note large differences in CBV between gray matter tissue regions and primarily vascular regions, as did (Kim and Kim, 2005). We found that CBF was not significantly different in subcortical than in cortical gray matter, in accord with Anwar et al. (1990), whose control group of rats had similar values of CBF (on the order of 70 mL/100 g/min, measured using 14-C-iodoantipyrine) for both cortical and thalamic regions, and with Adam et al. (2003), where CBF values in the parietal cortex and caudate putamen were 129±18 mL/100 g/min and 125± 17 mL/100 g/min respectively. CBV and CBF in cortical arterioles versus venules The differences in CBV, CBF, and flow speed between arterioles and venules correspond to what is known about arterioles and venules in general. The fraction of blood volume in venules was larger than that in arterioles (rCBVv/rCBVa = 1.2), as expected, although not as prominently as reported in the literature for the whole brain, where CBVv/CBVa is on the order of 2 to 3 (Duong and Kim, 2000; Mellander and Johansson, 1968). The flow speed in the cortex was higher for arterioles than for venules (va/vv = 1.1), though this difference was smaller than that observed using focal measurements (e.g. arteriolar vRBC = 2–4 mm/s and venular vRBC = 0.6–0.8 mm/s as measured using two-photon microscopy (Shih et al., 2012)). The resulting blood flow in the rat cortex was 10% higher in venules than in arterioles. The CBF–CBV relationship Understanding the CBF–CBV relationship is important not only in cerebrovascular disease (Iadecola, 2004) but also in health: for example, Mishra et al. (2011) found an apparent reversal of neurovascular coupling in the caudate putamen, where the BOLD fMRI signal decreased with increased neuronal activity (Mishra et al., 2011). In Lu et al. (2009), a region-dependent coupling between CBF and CBV in anesthetized rats was observed during hyperoxia and hypercapnia, although the cortical and subcortical areas were not significantly different (Lu et al., 2009). Our data on the CBF–CBV relationship does show a statistically significant difference between the exponents in cortical versus subcortical regions (α=0.37±0.13 and 0.75±0.16 respectively), thus strengthening the hypothesis that CBF–CBV coupling is not constant spatially across the brain (Sanganahalli et al., 2009), and that CBF changes more strongly with CBV in the subcortical region of the brain. It is also important to remember that isoflurane, the presently employed anesthetic, is a vasodilator that increases resting blood flow by 20–30% at the currently employed doses (Maekawa et al., 1986). While this moderate rise in resting blood flow may affect the CBF–CBV coupling, our cortical estimate of α was not different from either those reported originally by Grubb in phencyclidine anesthetized monkeys (Grubb et al., 1974) or the one found previously in urethane anesthetized rats undergoing whisker stimulation (Kennerley et al., 2005). Note that in comparing data from stimulation-response studies to steady-state studies such as this work we assume that the global relationship between CBF and CBV applies to focal changes as well. Another important note is that it has been shown that the much-used but simple power law model oversimplifies the vascular system and that, especially as imaging modalities become

1035

able to resolve the microvasculature and the many flow-regulating vessels therein, more complex models will be necessary (Piechnik et al., 2008). Methodological considerations: hemodynamic imaging with microultrasound While microbubbles provide an extremely useful contrast mechanism for ultrasound imaging, there can be several sources of variability related to the microbubbles, that can be categorized as factors related to the scanner, to the subject, and to the microbubbles themselves (Tang et al., 2011). For example, the size distribution of a bubble preparation changes over time. Most sources of variation can be mitigated by being very careful to prepare the bubbles identically for every experiment, and to use identical power and amplification settings on the scanner. Differences between subjects (composition of tissues, tissue motion) that might influence the signal from the microbubbles are all minimal in the case of a rigidly mounted rat head. The arteriovenular segmentation via color Doppler imaging is complicated by the potentially strong effects of various Doppler acquisition parameters on the size and position of the regions in the image reported as having flow towards or away from the transducer. The appendix summarizes these parameters and their optimal setting in the context of quantitative functional brain imaging. Interpretation of replenishment model parameters The mono-exponential model of microbubble replenishment was first applied by Wei et al. (1998) to study the canine heart and has since found widespread application in microbubble replenishment modeling. Even though the exponential model is not based on a physical theory of microbubble replenishment, it captures the essential features of the replenishment time–intensity curves, namely the monotonic increase of intensity and progressive decrease of the rate of intensity increase (Potdevin et al., 2006). The interpretation of the model parameters A, β, and their product Aβ as being proportional to blood volume, flow speed, and blood flow, respectively, rests on a number of assumptions. Models of replenishment assume that the flow occurs mainly through-plane, that is, that the vessels are oriented perpendicular to the plane of the ultrasound image. This is the case for many vessels in the pial and subcortical gray regions, but the cortical vessels are mostly in the plane of the image. However, since we do not compensate for the difference in beam thickness between the disruption beam and the imaging beam (Hudson et al., 2009), and since we do not observe a lag in the replenishment curves, this is not of practical concern. Another assumption is that the concentration of microbubbles in the blood remains constant. This assumption is met since we use a continuous infusion of bubbles (as opposed to a bolus injection) and since we start imaging only after a steady state has been reached. Conclusion We have established that high frequency contrast-enhanced ultrasound imaging can be used to quantify essential hemodynamic parameters in the rodent brain (CBF, CBV, flow speed) using the disruption–replenishment method. Segmentation of the image into arteriolar and venular irrigation regions by color Doppler imaging offers the ability to probe the arteriolar and venular functional responses separately. The initial experiments described here report on the steady-state parameters in healthy anesthetized rats. We anticipate this method to have an impact in fields as diverse as BOLD fMRI modeling, where quantitative interpretation of the signals measured can be corroborated with fMUS as a modality to measure flow and volume concurrently; in studies of vascular pathologies specific to

1036

M.E. van Raaij et al. / NeuroImage 63 (2012) 1030–1037

arteries such as cerebral amyloid angiopathy, where hemodynamic parameters can be quantified in mouse models of the disease (Chen and Zhang, 2011); and in pathologies specific to veins like venous collagenosis, where changes in stiffness of the veins with age lead to changes in resting-state CBV (Brown and Thore, 2011) that can be investigated using fMUS. Acknowledgments The authors thank Eva Chan, Carolyn Cesta, Stephanie Young, and Kogee Leung for their contributions to data acquisition for this study, John Sun for performing initial experiments, and John Hudson for valuable discussions on disruption–replenishment analysis. We gratefully acknowledge financial support from the Canadian Institutes of Health Research, the Natural Sciences and Engineering Research Council of Canada, the Terry Fox Foundation, the Ontario Research Fund and the VisualSonics, Inc.. None of our funding sources had any influence on study design, in the collection, analysis or interpretation of data, in the writing of the report, or in the decision to submit the paper for publication. Appendix A. Color Doppler imaging parameters Critical parameters whose influence should be understood by the operator of a high-frequency ultrasound system include focus depth, gain, color Doppler signal threshold, wall filter cutoff frequency, pulse repetition frequency (PRF), and the gate settings. The focus depth in the plane is determined by the electronics that control the individual elements in the array transducer. The signal-to-noise ratio (SNR) of the Doppler signal is greatest in the plane perpendicular to the axial direction of the beam. With our system, image regions more than about 2 mm away from the focus depth have a SNR below the display threshold and thus show no flow signal. The gain of the amplifier influences detectability of weak signals, but a high gain also results in many noise voxels (i.e., where at a low gain there would be no ‘colored’ voxel, there appears one at high gain). While this is not a significant problem in a real-time acquisition where the operator visually judges the location of the flow, in a single static image there is no way of knowing post-hoc whether any given voxel represents signal or just noise. Some amount of temporal signal averaging may be applied to address this issue. The color Doppler signal threshold is the signal level above which a given voxel will have a Doppler signal recorded. The threshold has a strong effect on the reported size of the arteriolar and venular regions associated with each vessel. The wall filter (also: clutter filter) removes artifacts due to motion of the vessel wall and surrounding tissue. Tissue moves slowly relative to blood in larger vessels, resulting in low Doppler frequencies that can be filtered out by a high pass filter. However, a high wall filter cutoff frequency will also reduce the signal from slow-moving capillary blood. Since the rat head is immobilized by the stereotaxic stage and gross motion is minimal, the wall filter should be set as low as possible. The pulse repetition frequency (prf) determines the maximum observable Doppler shift, and therefore the range of velocities that can be measured. Lowering the prf results in more sensitivity to slow flow, but at the cost of incurring aliasing artifacts as shown in Fig. 5A. (The upper limit of flow speed that can be measured is determined by the frame rate of the ultrasound image acquisition, whereas the lower limit is dictated by the length of the recording time.) RBC velocities in capillaries in the rat range from 0.15 to about 8.6 mm/s (Driscoll et al., 2009; Hutchinson et al., 2006; Unekawa et al., 2010), and in arterioles to about 20 mm/s (Jensen, 1996). A prf of 1 kHz will adequately sample these velocities in our experimental situation. Finally, the gate determines the axial sampling volume and therefore resolution: a small gate setting results in higher spatial resolution but poorer sensitivity. Taken together, these parameters can have a profound effect on the color Doppler image.

References Adam, J.-F., Elleaume, H., Le Duc, G., Corde, S., Charvet, A.-M., Tropres, I., Le Bas, J.-F., Esteve, F., 2003. Absolute cerebral blood volume and blood flow measurements based on synchrotron radiation quantitative computed tomography. J. Cereb. Blood Flow Metab. 23, 499–512. Anwar, M., Kissen, I., Weiss, H.R., 1990. Effect of chemodenervation on the cerebral vascular and microvascular response to hypoxia. Circ. Res. 67, 1365–1373. Attwell, D., Buchan, A.M., Charpak, S., Lauritzen, M., Macvicar, B.A., Newman, E.A., 2010. Glial and neuronal control of brain blood flow. Nature 468, 232–243. Becher, H., Burns, P.N., 2000. Handbook of Contrast Echocardiography. Springer, Berlin. Brown, W.R., Thore, C.R., 2011. Review: cerebral microvascular pathology in ageing and neurodegeneration. Neuropathol. Appl. Neurobiol. 37, 56–74. Calamante, F., Thomas, D.L., Pell, G.S., Wiersma, J., Turner, R., 1999. Measuring cerebral blood flow using magnetic resonance imaging techniques. J. Cereb. Blood Flow Metab. 19, 701–735. Chen, H., Zhang, J.H., 2011. Cerebral amyloid angiopathy-related microhemorrhages in Alzheimer's disease: a review of investigative animal models. Acta Neurochir. Suppl. 111, 15–17. Coles, J.P., 2006. Imaging of cerebral blood flow and metabolism. Curr. Opin. Anaesthesiol. 19, 473–480. D'Esposito, M., Deouell, L.Y., Gazzaley, A., 2003. Alterations in the BOLD fMRI signal with ageing and disease: a challenge for neuroimaging. Nat. Rev. Neurosci. 4, 863–872. Devor, A., Sakadzic, S., Srinivasan, V.J., Yaseen, M.A., Nizar, K., Saisan, P.A., Tian, P., Dale, A.M., Vinogradov, S.A., Franceschini, M.A., Boas, D.A., 2012. Frontiers in optical imaging of cerebral blood flow and metabolism. J. Cereb. Blood Flow Metab. 32 (7), 1259–1276 (July). Driscoll, J.D., Shih, A.Y., Drew, P.J., Valmianski, I., Kleinfeld, D., 2009. Quantitative twophoton imaging of blood flow in cortex. In: Helmchen, F., Konnerth, A. (Eds.), Optical Imaging in Neuroscience. Cold Spring Harbor Laboratory Press. Duong, T.Q., Kim, S.G., 2000. In vivo MR measurements of regional arterial and venous blood volume fractions in intact rat brain. Magn. Reson. Med. 43, 393–402. Ferrara, K., Pollard, R., Borden, M., 2007. Ultrasound microbubble contrast agents: fundamentals and application to gene and drug delivery. Annu. Rev. Biomed. Eng. 9, 415–447. Foster, F.S., Zhang, M.Y., Zhou, Y.Q., Liu, G., Mehi, J., Cherin, E., Harasiewicz, K.A., Starkoski, B.G., Zan, L., Knapik, D.A., Adamson, S.L., 2002. A new ultrasound instrument for in vivo microimaging of mice. Ultrasound Med. Biol. 28, 1165–1172. Goertz, D.E., Needles, A., Burns, P.N., Foster, F.S., 2005. High-frequency, nonlinear flow imaging of microbubble contrast agents. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 52, 495–502. Grubb Jr., R.L., Raichle, M.E., Eichling, J.O., Ter-Pogossian, M.M., 1974. The effects of changes in PaCO2 on cerebral blood volume, blood flow, and vascular mean transit time. Stroke 5, 630–639. Herman, P., Sanganahalli, B.G., Hyder, F., 2009. Multimodal measurements of blood plasma and red blood cell volumes during functional brain activation. J. Cereb. Blood Flow Metab. 29, 19–24. Hudson, J.M., Karshafian, R., Burns, P.N., 2009. Quantification of flow using ultrasound and microbubbles: a disruption replenishment model based on physical principles. Ultrasound Med. Biol. 35, 2007–2020. Hutchinson, E.B., Stefanovic, B., Koretsky, A.P., Silva, A.C., 2006. Spatial flow-volume dissociation of the cerebral microcirculatory response to mild hypercapnia. Neuroimage 32, 520–530. Iadecola, C., 2004. Neurovascular regulation in the normal brain and in Alzheimer's disease. Nat. Rev. Neurosci. 5, 347–360. Iadecola, C., Nedergaard, M., 2007. Glial regulation of the cerebral microvasculature. Nat. Neurosci. 10, 1369–1376. Jensen, J.A., 1996. Estimation of Blood Velocities Using Ultrasound: A Signal Processing Approach. Cambridge University Press, Cambridge; New York, USA. Kennerley, A.J., Berwick, J., Martindale, J., Johnston, D., Papadakis, N., Mayhew, J.E., 2005. Concurrent fMRI and optical measures for the investigation of the hemodynamic response function. Magn. Reson. Med. 54, 354–365. Kim, T., Kim, S.G., 2005. Quantification of cerebral arterial blood volume and cerebral blood flow using MRI with modulation of tissue and vessel (MOTIVE) signals. Magn. Reson. Med. 54, 333–342. Lampaskis, M., Averkiou, M., 2010. Investigation of the relationship of nonlinear backscattered ultrasound intensity with microbubble concentration at low MI. Ultrasound Med. Biol. 36, 306–312. Lindvere, L., Dorr, A., Stefanovic, B., 2010. Two-photon fluorescence microscopy of cerebral hemodynamics. Cold Spring Harb. Protoc. http://dx.doi.org/10.1101/pdb.prot5494. Lu, J., Dai, G., Egi, Y., Huang, S., Kwon, S.J., Lo, E.H., Kim, Y.R., 2009. Characterization of cerebrovascular responses to hyperoxia and hypercapnia using MRI in rat. Neuroimage 45, 1126–1134. Mace, E., Montaldo, G., Cohen, I., Baulac, M., Fink, M., Tanter, M., 2011. Functional ultrasound imaging of the brain. Nat. Methods 8, 662–664. Maekawa, T., Tommasino, C., Shapiro, H.M., Keifer-Goodman, J., Kohlenberger, R.W., 1986. Local cerebral blood flow and glucose utilization during isoflurane anesthesia in the rat. Anesthesiology 65, 144–151. Mellander, S., Johansson, B., 1968. Control of resistance, exchange, and capacitance functions in the peripheral circulation. Pharmacol. Rev. 20, 117–196. Mishra, A.M., Ellens, D.J., Schridde, U., Motelow, J.E., Purcaro, M.J., DeSalvo, M.N., Enev, M., Sanganahalli, B.G., Hyder, F., Blumenfeld, H., 2011. Where fMRI and electrophysiology agree to disagree: corticothalamic and striatal activity patterns in the WAG/Rij rat. J. Neurosci. 31, 15053–15064.

M.E. van Raaij et al. / NeuroImage 63 (2012) 1030–1037 Motulsky, H.J., Christopoulos, A., 2003. Fitting models to biological data using linear and nonlinear regression. A Practical Guide to Curvefitting. GraphPad Software Inc., San Diego CA. www.graphpad.com. Mulvagh, S.L., Rakowski, H., Vannan, M.A., Abdelmoneim, S.S., Becher, H., Bierig, S.M., Burns, P.N., Castello, R., Coon, P.D., Hagen, M.E., Jollis, J.G., Kimball, T.R., Kitzman, D.W., Kronzon, I., Labovitz, A.J., Lang, R.M., Mathew, J., Moir, W.S., Nagueh, S.F., Pearlman, A.S., Perez, J.E., Porter, T.R., Rosenbloom, J., Strachan, G.M., Thanigaraj, S., Wei, K., Woo, A., Yu, E.H.C., Zoghbi, W.A., 2008. American society of echocardiography consensus statement on the clinical applications of ultrasonic contrast agents in echocardiography. J. Am. Soc. Echocardiogr. 21, 1179–1201. Needles, A., Arditi, M., Rognin, N.G., Mehi, J., Coulthard, T., Bilan-Tracey, C., Gaud, E., Frinking, P., Hirson, D., Foster, F.S., 2010. Nonlinear contrast imaging with an array-based micro-ultrasound system. Ultrasound Med. Biol. 36, 2097–2106. Piechnik, S.K., Chiarelli, P.A., Jezzard, P., 2008. Modelling vascular reactivity to investigate the basis of the relationship between cerebral blood volume and flow under CO2 manipulation. Neuroimage 39, 107–118. Potdevin, T.C., Fowlkes, J.B., Moskalik, A.P., Carson, P.L., 2006. Refill model of rabbit kidney vasculature. Ultrasound Med. Biol. 32, 1331–1338. Powers, J., Averkiou, M., Bruce, M., 2009. Principles of cerebral ultrasound contrast imaging. Cerebrovasc. Dis. 27, 14–24. Sanganahalli, B.G., Herman, P., Blumenfeld, H., Hyder, F., 2009. Oxidative neuroenergetics in event-related paradigms. J. Neurosci. 29, 1707–1718.

1037

Shih, A.Y., Driscoll, J.D., Drew, P.J., Nishimura, N., Schaffer, C.B., Kleinfeld, D., 2012. Twophoton microscopy as a tool to study blood flow and neurovascular coupling in the rodent brain. J. Cereb. Blood Flow Metab. 32 (7), 1277–1309 (July). Sloan, H.L., Austin, V.C., Blamire, A.M., Schnupp, J.W., Lowe, A.S., Allers, K.A., Matthews, P.M., Sibson, N.R., 2010. Regional differences in neurovascular coupling in rat brain as determined by fMRI and electrophysiology. Neuroimage 53, 399–411. Sprague, M.R., Cherin, E., Goertz, D.E., Foster, F.S., 2010. Nonlinear emission from individual bound microbubbles at high frequencies. Ultrasound Med. Biol. 36, 313–324. Tang, M.X., Mulvana, H., Gauthier, T., Lim, A.K.P., Cosgrove, D.O., Eckersley, R.J., Stride, E., 2011. Quantitative contrast-enhanced ultrasound imaging: a review of sources of variability. Interface Focus 1, 520–539. Unekawa, M., Tomita, M., Tomita, Y., Toriumi, H., Miyaki, K., Suzuki, N., 2010. RBC velocities in single capillaries of mouse and rat brains are the same, despite 10-fold difference in body size. Brain Res. 1320, 69–73. van Raaij, M.E., Lindvere, L., Dorr, A., He, J., Sahota, B., Foster, F.S., Stefanovic, B., 2011. Functional micro-ultrasound imaging of rodent cerebral hemodynamics. Neuroimage 58, 100–108. Wei, K., Jayaweera, A.R., Firoozan, S., Linka, A., Skyba, D.M., Kaul, S., 1998. Quantification of myocardial blood flow with ultrasound-induced destruction of microbubbles administered as a constant venous infusion. Circulation 97, 473–483.