Functional micro-ultrasound imaging of rodent cerebral hemodynamics

NeuroImage 58 (2011) 100–108

Contents lists available at ScienceDirect

NeuroImage j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / y n i m g

Functional micro-ultrasound imaging of rodent cerebral hemodynamics Martijn E. van Raaij ⁎, 1, Liis Lindvere 1, 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: Received 22 February 2011 Revised 11 April 2011 Accepted 23 May 2011 Available online 24 June 2011 Keywords: Functional micro-ultrasound imaging (fMUS) Cerebral blood volume (CBV) Primary somatosensory cortex: forelimb region (S1FL) Functional neuroimaging Cerebral hemodynamics Rat

a b s t r a c t Healthy cerebral microcirculation is crucial to neuronal functioning. We present a new method to investigate microvascular hemodynamics in living rodent brain through a focal cranial window based on high-frequency ultrasound imaging. The method has a temporal resolution of 40 ms, and a 100 μm in-plane and 600 μm throughplane spatial resolution. We use a commercially available high-frequency ultrasound imaging system to quantify changes in the relative cerebral blood volume (CBV) by measuring the scattered signal intensity from an ultrasound contrast agent circulating in the vasculature. Generalized linear model analysis is then used to produce effect size and significance maps of changes in cerebral blood volume upon electrical stimulation of the forepaw. We observe larger CBV increases in the forelimb representation of the primary somatosensory cortex than in the deep gray matter with stimuli as short as 2 s (5.1± 1.3% vs. 3.3 ± 0.6%). We also investigate the temporal evolution of the blood volume changes in cortical and subcortical gray matter, pial vessels and subcortical major vessels, and show shorter response onset times in the parenchymal regions than in the neighboring large vessels (1.6± 1.0 s vs. 2.6 ± 1.3 s in the cortex for a 10 second stimulus protocol). This method, which we termed functional micro-ultrasound imaging or fMUS, is a novel, highly accessible, and cost-effective way of imaging rodent brain microvascular topology and hemodynamics in vivo at 100 micron resolution over a 1-by-1 cm field of view with 10s–100s frames per second that opens up a new set of questions regarding brain function in preclinical models of health and disease. © 2011 Elsevier Inc. All rights reserved.

Introduction Healthy functioning of neurons critically depends on a stable level of cerebral blood flow. Further, changes in neuronal activity induce rapid adjustments in flow and volume of the surrounding microvessels. This spatiotemporal correlation between neuronal and vascular state, termed neurovascular coupling, was first described over two centuries ago by Angelo Mosso (Mosso, 1881) and arises from a highly complex system controlling the cerebral hemodynamics. Neurovascular coupling remains a major research focus (Attwell et al., 2010) — in no small part due to the recent recognition that neurovascular coupling gets perturbed in virtually all brain disorders and diseases (D'Esposito et al., 2003; Iadecola, 2004). Neurovascular coupling is also the basis of a range of widely employed methods for brain function imaging — from positron emission tomography, optical intrinsic signal imaging and near-infrared spectroscopy to functional magnetic resonance imaging. Each of these methods offers a different set of tradeoffs among various imaging

Abbreviations: fMUS, functional micro-ultrasound imaging; GLM, generalized linear model; HRF, hemodynamic response function. ⁎ 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] (M.E. van Raaij). 1 These authors contributed equally to this work. 1053-8119/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2011.05.088

parameters – most importantly spatiotemporal resolution, brain coverage, range of sources of contrast, and accessibility/ease of use – and thus allows interrogation of a distinct set of hypotheses regarding brain functioning. The current work introduces another hemodynamically weighted method for investigation of brain function in animal models, namely high frequency ultrasound imaging with ultrasound contrast agents. Although not typically used in neuroimaging, ultrasound is one of the most widely available and pervasive technologies employed in clinical imaging today. Ultrasound images are essentially maps of the elastic properties of tissue, formed by sending high frequency acoustic pulses into the tissue and recording the echoes generated when those pulses are scattered by various structures in the tissue (Shung, 2005; Cobbold, 2008). The simplicity, ease of use, speed, and safety of ultrasound have led to a significant role in diagnosis, treatment assessment, follow-up, and guidance of therapy (Rumack et al., 1998). The maximum imaging frequencies of clinical systems are typically 12 – 15 MHz with in-plane spatial resolution reaching about 300 μm. To improve spatial resolution and allow for imaging of structures in small animal models, high frequency (15 – 50 MHz) imaging systems have been developed over the past decade (Foster et al., 2002). These systems provide imaging depths of 15–20 mm with in-plane resolution between 40 and 200 μm and have found wide applications in cardiovascular research (Ino et al., 1996; Zhou et al., 2003; Zhou et al., 2004; Liu et al.,

M.E. van Raaij et al. / NeuroImage 58 (2011) 100–108

2007; Trivedi et al., 2007; Lee et al., 2008), and in cancer research (Wirtzfeld et al., 2005; Olive and Tuveson, 2006; Wirtzfeld et al., 2006; Goessling et al., 2007; Kiguchi et al., 2007; Xuan et al., 2007). In 2009, systems based on linear transducer array technology were introduced to improve depth of field and allow realtime assessment of blood flow using Doppler and contrast agent techniques in small animal models (Foster et al., 2009). The latter technology is used in this study. The development of ultrasound contrast agents over the past decade has provided a convenient means for visualization of the microcirculation (Becher and Burns, 2000; Ferrara et al., 2007). Microbubble contrast agents are typically composed of a lipid membrane encapsulating a gas such as perfluorobutane and have a diameter on the order of 1 to 5 μm. The size of the microbubbles is such that they resonate at ultrasound frequencies, creating strong nonlinear signals that can be isolated from the (linearly scattered) tissue signal. Because the microbubbles are strictly intravascular they make an excellent tracer for the study of cerebral hemodynamics. Microbubbles are used as ultrasound contrast agents both in preclinical animal models and in clinical practice in many countries, and their biosafety has been investigated extensively (see e.g. Mulvagh et al., 2008). Ultrasound is not typically used in functional neuroimaging of mice and rats because the bones of the skull reflect most of the ultrasound energy. If a focal craniotomy may be performed, however, ultrasound has unique advantages which make it an attractive alternative to other modalities in this arena. The current work presents an initial demonstration of the utility of high frequency ultrasound imaging for noninvasive investigation of cerebral microvascular topology and hemodynamics in an anesthetized rodent model. Color Doppler imaging is employed for mapping of the blood flow direction in the cortical penetrating vessels. Thereafter, contrast-enhanced ultrasound imaging, during continuous infusion of microbubbles, is performed to map the spatiotemporal dynamics of the cortical vs subcortical cerebral blood volume response induced by electrical forepaw stimulation. Materials and methods Animal preparation The surgical procedures employed here are described in detail in (Lindvere et al., 2010). Briefly, male Sprague–Dawley rats (m = 174 ± 39 g, Charles River Laboratories Inc, Wilmington, MA) were induced with 5% isoflurane and maintained throughout surgery with 2–3% isoflurane in oxygen-enriched medical air. Animals were tracheotomized, mechanically ventilated (SAR 830/P, CWE Inc., Ardmore, PA) and placed on an electric heating pad. The tail vein, femoral vein and femoral artery were cannulated for delivery of ultrasound contrast agent,

101

infusion of α-chloralose anesthetic, and blood gas sampling and blood pressure monitoring, respectively. A craniotomy was performed removing a rectangular section of skull (−4.0 to 2.5 mm AP and 2 to 5 mm ML) over the primary somatosensory cortex, while the dura remained intact. After surgery was complete, isoflurane was discontinued and anesthetic was changed to α-chloralose (80 mg/kg bolus induction, followed by constant infusion at 26.7 mg/kg per hour) to minimize the impact on neurovascular coupling (Ueki et al., 1992; Lindauer et al., 1993). Electrodes were placed into the forelimb pad between digits 2 and 3, and 4 and 5, contralateral to the craniotomy. During the surgery and the entire imaging protocol, the physiological state of the animal was monitored and recorded (Biopac MP150, Biopac Systems Inc., Goleta, CA): temperature using a rectal temperature probe connected to a DC proportional temperature controller (SWE Inc, Ardmore, PA), heart rate and arterial blood oxygenation using a pulse oximeter (MouseOx, StarrLite Sciences, Oakmont, PA), arterial blood pressure using a pressure transducer connected to the femoral artery catheter, the end-tidal respiratory pressure via the ventilator, and blood pH, pCO2 and pO2 through hourly arterial blood gas analyses. Data recorded when temperature, blood pressure or heart rate were outside of the physiological range as defined in (Lindvere et al., 2010) were excluded from the analysis. All animal experiments reported on in this study have been approved by the Animal Care Committee of Sunnybrook Health Sciences Centre. Functional ultrasound image acquisition Ultrasound imaging was performed using 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). The transducer was held in a stereotactic frame (Fig. 1) and placed in a near-sagittal plane (−9° relative to the midline), 4 mm lateral to the midline, so as to intersect the forelimb representation of the primary somatosensory cortex (S1FL) and the ventro-posterolateral nucleus of the thalamus (VPL) (Fig. 3B). The ultrasound was coupled into the brain tissue via standard ultrasound gel and the layer of agarose covering the craniotomy. An initial overview standard ultrasound ‘B-mode’ image was acquired to provide a scout image for the functional imaging and assess the position of the field of view. A Color Doppler image, showing motion of blood away from and towards the transducer, was acquired to identify cortical penetrating vessels as arterioles and venules, respectively. An ultrasound contrast agent (Vevo MicroMarker, untargeted, VisualSonics Inc.) consisting of a lipid shell with a diameter ranging from 0.5 to 5 μm, filled with perfluorocarbon gas, was dissolved in saline to a concentration of 6 · 108 bubbles/ml. The contrast agent was infused using an infusion pump at a constant rate of 40 μl/min for 7 min, yielding

Fig. 1. Photographs of the functional micro-ultrasound imaging system. (A) shows from left to right the ultrasound acquisition system, the Faraday cage in which the experiment is performed, the physiological monitoring equipment and the surgical table. (B) Close-up of the ultrasound transducer (t) mounted in the stereotactic stage (s) with heating pad (h).

102

M.E. van Raaij et al. / NeuroImage 58 (2011) 100–108

a ‘bubble hematocrit’ of no more than 0.002% in our subjects, assuming a total blood volume of 12 ml. The bubble concentration in the blood reached a steady state after 2.5 min (as established by preliminary experiments), at which point functional imaging was started. The bubbles circulate freely in the bloodstream until they are cleared out by the liver, kidneys and lungs: once the infusion is discontinued, the signal drops to 50% in approximately 2 min. No more than 7 infusions were done per animal to limit the total volume of infused contrast agent suspension to approximately 2 ml, with 10-15 min between infusions, over a period of about 3 h. The first infusion of contrast agent was used to establish a steady state map of baseline blood flow and volume using the disruptionreplenishment technique described in (Sun et al., 2010) (data not shown). The stimulus patterns were presented during the subsequent six infusions. We estimated CBV changes by using a ‘nonlinear contrast’ (NLC) ultrasound imaging mode, which employs an ultrasound pulse sequence designed to receive only backscattered ultrasound from nonlinear scatterers (i.e., microbubbles) and reject signal from linear scatterers like tissue or red blood cells (Goertz et al., 2005; Needles et al., 2010). For the low concentrations of bubbles employed here, the signal intensity in this imaging mode is linearly dependent on bubble concentration, effectively making the signal a relative measure of local blood plasma volume (Lampaskis and Averkiou, 2010). Since the hematocrit is largely unaffected by functional activation (Herman et al., 2009), the measured plasma volume changes provide an estimate of total blood volume change. The field of view was 8 mm wide and 10 mm deep measured from the surface of the cortex. The cross section of an ultrasound beam emanating from a linear array transducer such as the one used in this work varies with distance away from the active elements. The imaging slice thickness (for the near-sagittal sections imaged here, slice thickness is the extent in the lateral-to-medial direction) is hourglassshaped by an acoustic lens with a fixed focal distance. The slice thickness in our images is greatest (~2 mm) at the cortical surface and reduces to ~600 μm at a depth of 10 mm. We utilized two forepaw electrical stimulation paradigms: a ‘long stimulus protocol’ with three blocks of 10 s off, 10 s on (Fig. 2A), and a ‘short stimulus protocol’ with three blocks of 5 s off, 2 s on (Fig. 2B). Both protocols used 2 mA pulses of 0.3 ms duration played out at 3 Hz, to effect maximal cortical hemodynamic response in this model (Silva et al., 1999). Image sequences were recorded at 10 frames per second (fps) for the long stimulus paradigm with a run time of 100 s and 25 fps for the short stimulus paradigm with a run time of 40 s. The run time (and therefore the interstimulus rest interval) was limited by available memory in the ultrasound system. Recordings of physiological parameters and stimulus presentation as described above were triggered by the start of the ultrasound imaging sequence. We recorded up to 9 repetitions of the long stimulus protocol and 12 of the short stimulus protocol per subject. Imaging for the long stimulus protocol was always completed before imaging of the short protocol.

Fig. 2. Stimulation paradigms and modeled response. Time courses of (A) long stimulus protocol and (B) short stimulus protocol. Shaded bars show on-times of stimulation blocks. Solid lines show the modeled local blood volume response. Each block consists of a train of 300 μs, 2 mA pulses played out at 3 Hz.

Finally, raw nonlinear-contrast image sequences (hereafter referred to as runs) and the corresponding physiological monitoring timetraces were exported from their respective acquisition systems for offline analysis. Electrophysiological recordings In a subset of the subjects (N = 4), somatosensory evoked potentials were measured after ultrasound imaging to assess neuronal response to the stimuli (Ngai et al., 1998; Stefanovic et al., 2006): three burr holes (0.3 mm diameter, no more than 1 mm below the skull) were made over the left S1FL region, right S1FL region, and the midline of the cerebellum. Tripolar electrodes were sealed into the burr holes to record the differential voltage between left and right S1FL areas, with the cerebellum location as a reference, during stimulation identical to that used in the imaging protocol. The subjects were not moved between ultrasound imaging and electrophysiology recordings; experiments were performed in a Faraday cage to minimize RF noise. Functional ultrasound data analysis Generalized linear model (GLM) analysis was applied to identify regions of statistically significant stimulus correlation at the omnibus significance level of 0.05 after correction for multiple comparisons. The analysis was performed using the AFNI suite of programs (Cox, 1996). Anatomical images showing the location of the blood vessels were created by making an average intensity projection over time of each run. Within each subject, all runs were spatially aligned to the first run in the imaging plane by registering their average intensity projections using local Pearson correlation as the cost functional and allowing only inplane translations (AFNI: 3dAllineate). Efficacy of the alignment procedure was inspected visually by overlaying average intensity projections before and after alignment. Data were temporally smoothed using a 0.5 s Hamming window to reduce high temporal frequency noise, and spatially smoothed with a Gaussian filter with a 150 μm full width at half maximum to reduce high spatial frequency noise. A manually defined mask was applied to exclude out-of-brain voxels. The combined time-series data for all runs in a given subject was ordinary-least-squares fitted in each voxel using a generalized linear model (AFNI: 3dDeconvolve) (Cox, 1996). The GLM incorporated a modeled hemodynamic response and a first order polynomial for each run to account for baseline variations. The modeled hemodynamic response was obtained by convolving an assumed hemodynamic impulse response function (HRF) with the block stimulus pattern. The HRF was based on the blood volume response measured via contrast enhanced MRI in the rat cortex after electrical stimulation of the forepaw with the same functional paradigm and the same anesthesia protocol as those used in the present work (Hirano et al., 2011). The hemodynamic response function was defined using AFNI: WAVER, (Cox, 1996) to have an onset time of 0.34 s, a time-to-peak of 2.7 s, and a FWHM of 3.6 s (Fig. 2). A t-statistic map was used to determine incidence of activation. The inferences were corrected for multiple comparisons by determining the value of the t-threshold given a false-discovery rate (FDR) of at most 5% (i.e. q b 0.05) (Genovese et al., 2002). Activated voxels were also clustered with a minimum cluster size of 190 voxels (corresponding to a circle with a diameter of 480 μm). This minimum cluster size is based on a random-noise simulation which suggested that given the number and size of voxels in our dataset, random noise thresholded at a voxelwise p-threshold of 0.05 would yield a false-positive cluster of 190 voxels 1.0% of the time, i.e., αcluster = 0.010 (AFNI: 3dClustSim). Effect size maps in units of percent difference were calculated as the estimate of the amplitude of the modeled response regressor divided by the run-averaged constant baseline estimate in each voxel. The images were segmented manually into four anatomical regions of interest

M.E. van Raaij et al. / NeuroImage 58 (2011) 100–108

(ROIs): pial vessel region, cortex region, deep gray region, and major vessels in deep gray region. The major vessel region was defined as vessels in subcallosal gray matter with an apparent diameter larger than 200 μm. The vessel density and signal intensity in the anatomical images were used to guide ROI deliniation. Time-intensity traces for activated voxels in each region were normalized to their constant baseline as reported by the GLM analysis and averaged over all runs in each subject. The three replicates of the stimulus block in each run-average trace were then averaged to produce subject-average response traces. These were then averaged across N = 11 rats (long stimulus protocol) and 8 rats (short-stimulus protocol) respectively to yield the final cerebral blood volume response time-intensity traces. The percent signal changes estimated from the GLM analysis were used in linear mixed effects modeling, by restricted maximum likelihood (‘lme’ function in R statistical software, http://www.r-project.org), to investigate the effect of the region of interest on the relative response amplitude, after controlling for inter-subject variability, in both short and long stimulation paradigms. To quantify the temporal dynamics of the CBV responses, we estimated onset time, time-to-peak, peak amplitude and offset time for each subject individually. After a 10 × interpolation to improve the

103

accuracy of parameter estimation, we estimated the onset time as the time when the response rises over (μ + σ) of the pre-stimulus baseline for at least 1 s (Hirano et al., 2011), and the offset time as the time after the stimulus has ended when the response falls to below (μ + σ) of the pre-stimulus baseline for at least 1 s. The time-to-peak and peak amplitude estimates were obtained by a gamma variate fit to the response curves. All inferences in the results section have p b 0.05 unless otherwise noted. Results Anatomical ultrasound imaging Skull appears hyperintense in standard ultrasound B-mode images of the rat brain (Fig. 3C) due to high scattering of the acoustic wave by bone, but little contrast is evident in the brain itself in a B-mode image. The topography of the microvasculature in the rat brain can be appreciated in the average intensity projection of a nonlinear contrast image over time as shown in Fig. 3D. In this image we can identify the cortex, with most of the larger vessels running perpendicular to the brain surface, the less well-perfused corpus callosum, and a deep gray region with larger, less regularly oriented vessels. Color Doppler mode

Fig. 3. In vivo micro-ultrasound imaging has several imaging modes that offer complementary information on brain microvascular topography and function. (A) Micro-CT image of a rat cranium with a 3.5 × 4.5 mm cranial window and anatomical landmarks L, lambda; and B, bregma (typical subjects had only one cranial window in this study). (B) Atlas image showing the cross-section of the rat brain imaged by the transducer (red box). The imaging slice was near-sagittal: − 4 mm lateral of the midline, at a 9º angle to the sagittal plane, so as to include S1FL, CPu, and VPL. The ultrasound field of view is 8.0 × 11.0 mm. Annotations clockwise from top left: Tr, primary somatosensory cortex: trunk region; SH, -shoulder region; S1FL, -forelimb region; S1J, -jaw region; CPu, caudate putamen; Pir, piriform cortex; GP, globus pallidus; Rt, reticular thalamic nucleus; VPL, ventral posterolateral thalamic nucleus. Anterior side of animal (marked A) is to the right in the image. Atlas image produced using Brain Navigator (Paxinos and Watson, 2007; Paxinos and Watson, 2009). (C) Ultrasound B-mode image showing the position of the cranial window in the skull but little contrast in the brain itself. (D) Average intensity projection of a nonlinear contrast mode time series showing the topography of microvasculature in the imaging plane. (E) Color Doppler imaging identifies the direction and velocity of red blood cell flow in each of the vessels. The color scale represents blood flow velocity in mm/s. Panels C-E show the same field of view in the same subject.

104

M.E. van Raaij et al. / NeuroImage 58 (2011) 100–108

images (Fig. 3E) allow for identification of individual penetrating vessels as arterioles (blood flowing away from the transducer, down into the brain, shown in red) and venules (blood flowing towards the transducer, toward the brain surface, shown in blue), suggesting a ratio of cortical penetrating arterioles to venules of approximately 2:1, which is well in the range of previously reported values for the ratio of cortical penetrating arterioles to venules (Bar, 1981; McHedlishvili and Kuridze, 1984; Woolsey et al., 1996). Functional micro-ultrasound imaging Voxel-wise effect size maps for a sample subject are shown in Fig. 4. Regions of interest are defined based on the blood vessel topography as evident from the average intensity projection (Fig. 4A) and are labeled as follows: p, pial region; c, cortex; d, deep gray; and m, major vessels in deep gray. The top row of images (Figs. 4B–D) shows the results of the long stimulus protocol; the bottom row (Figs. 4E–G) shows the results of the short stimulus protocol. The effect size maps (Figs. 4B and E, with the color scale representing the change in CBV in %), and the corresponding t-statistic maps (Figs. 4C and F), reveal a region of strong activation in the cortex and several smaller patches of apparent activation in other parts of the cortex and deep gray regions. Average time-intensity traces from all voxels deemed activated within each ROI (Figs. 4D and G) suggest stronger stimulus-correlated signal changes in the cortex than in the deep gray region. For the short stimulus protocol, we frequently observed a slowly rising signal in the pial region, possibly resulting from the short interstimulus interval relative to post-stimulus offset dynamics of pial vessels.

The contrast-to-noise ratio (CNR) of the CBV responses, calculated as the region-mean amplitude of the signal change (the β-coefficient of the response regressor) divided by the standard deviation of the pre-stimulus baseline, was significantly higher in the parenchyma (cortex and the deep gray regions) than in the vascular regions (pial vessels and major vessels in deep gray) due to the higher baseline signal variation in the intravascular space. Furthermore, the CNR is higher in the short than in the long protocol, with the subject-mean CNR in the cortex being 3.3 ± 1.7 (long protocol) and 4.0 ± 1.6 (short protocol) respectively. In our current implementation, fMUS allows for imaging with a temporal resolution of up to 25 frames per second, which enables temporal characterization of the local blood volume response on a region-specific level. The region-average normalized local blood volume response timetraces within each subject and each stimulation protocol were first averaged within each subject, and then across all subjects, to yield the single-response time series shown in Fig. 5 for each region and protocol. The shaded region behind each curve indicates the standard deviation across subjects for that response. The temporal characteristics of these responses have been quantified in Fig. 6. The onset times (Figs. 6A and D), time-to-peak (Figs. 6B and E), peak amplitude (Figs. 6C and F) and offset time (Fig. 6G) are reported as mean and standard deviation across subjects. Linear mixed effects modeling showed that after controlling for inter-subject variability, for the long stimulus paradigm, the cortical and deep gray onset times were not distinguishable, but they were both shorter than the onset time in major deep gray vessels. Cortical onset time was also shorter than that in the pial vessels region. In the short stimulus paradigm, no statistically significant differences

Fig. 4. Functional micro-ultrasound images of the response to somatosensory stimulation. The imaging plane is a near-sagittal section through S1 and VPL with the anterior direction to the right as shown in Fig. 3. (A) Anatomical image with manually segmented regions of interest (p, pial vessels; c, cortex; d, deep gray; m, major vessels in deep gray). (B-D) refer to the long stimulation protocol and (E-G) to the short stimulation protocol in the same subject. (B and E) Effect size map in %-signal change as output by the GLM analysis. (C and F) Corresponding t-statistic map (t-threshold was 2.5, corresponding to p b 0.05). (D and G) Average timetraces of activated voxels in each of the ROIs. Shaded bars indicate stimulus-on intervals; solid red line indicates modeled blood volume response (mHRF). The GLM analysis for this subject included 6 runs (long protocol) and 9 runs (short protocol) respectively. The contrast-to-noise ratio of the CBV response in the cortex was 4.6 (long protocol) and 6.6 (short protocol) for this subject.

M.E. van Raaij et al. / NeuroImage 58 (2011) 100–108

Fig. 5. Temporal evolution of blood volume response in different brain regions. The graph shows the responses that have been averaged within a subject across runs and repetitions and then across subjects, for the long stimulus protocol (left hand side) and the short stimulus protocol (right hand side) for all regions of interest. p, pial vessels; c, cortex; d, deep gray; m, major vessels in deep gray. The lowest trace is the modeled hemodynamic response function. The vertical shaded bars denote the stimulus-on intervals (0 to 10 s for the long stimulus protocol, 0 to 2 s for the short stimulus protocol); the shaded regions behind each response curve denote the standard deviation across subjects for that response curve. N = 11 and 8 subjects for the long and short stimulus protocol, respectively.

between any regional onset times were observed even after controlling for inter-subject variability. For the long paradigm, cortical time to peak (TTP) was shorter than the TTPs in all other regions. For the short paradigm, pial vessel TTP was longer than the TTP of all other regions. Peak response amplitude was higher in the cortex than in the deep gray for both short and long stimulation protocols. In the long stimulus protocol, peak cortical response amplitude was also higher than that in both intravascular regions of interest. Further, peak response amplitude was higher in the long than short stimulation protocol, though the short interstimulus interval makes interpretation of this contrast difficult. Finally, in the long stimulus paradigm, we found that the off-time between stimulus-on blocks was too short to allow the signal to return to the pre-stimulus baseline, which precluded an accurate estimate of the offset time. In the short stimulus protocol, the pial region response did not drop back to baseline within the timeframe of the experiment, and the offset times in cortex, deep gray and major vessels were not significantly different. Discussion We have shown that contrast-enhanced high-frequency ultrasound imaging can be used for studying the cerebral blood volume response to neuronal stimulation. With an in-plane spatial resolution of around 100 μm, a slice thickness down to 600 μm (dependent on the distance

105

between the region of interest and the transducer surface), and a temporal resolution of 40 ms, functional micro-ultrasound imaging or fMUS opens new possibilities for investigation of the hemodynamics of rodent cerebral vasculature in health and disease. Two-photon fluorescence microscopy (2 PM) has unsurpassed spatial resolution, but fMUS has a much deeper penetration depth and a higher temporal resolution. Compared to fMRI, fMUS has a comparable spatial and much better temporal resolution, but as of yet lacks the capability of imaging the whole brain. Extension to 3D imaging is currently under development. Like 2 PM but unlike fMRI, fMUS as presently implemented requires the opening of a cranial window. An advantage of fMUS over both other modalities is that fMUS is relatively inexpensive: at the time of this writing, the cost of the micro-ultrasound system used here is less than a quarter of the price of a high-end commercial two-photon microscope, and approximately 10 × less than a high-field small-animal MRI system. Furthermore, ultrasound systems do not require dedicated technicians to operate and maintain the system. The analysis performed in this work represents the first exploration of the data produced by this method. Since the contrast mechanism employed by fMUS differs from that of fMRI, the pre-processing, analysis and post-processing algorithms that have been developed in the fMRI literature are not necessarily the most appropriate or the most fruitful in the case of fMUS. In addition to the positive activation patterns explored in this paper, we have also observed regions with negative correlations to the stimulus paradigm. These will be reported on in future work. The rich dataset produced by high frequency ultrasound imaging gives insight into the topography of blood vessels through the nonlinear contrast average intensity projection, maps the local blood volume through nonlinear contrast mode and power Doppler mode, reveals both direction and velocity of flow in individual vessels via color Doppler imaging, and yields maps of blood flow through contrast agent disruption/replenishment (Sun et al., 2010). In disruption-replenishment mode, the imaging plane is cleared of contrast agent by a disruption pulse, and the time course of the replenishment of the imaging plane with contrast agent provides a voxel-by-voxel measure of CBF with the same spatial resolution as the CBV measurements. Imaging parameters of functional micro-ultrasound imaging The temporal resolution of micro-ultrasound can be as high as 500– 1000 fps, limited physically by the propagation velocity of the acoustic signal in tissue and the necessity to wait for the current pulse-echo to be received before transmission of the next one. In practice, the maximum achievable temporal resolution is restricted by memory and data transfer rate limits given the desired matrix size and acquisition period duration. With current technology, these restrictions results in acquisitions of 1000 frames per run; our 10-second stimulus protocol was recorded at 10 frames per second and our 2-second stimulus protocol at 40 fps. The contrast-to-noise ratio of the CBV responses measured using fMUS depends on the duration of the protocol and on whether the region of interest consists mainly of parenchymal versus intravascular voxels, and ranges between 1.2 and 4.5. The spatial resolution of an ultrasound imaging system generally depends on the position in the field of view (Cobbold, 2008, par. 8.3). In the axial direction of the ultrasound beam, which in our scanning geometry corresponds to the superior-to-inferior direction, resolution depends primarily on the ultrasound frequency and pulse length and is nearly constant at ~80 μm. The lateral resolution of the transducer (anatomically, the posterior-to-anterior direction) is also nearly constant across the field of view at about 160 μm. In-plane spatial resolution can be increased at the expense of penetration depth by using a transducer with a higher ultrasound frequency. The physical resolution in the medial-to-lateral direction is the imaging slice thickness. As noted in the Materials and methods section, the slice thickness is largest in the cortex and smaller in the deep regions of the image. The higher sensitivity associated with a narrower beam profile

106

M.E. van Raaij et al. / NeuroImage 58 (2011) 100–108

Fig. 6. Temporal parameters of CBV response to stimulation in different regions. (A and D) Onset time, (B and E) time to peak, (C and F) peak amplitude, and (G) offset time for 10second stimulation (A-C) and 2-second stimulation protocols (D-G). All bars represent average ± standard deviation across N = 11 (long protocol) and N = 8 (short protocol) subjects. Offset time could not be determined for the long stimulus protocol and the pial region in the short stimulus protocol because the response did not return to baseline within the inter-stimulus period.

deeper in the brain keeps the signal-to-noise ratio more or less constant across depth despite the more severe attenuation experienced by an ultrasound wave traveling to and from a deeper tissue layer. The contrast-to-noise ratio of the CBV response timetraces reported in this work depends mainly on the number of activated voxels in each region, which tends to be larger in the cortical than in the deep gray regions. The physiological limitations, such as potential brain swelling caused by the need for a cranial window may be alleviated by closing the window. However, it is not possible to use a microscope coverslip or other glass window because glass reflects most of the incident ultrasound energy. A better choice of window material may be rexolite or mylar, which provide some structural stability while still being reasonably transparent to acoustic waves. We explored these materials and found they produced an average signal intensity drop of ~30% and introduced reflection artifacts. Other alternatives include the use of mice, whose skull is considerably thinner, and/or a skull-thinning preparation, either mechanically (Drew et al., 2010) or by laser ablation (Vitek et al., 2010). Under certain conditions, ultrasound imaging in the brain in the presence of microbubble contrast agents can lead to transient disruption of the blood–brain barrier (BBB) (Hynynen, 2008). This phenomenon has been investigated extensively at clinical ultrasound frequencies (up to 2 MHz), but not at the high frequencies employed in this work. Lowfrequency studies on rabbit brain (McDannold et al., 2008) suggest that at a threshold mechanical index (MI) of 0.46 and higher, transient BBB disruption occurs in 50% of cases, and that this threshold is constant with frequency in the range of fc = 0.2 to 2 MHz. If this relationship can be extrapolated to fc = 18 MHz, our peak negative pressure of ~1.8 MPa and MI of ~0.4 would be close to that threshold. However, a critical parameter for BBB disruption is the pulse length, which in our case is very short (4 cycles at 18 MHz is 0.2 μs) compared to the pulse lengths typically used to purposely induce BBB disruption, which are on the order of 10 ms. For this reason, supported by the lack of observable microbubble extravasation in the course of our experiments, we assume that our imaging protocol does not induce BBB disruption. We note that if BBB disruption was desired, for example to study the immediate effects of local drug or gene delivery (Hynynen, 2008) on the functional

CBV response to stimulation, ultrasound insonation parameters could be chosen that would likely induce a transient opening of the BBB. Magnitude of the CBV response The estimation of stimulation-induced changes in CBV using fMUS bears similarities to that using fMRI. Both modalities use contrast agents that are strictly intraplasmatic in the presence of an intact blood–brain barrier and both provide a relative (as opposed to absolute) measure of CBV. Notable differences include the use of a continuous infusion of contrast agent in fMUS versus a bolus injection of contrast agent that is typical of fMRI (Belliveau et al., 1991). The signal in fMUS is more directly related to the CBV itself than the signal in fMRI because in fMUS, a local increase in the amount of microbubbles directly increases the amount of backscattered ultrasound and thus the measured signal from the vasculature only, whereas in fMRI a local increase in the amount of contrast agent affects the effective magnetic field experienced by, and thus the signal recorded from, proton spins in the surrounding tissue regions. We observed robust stimulation-induced changes both cortically and subcortically. Given the stereotaxic positioning of the imaging plane, the strongest response to electrical stimulation of the forepaw in most subjects takes place in the forelimb region of the primary somatosensory cortex (S1FL), with additional smaller activated regions mainly localized in the caudate putamen (CPu). This global pattern of activation measured using fMUS corresponds coarsely to that observed by (Keilholz et al., 2006) using CBV-fMRI. Whereas the presence of activation due to extraneous stimuli can never be ruled out in functional brain imaging experiments, our observations in multiple subjects (N = 13 different rats, of which 11 showed activation in the long stimulus protocol and 8 in the short protocol) gives us confidence that fMUS allows robust detection of activation elicited by electrical forepaw stimulation. Three-dimensional functional imaging, which is currently under development, will allow for improved localization of the activated regions. The observed difference in cortical vs subcortical ΔCBV likely results from a very different microvascular topology in the two regions (as reported in the literature and readily observable using 3D Power

M.E. van Raaij et al. / NeuroImage 58 (2011) 100–108

Doppler imaging, a subject of a separate manuscript) and hence likely distinct neurovascular coupling, rather than being a direct reflection of relative magnitudes in the neuronal responses in the respective areas (Sanganahalli et al., 2009). Furthermore, it is important to note that the validity of the cortical vs. subcortical signal change comparison critically depends on the present focus on relative rather than absolute CBV changes, as signal attenuation increases with distance from the transducer. The reported relative changes in the cortical cerebral blood volume lie well within the range of forepaw stimulation-induced CBV increases in S1FL of alpha-chloralose anesthetized rats (Mandeville et al., 1998; Herman et al., 2009; Sanganahalli et al., 2009; Hirano et al., 2011). Similarly, the magnitude of subcortical CBV increases is in general agreement with prior MRI measurements, following 45-s of the same functional challenge (Keilholz et al., 2006). Clearly, this concordance is not meant as a validation, but as an illustration that fMUS produces ΔCBV estimates in the physiological range. As with forementioned studies, the contrast agent currently employed distributes evenly in blood plasma, so that signal changes reflect adjustments in the cerebral plasma volume (CPV). Notwithstanding, relative changes in CPV and CBV are equal for constant hematocrit, an assumption that is frequently made in these studies and indeed supported by multitracer studies in rodent brain during hypercarbia (Todd et al., 1993) and more recent MRI and laser Doppler flowmetry measurements of CPV and RBC volume changes following forepaw stimulation (Herman et al., 2009).

107

izing cortical hemodynamics in rodent somatosensory vs. cat visual cortices. Conclusion This work introduces high frequency microultrasound in combination with exogenous nonlinear contrast agents as a new, cost-effective and highly accessible modality for imaging of brain function in preclinical animal models. The potential of this methodology is demonstrated by interrogation of functional stimulation elicited changes following forepaw stimulation in anesthetized rats prepared with a cranial window. The technique offers the potential for whole brain coverage at 100 μm and 10s of ms resolution and, when combined with other readily available acoustic contrast modes, offers a rich characterization of both microvascular topology and hemodynamics in vivo. Future developments include using the arterio-venous identification from color Doppler images as a region mask for the functional images and studying differential CBV changes in arterial vs. venous vessels as a function of the functional challenge parameters. Furthermore, we are currently working on extending the modality to 3D, which will result in a better localization of activated areas and improved anatomical registration as well as allow probing of connectivity of distant brain regions.

Temporal evolution of the CBV response

Disclosure

The subsecond cortical onset times observed with the 2-s stimulus duration were similar to those reported recently with fMRI (Hirano et al., 2011). The cortical onset times for the 10-s stimulus protocol are in agreement with the prior estimates of total hemoglobin concentration changes in response to the same function stimulus, measured using diffuse optical tomography (~5-s TTP in 30-s stimulation paradigm in (Siegel et al., 2003)). In the long stimulation paradigm, cortical OT was shorter than that in pial region, consistent with the notion of propagation of the CBV response from the middle of the cortex outward (Hirano et al., 2011). Further, parenchymal OTs were shorter than those in the neighboring large vessels, in support of the hemodynamic response originating in the tissue and spreading to the major vessels (Yu et al., 2010). While subcortical neuronal activity precedes that in S1FL, we were not able to detect differences between subcortical and cortical OTs, at least in part reflective of SNR limitations (which likely also precluded detection of inter-regional OT variation in the 2-s stimulation protocol) and significantly lower CNR of subcortical responses, as noted earlier. As expected, time-to-peak correlated strongly with the duration of the stimulus. Further, the average cortical TTP for the short stimulation protocol agrees very well with those observed using fMRI for the same functional protocol (Hirano et al., 2011) and slightly longer functional protocols (3 s TTP for 6-s stimulus duration in (Marota et al., 1999)). As with OTs, the longer TTP of pial vessel region is consistent with the delayed superficial relative to parenchymal dilatation. In agreement with previous MRI studies and given the relatively long stimulus duration (2-s) of even the brief stimulation protocol, the CBV response exhibited a slow return to baseline following cessation of the stimulus (Culver et al., 2005; Keilholz et al., 2006; Kida et al., 2007; Lu et al., 2007; Silva et al., 2007; Hirano et al., 2011), presumably due to delayed compliance of intracortical drainage vessels (Mandeville et al., 1998; Hirano et al., 2011). The slow return of CBV to prestimulus baseline level was characteristic of all regions considered: we did not observe the fast return to baseline of the CBV signal in the pial vessel region reported previously in isoflurane anesthetized cats, in response to visual stimulation (Yacoub et al., 2006). Moreover, we found no evidence of a post-stimulus undershoot in CBV in any of the stimulation protocols, in contrast to earlier findings in isoflurane anesthetized cats in response to visual stimulation (Kim and Kim, 2010). We believe both of these to stem from differences in the temporal parameters character-

F. Stuart Foster is a consultant to VisualSonics Inc., Toronto, Canada. 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 his contributions to preliminary experiments, and Afonso Silva for a critical reading of the manuscript. 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 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. References 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. Bar, T., 1981. Distributions of radially penetrating arteries and veins in the neocortex of rat. In: Cervos-Navarro, J., Fritschka, E. (Eds.), Cerebral Microcirculation and Metabolism. Raven Press, New York. Becher, H., Burns, P.N., 2000. Handbook of Contrast Echocardiography. Springer, Berlin. Belliveau, J.W., Kennedy Jr., D.N., McKinstry, R.C., Buchbinder, B.R., Weisskoff, R.M., Cohen, M.S., Vevea, J.M., Brady, T.J., Rosen, B.R., 1991. Functional mapping of the human visual cortex by magnetic resonance imaging. Science 254, 716–719. Cobbold, R., 2008. Foundations of Biomedical Ultrasound. Oxford University Press, New York. Cox, R.W., 1996. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput. Biomed. Res. 29, 162–173. Culver, J.P., Siegel, A.M., Franceschini, M.A., Mandeville, J.B., Boas, D.A., 2005. Evidence that cerebral blood volume can provide brain activation maps with better spatial resolution than deoxygenated hemoglobin. NeuroImage 27, 947–959. 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. 4, 863–872. Drew, P.J., Shih, A.Y., Driscoll, J.D., Knutsen, P.M., Blinder, P., Davalos, D., Akassoglou, K., Tsai, P.S., Kleinfeld, D., 2010. Chronic optical access through a polished and reinforced thinned skull. Nat. methods 7, 981–984. 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.

108

M.E. van Raaij et al. / NeuroImage 58 (2011) 100–108

Foster, F.S., Mehi, J., Lukacs, M., Hirson, D., White, C., Chaggares, C., Needles, A., 2009. A new 15–50 MHz array-based micro-ultrasound scanner for preclinical imaging. Ultrasound Med. Biol. 35, 1700–1708. Genovese, C.R., Lazar, N.A., Nichols, T., 2002. Thresholding of statistical maps in functional neuroimaging using the false discovery rate. NeuroImage 15, 870–878. 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. Goessling, W., North, T.E., Zon, L.I., 2007. Ultrasound biomicroscopy permits in vivo characterization of zebrafish liver tumors. Nat. Methods 4, 551–553. Herman, P., Sanganahalli, B.G., Hyder, F., 2009. Multimodal measurements of blood plasma and red blood cell volumes during functional brain activation. J. Cerebr. Blood F. Met. 29, 19–24. Hirano, Y., Stefanovic, B., Silva, A.C., 2011. Spatiotemporal evolution of the functional magnetic resonance imaging response to ultrashort stimuli. J. Neurosci. 31, 1440–1447. Hynynen, K., 2008. Ultrasound for drug and gene delivery to the brain. Adv. Drug Deliv. Rev. 60, 1209–1217. Iadecola, C., 2004. Neurovascular regulation in the normal brain and in Alzheimer's disease. Nat. Rev. 5, 347–360. Ino, T., Akimoto, K., Ohkubo, M., Nishimoto, K., Yabuta, K., Takaya, J., Yamaguchi, H., 1996. Application of percutaneous transluminal coronary angioplasty to coronary arterial stenosis in Kawasaki disease. Circulation 93, 1709–1715. Keilholz, S.D., Silva, A.C., Raman, M., Merkle, H., Koretsky, A.P., 2006. BOLD and CBVweighted functional magnetic resonance imaging of the rat somatosensory system. Magn. Reson. Med. 55, 316–324. Kida, I., Rothman, D.L., Hyder, F., 2007. Dynamics of changes in blood flow, volume, and oxygenation: implications for dynamic functional magnetic resonance imaging calibration. J. Cerebr. Blood F. Met. 27, 690–696. Kiguchi, K., Ruffino, L., Kawamoto, T., Franco, E., Kurakata, S., Fujiwara, K., Hanai, M., Rumi, M., DiGiovanni, J., 2007. Therapeutic effect of CS-706, a specific cyclooxygenase-2 inhibitor, on gallbladder carcinoma in BK5.ErbB-2 mice. Mol. Cancer Ther. 6, 1709–1717. Kim, T., Kim, S.-G., 2010. Temporal dynamics and spatial specificity of arterial and venous blood volume changes during visual stimulation: implication for BOLD quantification. J. Cerebr. Blood F. Met. doi:10.1038/jcbfm.2010.226. 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. Lee, D.J., Lyshchik, A., Huamani, J., Hallahan, D.E., Fleischer, A.C., 2008. Relationship between retention of a vascular endothelial growth factor receptor 2 (VEGFR2)targeted ultrasonographic contrast agent and the level of VEGFR2 expression in an in vivo breast cancer model. J. Ultrasound Med. 27, 855–866. Lindauer, U., Villringer, A., Dirnagl, U., 1993. Characterization of CBF response to somatosensory stimulation: model and influence of anesthetics. Am. J. Physiol. Heart Circ. Physiol. 264, H1223–H1228. Lindvere, L., Dorr, A., Stefanovic, B., 2010. Two-photon fluorescence microscopy of cerebral hemodynamics. Cold Spring Harb Protoc. doi:10.1101/pdb.prot5494. Liu, J., Du, J., Zhang, C., Walker, J.W., Huang, X., 2007. Progressive troponin I loss impairs cardiac relaxation and causes heart failure in mice. Am. J. Physiol. Heart Circ. Physiol. 293, H1273–H1281. Lu, H., Scholl, C.A., Zuo, Y., Stein, E.A., Yang, Y., 2007. Quantifying the blood oxygenation level dependent effect in cerebral blood volume-weighted functional MRI at 9.4 T. Magn. Reson. Med. 58, 616–621. Mandeville, J.B., Marota, J.J.A., Kosofsky, B.E., Keltner, J.R., Weissleder, R., Rosen, B.R., Weisskoff, R.M., 1998. Dynamic functional imaging of relative cerebral blood volume during rat forepaw stimulation. Magn. Reson. Med. 39, 615–624. Marota, J.J.A., Ayata, C., Moskowitz, M.A., Weisskoff, R.M., Rosen, B.R., Mandeville, J.B., 1999. Investigation of the early response to rat forepaw stimulation. Magn. Reson. Med. 41, 247–252. McDannold, N., Vykhodtseva, N., Hynynen, K., 2008. Blood–brain barrier disruption induced by focused ultrasound and circulating preformed microbubbles appears to be characterized by the mechanical index. Ultrasound Med. Biol. 34, 834–840. McHedlishvili, G., Kuridze, N., 1984. The modular organization of the pial arterial system in phylogeny. J. Cereb. Blood Flow Metab. 4, 391–396. Mosso, A., 1881. Ueber den Kreislauf des Blutes im Menschlichen Gehirn. Verlag Von Veit & Comp, Leipzig. Mulvagh, S.L., et al., 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 arraybased micro-ultrasound system. Ultrasound Med. Biol. 36, 2097–2106.

Ngai, A.C., Meno, J.R., Jolley, M.A., Winn, H.R., 1998. Suppression of somatosensory evoked potentials by nitric oxide synthase inhibition in rats: methodological differences. Neurosci. Lett. 245, 171–174. Olive, K.P., Tuveson, D.A., 2006. The use of targeted mouse models for preclinical testing of novel cancer therapeutics. Clin. Cancer Res. 12, 5277–5287. Paxinos, G., Watson, C., 2007. The Rat Brain in Stereotaxic Coordinates6th ed. Elsevier Academic Press, San Diego. Paxinos, G., Watson, C. (Eds.), 2009. Brain Navigator™ Release 2.0. Elsevier, Boston, MA. Rumack, C.M., Wilson, S.R., Charboneau, J.W. (Eds.), 1998. Diagnostic Ultrasound. Mosby, New York. Sanganahalli, B.G., Herman, P., Blumenfeld, H., Hyder, F., 2009. Oxidative neuroenergetics in event-related paradigms. J. Neurosci. 29, 1707–1718. Shung, K., 2005. Diagnostic Ultrasound: Imaging and Blood Flow Measurements. CRC Press, Boca Raton FL USA, pp. 72–73. Siegel, A.M., Culver, J.P., Mandeville, J.B., Boas, D.A., 2003. Temporal comparison of functional brain imaging with diffuse optical tomography and fMRI during rat forepaw stimulation. Phys. Med. Biol. 48, 1391–1403. Silva, A.C., Lee, S.P., Yang, G., Iadecola, C., Kim, S.G., 1999. Simultaneous blood oxygenation level-dependent and cerebral blood flow functional magnetic resonance imaging during forepaw stimulation in the rat. J. Cereb. Blood Flow. Metab. 19, 871–879. Silva, A.C., Koretsky, A.P., Duyn, J.H., 2007. Functional MRI impulse response for BOLD and CBV contrast in rat somatosensory cortex. Magn. Reson. Med. 57, 1110–1118. Stefanovic, B., Bosetti, F., Silva, A.C., 2006. Modulatory role of cyclooxygenase-2 in cerebrovascular coupling. NeuroImage 32, 23–32. Sun, J., Lindvere, L., van Raaij, M.E., Dorr, A., Stefanovic, B., Foster, F.S., 2010. In vivo imaging of cerebral hemodynamics using high-frequency micro-ultrasound. Cold Spring Harb Protoc. doi:10.1101/pdb.prot5495. Todd, M.M., Weeks, J.B., Warner, D.S., 1993. Microwave fixation for the determination of cerebral blood-volume in rats. J. Cerebr. Blood F. Met. 13, 328–336. Trivedi, C.M., Luo, Y., Yin, Z., Zhang, M., Zhu, W., Wang, T., Floss, T., Goettlicher, M., Noppinger, P.R., Wurst, W., Ferrari, V.A., Abrams, C.S., Gruber, P.J., Epstein, J.A., 2007. Hdac2 regulates the cardiac hypertrophic response by modulating Gsk3 beta activity. Nat. Med. 13, 324–331. Ueki, M., Mies, G., Hossmann, K.A., 1992. Effect of alpha-chloralose, halothane, pentobarbital and nitrous oxide anesthesia on metabolic coupling in somatosensory cortex of rat. Acta Anaesthesiol. Scand. 36, 318–322. Vitek, D.N., Adams, D.E., Johnson, A., Tsai, P.S., Backus, S., Durfee, C.G., Kleinfeld, D., Squier, J.A., 2010. Temporally focused femtosecond laser pulses for low numerical aperture micromachining through optically transparent materials. Opt. Expr. 18, 18086–18094. Wirtzfeld, L.A., Wu, G., Bygrave, M., Yamasaki, Y., Sakai, H., Moussa, M., Izawa, J.I., Downey, D.B., Greenberg, N.M., Fenster, A., Xuan, J.W., Lacefield, J.C., 2005. A new three-dimensional ultrasound microimaging technology for preclinical studies using a transgenic prostate cancer mouse model. Cancer Res. 65, 6337–6345. Wirtzfeld, L.A., Graham, K.C., Groom, A.C., Macdonald, I.C., Chambers, A.F., Fenster, A., Lacefield, J.C., 2006. Volume measurement variability in three-dimensional highfrequency ultrasound images of murine liver metastases. Phys. Med. Biol. 51, 2367–2381. Woolsey, T.A., Rovainen, C.M., Cox, S.B., Henegar, M.H., Liang, G.E., Liu, D., Moskalenko, Y.E., Sui, J., Wei, L., 1996. Neuronal units linked to microvascular modules in cerebral cortex: response elements for imaging the brain. Cereb. Cortex 6, 647–660. Xuan, J.W., Bygrave, M., Jiang, H., Valiyeva, F., Dunmore-Buyze, J., Holdsworth, D.W., Izawa, J.I., Bauman, G., Moussa, M., Winter, S.F., Greenberg, N.M., Chin, J.L., Drangova, M., Fenster, A., Lacefield, J.C., 2007. Functional neoangiogenesis imaging of genetically engineered mouse prostate cancer using three-dimensional power Doppler ultrasound. Cancer Res. 67, 2830–2839. Yacoub, E., Ugurbil, K., Harel, N., 2006. The spatial dependence of the poststimulus undershoot as revealed by high-resolution BOLD- and CBV-weighted fMRI. J. Cerebr. Blood F. Met. 26, 634–644. Yu, X., Dodd, S., Hirano, Y., Glen, D., Saad ZS, R.C.R., Silva, A.C., Koretsky, A.P., 2010. Mapping the early spatiotemporal BOLD fMRI response in the barrel cortex of rats. Annual Meeting of the International Society for Magnetic Resonance in Medicine 2010. Zhou, Y.Q., Foster, F.S., Parkes, R., Adamson, S.L., 2003. Developmental changes in left and right ventricular diastolic filling patterns in mice. Am. J. Physiol. Heart Circ. Physiol. 285, H1563–H1575. Zhou, Y.Q., Foster, F.S., Nieman, B.J., Davidson, L., Chen, X.J., Henkelman, R.M., 2004. Comprehensive transthoracic cardiac imaging in mice using ultrasound biomicroscopy with anatomical confirmation by magnetic resonance imaging. Physiol. Genomics 18, 232–244.