NeuroImage 46 (2009) 658–664

Contents lists available at ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r.com/ locate /yn img

Microbubbles as a novel contrast agent for brain MRI

Jerry S. Cheung a,b, April M. Chow a,b, Hua Guo a,b, Ed X. Wu a,b,⁎a Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Pokfulam, Hong Kong, Chinab Department of Electrical and Electronic Engineering, The University of Hong Kong, Pokfulam, Hong Kong, China

⁎ Corresponding author. Laboratory of Biomedical ImaUniversity of Hong Kong, Hong Kong, China.

E-mail address: ewu@eee.hku.hk (E.X. Wu).

1053-8119/$ – see front matter © 2009 Elsevier Inc. Aldoi:10.1016/j.neuroimage.2009.02.037

a b s t r a c t

a r t i c l e i n f o

Article history:Received 16 January 2009Revised 19 February 2009Accepted 24 February 2009Available online 6 March 2009

Gas-filled microbubbles have the potential to become a unique MR contrast agent due to their magneticsusceptibility effect, biocompatibility and localized manipulation via ultrasound cavitation. In this study, twotypes of microbubbles, custom-made albumin-coated microbubbles (A-MB) and a commercially availablelipid-based clinical ultrasound contrast agent (SonoVue®), were investigated with in vivo dynamic brain MRIin Sprague–Dawley rats at 7 T. Microbubble suspensions (A-MB: 0.2 mL of ∼4% volume fraction; SonoVue®:0.2 mL of ∼3.5% volume fraction) were injected intravenously. Transverse relaxation rate enhancements(ΔR2⁎) of 2.49±1.00 s−1 for A-MB and 2.41±1.18 s−1 for SonoVue® were observed in the brain (N=5).Brain ΔR2⁎ maps were computed, yielding results similar to the cerebral blood volume maps obtained with acommon MR blood pool contrast agent. Microbubble suspension ΔR2⁎ was measured for different volumefractions. These results indicate that gas-filled microbubbles can serve as an intravascular contrast agent forbrain MRI at high field. Such capability has the potential to lead to real-time MRI guidance in variousmicrobubble-based drug delivery and therapeutic applications in the central nervous system.

© 2009 Elsevier Inc. All rights reserved.

Introduction

MRI provides superb soft tissue contrast with high spatialresolution when compared with other imaging modalities. While MRimage contrast can be flexibly controlled by varying pulse sequencesand parameters, it is determined by the intrinsic tissue properties suchas proton density, longitudinal relaxation time (T1) and transverserelaxation time (T2). At present, the exogenous contrast agentsavailable for brainMRImainly fall into three categories, i.e. gadoliniumchelates, manganese chelates, and superparamagnetic iron oxideparticles. Their effects are usually described by longitudinal relaxationrate (R1) and transverse relaxation rate (R2/R2⁎), where R1, R2 and R2⁎

are defined as 1/T1,1/T2 and 1/T2⁎ respectively. Susceptibility contrastagents exhibit large R2⁎/R1 ratios and predominantly induce signalloss through spin dephasing by strong magnetic susceptibility effects.Their T2⁎ shortening effects are usually much stronger than thebaseline T2 effects. Dynamic susceptibility contrast MRI is an effectivetool to measure cerebral blood volume and perfusion (Belliveau et al.,1990). Gadolinum chelates have been widely used as susceptibilitycontrast agents to assess vascular characteristics through dynamicimaging by investigating first-pass effects of bolus injections in brain(Rempp et al., 1994; Rosen et al., 1991), while superparamagnetic ironoxide particles have also been used for the same purpose by utilizing

ging and Signal Processing, The

l rights reserved.

both thefirst-pass (Simonsen et al.,1999) and steady-state effects (vanBruggen et al., 1998; Wu et al., 2003, 2004a, b, c, d).

Gas-filled microbubbles were originally developed as an intravas-cular contrast agent to enhance backscattering in ultrasound imaging.Microbubbles can potentially be used as a MR susceptibility contrastagent in vivo due to the induction of large local magnetic susceptibilitydifferences by the gas–liquid interface. Moreover, microbubbles can belocally cavitated by spatially focused ultrasound (Bouakaz et al.,2005), and hence the MR signals can be temporally and spatiallymanipulated by external ultrasound irradiation because microbubbledestruction will diminish the susceptibility effect. Due to their uniquecavitation (Liu et al., 2006; Unger et al., 2001) and sonoporation(Mehier-Humbert et al., 2005; Wu et al., 2006) properties, gas-filledmicrobubbles play an expanding role in therapeutic applications. Site-specific release of incorporated drugs or genes inside microbubblescan be potentially achieved by local microbubble cavitation, whilemicrobubble-mediated sonoporation dramatically increases cell per-meability and intracellular uptake. Microbubble-mediated therapyhas been used to deliver genes or drugs to specific tissues (Bekeredjianet al., 2003; Hauff et al., 2005; Shimamura et al., 2004; Taniyama et al.,2002) utilizing microbubble cavitation and sonoporation effects,including neural tissues, skeletal muscles, myocardium, kidneys,vessels, and tumors. Furthermore, microbubble cavitation phenom-enon has been put into practical use in achieving several therapeuticinterventions. Sonothrombolysis, which employs the local shockwaves produced by microbubble cavitation to fragment clots onmicroscopic scale and restores blood flow, has been developed as aminimally invasive recanalization technique in treating vascular

659J.S. Cheung et al. / NeuroImage 46 (2009) 658–664

thrombosis (Culp et al., 2004; Daffertshofer and Hennerici, 2003).Besides, transient opening of blood–brain barrier through micro-bubble-enhanced sonoporation can be accomplished by applyingtranscranial ultrasound with intravenous injection of microbubbleswithout damaging the neurons (Sheikov et al., 2004). Delivery of bothlow and high molecular weight therapeutic compounds to the centralnervous system can be potentially attained through this noninvasiveprocedure. Microbubbles are also used to enhance the noninvasivehigh intensity focused ultrasound therapy by increasing the localheating rate (Kaneko et al., 2005).

Early experiment with Albunex®, an ultrasound contrast agentconsisting of air-filled microbubbles coated with human albuminshells, suggested the potential of air-filled microbubbles as a MRsusceptibility contrast agent for tumor imaging (Moseley et al., 1991).Feasibility of microbubbles as a MR pressure sensor, based on thesusceptibility change caused by pressure-induced microbubble sizechange, has been explored through theoretical (Dharmakumar et al.,2002) and phantom (Alexander et al., 1996) studies. The first in vivoinvestigation of susceptibility contrast induced by microbubbles wasreported in liver by our group previously using Optison®, micro-bubbles of human albumin shells with perfluorocarbon as core gas, at7 T in rat liver (Wong et al., 2004). Dependency of R2⁎ on microbubblevolume fractions was also reported using Levovist®, air-filled micro-bubbles with palmitic acid shells, through a phantom study at 1.5 T(Ueguchi et al., 2006). Magnetic susceptibility enhancements inducedby gas–liquid interface were demonstrated recently by simulationsand MR experiments using air-filled cylinders in water (De Guio et al.,2008), consolidating the feasibility of gas-filled microbubbles as a MRsusceptibility contrast agent. In this study, we aim to furtherinvestigate and demonstrate the in vivo MR susceptibility effectinduced by both custom-made albumin-coated microbubbles andcommercially available lipid-based microbubbles in rat brain at 7 Tusing dynamic susceptibility weighted MRI.

Methods

All MRI experiments were performed on a 7 T MRI scanner with amaximum gradient of 360 mT/m (70/16 PharmaScan, Bruker BiospinGmbH, Germany). All animal experiments were approved by the localinstitutional animal ethics committee.

Microbubble preparation

Two types of gas-filled microbubbles, custom-made albumin-coatedmicrobubbles (A-MB) and a commercially available clinical ultrasoundcontrast agent (SonoVue®, Bracco, Milan, Italy), were used in this study.SonoVue®microbubbles consist of sulphurhexafluoride gas stabilized inan aqueous dispersion by a phospholipid monolayer. Air-filled A-MBwere produced by sonication aspreviously described (Cernyet al.,1990).Briefly, a 5% solution of bovine serum albumin (10857, USB Corporation,Cleveland, OH, USA) was preheated to about 70 °C and sonicated usingultrasound frequency of 20 kHz. Concentrated microbubble suspensionwas achieved by draining off excess albumin solution below the floatingmicrobubbles. Micrographs of A-MB and SonoVue® were taken usinglight microscope. To estimate microbubble diameter distribution,micrographs were analyzed using ImageJ (National Institute of Health,USA) for histogram of size distribution.

Measurement of ΔR2⁎ of albumin-coated and SonoVue® microbubblesuspensions

Microbubble phantom study was performed with 23-mm quad-rature resonator for radiofrequency (RF) transmission and receiving.Microbubbles were diluted from a well-mixed microbubble suspen-sion to three different concentrations (8%, 4% and 2% volume fractionsfor A-MB; 4%, 2% and 1% volume fractions for SonoVue®) with the

addition of saline, and were then placed in separate 4-cm long, 1-cmdiameter, 2-mL cylindrical phantom tubes. Fresh microbubble vialswere used to make the phantoms. Each phantom tube was slowlywarmed to room temperature and gently mixed for 2 min outside themagnet prior to MR measurements. To ensure uniform suspension ofmicrobubbles, the phantomwas then continuously stirred by rotationinside the magnet. It was then arrested in horizontal positionimmediately prior to MRI.

Apparent transverse relaxation rate enhancement (ΔR2⁎) wasmeasured by acquiring multi-echo gradient-echo (GE) signals con-tinuously without phase encoding for 2 min from an axial 1-mm sliceatmiddle of the phantom. Themeasurementwas repeated three timesfor each microbubble phantom. The parameters were repetition time(TR)=1000 ms, echo time (TE)=5, 10, 15, 20, 25, 30, 35, 40 ms, flipangle (FA)=30° and number of signal averages (NEX)=1. Initially,there was a uniform suspension of microbubbles. As GE signals wereacquired, microbubbles started to migrate upward; therefore, in thefinal state the microbubbles aggregated in the upper part of the tube.Phantom R2⁎ values were computed by monoexponential fitting ofthe peak magnitudes of the multi-echo GE signals using a softwaretoolkit developed in MATLAB (MathWorks, Inc., Nitick, MA, USA).Microbubble-induced ΔR2⁎ was then calculated as the differencebetween R2⁎ in the initial state and that in the final state.

Animal preparation

Ten normal adult Sprague–Dawley rats (∼250–350 g) were used inthis study. Each rat was anesthetized with intraperitoneal injection ofketamine (100 mg/kg) and xylazine (10 mg/kg). Femoral veincatheterization was performed with a 1-m long tube (MicroRe-nathane®, MRE-0.25, Braintree Scientific, Inc., Braintree, MA) con-nected to a 27-gauge needle. The dead space in the catheter was about0.2 mL. During imaging, animals were anesthetized with isoflurane/air using 1.0–1.5% via a nose cone with respiratory monitoring.

Administration of microbubble suspensions

Microbubbles were first warmed slowly to room temperature. A-MB suspension was resuspended by inversion and rotation for 2 minuntil a homogenous milky-white suspension formed. For SonoVue®,1.2 mL sodium chloride 0.9% w/v solution was used for dispersion ofthe powder in the vial. Resuspension was performed by vortexing for∼20 s until a homogenous milky-white suspension formed. For eachimaging session, 0.2 mL of microbubble suspension (of ∼4% volumefraction for A-MB and ∼3.5% volume fraction for SonoVue®) wasslowly injected (over ∼10 s) into femoral vein at a rate of 1.2 mL/minto avoid possible microbubble destruction due to high pressure andshear stress.

Susceptibility contrast imaging of rat brain

In vivo brain imaging was performed in five rats in supine positionusing a 38-mm quadrature resonator for RF transmission and receiving.High resolution anatomical images were acquired with two-dimen-sional (2D) fast low-angle shot (FLASH) sequence using TR=500 ms,TE=10 ms, FA=30°, field of view (FOV)=29 mm×29 mm, slicethickness=0.7 mm, acquisition matrix=256×256, spatialresolution=0.11×0.11×0.7 mm3 and NEX=4. Dynamic susceptibilityweighted MRI was performed with single-shot GE echo-plannerimaging (GE-EPI) sequence using TR=1000 ms, TE=30 ms, FA=90°,FOV=29 mm×29 mm, slice thickness=0.7 mm, acquisitionmatrix=96×96, spatial resolution=0.3×0.3×0.7 mm3, samplingbandwidth (BW)=221 kHz and NEX=1. Microbubble suspensionwas injected about 5min after the start of dynamic imaging. Aminimumlapse of 10 min was used to ensure sufficient clearance of themicrobubbles before the next injection. The susceptibility effect of

Fig. 1. Representative light micrograph (magnification=200×) of (a) albumin-coated microbubble (A-MB) and (b) SonoVue® microbubble suspensions. Histograms showingdiameter distribution for a representative batch of (c) A-MB and (d) SonoVue®. The mean diameters of A-MB and SonoVue® microbubbles are 9.21 μm and 2.95 μm, respectively.

Fig. 2. (a) Dynamic T2⁎-weighted GE-EPI images of an A-MB suspension phantomillustrating the upwards flotation of microbubbles from the initial well-suspended stateto thefinal state. (b) Signal time coursewith the circularmeasurement ROI shown in (a).

660 J.S. Cheung et al. / NeuroImage 46 (2009) 658–664

microbubbles was compared with that of a well-established intravas-cular contrast agent, monocrystalline iron oxide nanoparticles (MION;MGH Center of Molecular Imaging Research, MA), in brain by a singledose of 0.6 mg Fe/kg injection using identical injection protocol andimaging sequence.

In vivo data analysis

GE-EPI images were first co-registered using AIR5.2.5 (Woods etal., 1998). ΔR2⁎ maps were computed on a pixel-by-pixel basis asΔR2⁎=ln (Spre/Savg-post)/TE (Wu et al., 2004b), where Spre is theaverage intensity of 100 preinjection images and Savg-post is theaverage intensity of 40 postinjection images with maximum suscept-ibility contrast for microbubbles (or 100 postinjection images atsteady-state contrast for MION). To estimate brain ΔR2⁎ in each rat, alarge region of interest (ROI) was manually drawn in the cortex basedon the high resolution FLASH images. Wilcoxon matched pairs testwas employed to compare the ΔR2⁎ values induced by A-MB andMION, and those induced by SonoVue® and MION. Assuming thatΔR2⁎ is proportional to microbubble concentration C(t) at time t, C(t)can be estimated as C(t)=k ln {Spre/S(t)}/TE+CB, where S(t) is theimage intensity at time t, k a proportionality constant, and CB aconstant residue to account for any postinjection baseline (Wong etal., 2004). Note that k depends on the microbubble properties, tissuecharacteristics, pulse sequence and field strength. Given the relativelylong injection time and the limited lifetime of microbubbles in vivo, C(t) were approximately modeled with a gamma-variate function bycurve fitting (Pickens, 1992). Full width at half maximum (FWHM)and time-to-peak were then measured from the fitted C(t) timecourses.

Results

Characterization of microbubble suspensions

The light micrographs of a representative batch of A-MB andSonoVue® microbubbles were depicted in Figs. 1a and b, respectively.The estimated size distribution was from 1 to 23 μm (with 9.21 μm

mean diameter) for A-MB and 1 to 10 μm (with 2.95 μm meandiameter) for SonoVue®, as shown in Figs. 1c and d. The estimatedmean diameter of SonoVue® was slightly higher than that provided bythe manufacturer (2.5 μm). This was likely due to the aggregation ofSonoVue® microbubbles in suspension (seen in Fig. 1b), which couldlead to overestimation of the mean diameter in the micrographanalysis.

Fig. 3. Multi-echo GE signals of A-MB phantom at ∼2% volume fraction in (a) its initial well-suspended state and (b) the final state, with monoexponential fitting as dotted line. (c)R2⁎ versus time for A-MB phantom at ∼2% volume fraction.Multi-echo GE signals of SonoVue® phantom at ∼2% volume fraction in (d) its initial well-suspended state and (e) the finalstate. (f) R2⁎ versus time for SonoVue® phantom at ∼2% volume fraction.

Fig. 4. Measured ΔR2⁎ of (a) A-MB and (b) SonoVue® microbubble suspensions versusmicrobubble volume fractions. The error bars represent standard deviation.

661J.S. Cheung et al. / NeuroImage 46 (2009) 658–664

Fig. 2a shows the dynamic susceptibility weighted GE-EPI imagesof an A-MB suspension phantom. Initially, the image intensity was lowas the microbubble suspension was uniform; it gradually increased asmicrobubbles floated upwards. Fig. 2b shows the signal time coursefor a ROI in the center of the phantom as indicated in Fig. 2a,demonstrating the strong upwards migration of microbubbles due tobuoyant force.

Measurement of ΔR2⁎ of albumin-coated and SonoVue®

microbubble suspensions

Fig. 3a shows the multi-echo GE signals of A-MB suspensionphantomat 2% volume fraction in its initialwell-suspended state. Fig. 3(b) shows themulti-echoGE signals in itsfinal state.Monoexponentialfittingwas shown as dotted lines. R2⁎ of A-MBwas plotted against timein Fig. 3c. Similarly, Figs. 3d and e show the multi-echo GE signals ofSonoVue® phantom at 2% volume fraction in its initial well-suspendedstate and its final state, respectively. Fig. 3f shows the R2⁎ versus time.ΔR2⁎ induced by microbubbles was calculated as the R2⁎ decreasebetween the initial and final microbubble-free state. The average ΔR2⁎

wasmeasured to be 107.9±23.2 s−1 and 109.6±4.1 s−1 for A-MB andSonoVue® microbubbles, respectively, at 2% volume fraction.

Fig. 4 shows the dependency of ΔR2⁎ on microbubble volumefractions in A-MB and SonoVue® microbubble suspension phantoms.An approximately linear relationship was observed. The relaxivitiesestimated from the fitted slopes were 58.52 (s volume fraction)−1 and52.94 (s volume fraction)−1 for A-MB and SonoVue®, respectively.

In vivo rat brain imaging

Microbubble susceptibility contrast enhancements were consis-tently observed in all five rats studied. During microbubble injections,the respiratory rates were observed to increase slightly (by ∼10% fromthe typical 60–65 breaths/min). No other apparent adverse effects

were observed during and after imaging sessions. Fig. 5 illustratesthe rat brain images typically observed during A-MB injection(0.2 mL of ∼4% volume fraction). Fig. 5a shows the high resolutionanatomical FLASH image. Fig. 5b shows one of the preinjection GE-EPI T2⁎-weighted images. Fig. 5c shows the postinjection GE-EPIT2⁎-weighted image with the maximum susceptibility contrast. Fig.5(d) depicts the typical T2⁎-weighted signal time courses during A-MB injection with ROIs indicated in Fig. 5b. Three ROIs wereselected from cortex (Ctx), cauduate putamen (CPu) and bloodvessel (BV) regions. The computed ΔR2⁎ map is shown in Fig. 5e.ΔR2⁎ map measured for 0.6 mg Fe/kg MION in the same animal wasdepicted in Fig. 5f, similar to that shown in Fig. 5e.

Fig. 5. Corresponding images from the same rat brain during A-MB injection (0.2 mL of ∼4% volume fraction over ∼10 s): (a) anatomical image, (b) preinjection GE-EPI T2⁎-weightedimage and (c) postinjection GE-EPI T2⁎-weighted image with the maximum susceptibility contrast. (d) T2⁎-weighted signal time courses in different brain regions during A-MBinjection. ΔR2⁎ maps for (e) A-MB and (f) intravascular contrast agent MION at 0.6 mg Fe/kg. Three ROIs used for time course measurement are shown in (b) and gamma-variatefitted data shown in sold lines.

662 J.S. Cheung et al. / NeuroImage 46 (2009) 658–664

Fig. 6 illustrates the images typically observed during SonoVue®

injection (0.2 mL of ∼3.5% volume fraction). Fig. 6a shows the highresolution FLASH image. Fig. 6b shows one of the preinjection imageswhile Fig. 6c shows the postinjection image with the maximumsusceptibility contrast. Signal time courses are depicted in Fig. 6(d).

Fig. 6. Corresponding images from the same rat brain during SonoVue® injection (0.2 mL ofweighted image and (c) postinjection GE-EPI T2⁎-weighted image with the maximum susceSonoVue® injection. ΔR2⁎ maps for (e) SonoVue® and (f) 0.6 mg Fe/kg MION.

Similar ΔR2⁎ maps were observed for SonoVue® and 0.6 mg Fe/kgMION.

The signal time courses demonstrate that the negative enhance-ments started to slowly disappear a few minutes after injection as aresult of the limited lifetime of microbubbles in vivo. Similar signal

∼3.5% volume fraction over ∼10 s): (a) anatomical image, (b) preinjection GE-EPI T2⁎-ptibility contrast. (d) T2⁎-weighted signal time courses in different brain regions during

Table 1In vivo measurements of relaxation rate enhancement (ΔR2⁎), full width at halfmaximum (FWHM) and time-to-peak of the concentration time courses for albumin-coated microbubbles (A-MB; 0.2 mL of ∼4% volume fraction), SonoVue® (0.2 mL of∼3.5% volume fraction) and monocrystalline iron oxide nanoparticles (MION; 0.6 mgFe/kg) in rat brain cortex (mean±standard deviation, N=5).

ΔR2⁎ (s−1) FWHM (s) Time-to-peak (s)

A-MB 2.49±1.00 114±39 57±20SonoVue® 2.41±1.18 86±16 48±12MION 1.98±0.36 N.A. 24±2

663J.S. Cheung et al. / NeuroImage 46 (2009) 658–664

time courses were observed in Ctx, CPu and BV (Figs. 5d and 6d forboth A-MB and SonoVue® microbubbles. Table 1 shows the in vivomeasurements of ΔR2⁎, FWHM and time-to-peak of A-MB andSonoVue® as well as ΔR2⁎ and time-to-peak of MION in cortex areaamong all rats studied. FWHM measured from the concentrationtime courses in five rats was 114±39 s and 86±16 s for A-MB andSonoVue®, respectively. Time-to-peak was 57±20 s and 48±12 sfor A-MB and SonoVue®, respectively. With identical injectionprotocol, time-to-peak was found to be shorter (24±2 s) forMION. ΔR2⁎ was 2.49±1.00 s−1, 2.41±1.18 s−1 and 1.98±0.36 s−1

for A-MB, SonoVue® and MION, respectively. No statisticallysignificant differences were found between the ΔR2⁎ values inducedby A-MB and MION (P=1.00), and between those induced bySonoVue® and MION (P=1.00). This indicates that the in vivosusceptibility effects of A-MB and SonoVue® at the dosage used arecomparable to that of 0.6 mg Fe/kg MION in brain tissue at 7 T. Inaddition, the susceptibility effects induced by two types ofmicrobubbles are comparable in vivo.

Discussions

In this study, gas-filled microbubbles were successfully demon-strated as a novel intravascular contrast agent for brain MRI for thefirst time. Two different types of microbubbles were investigated fortheir susceptibility effects by dynamic brain imaging at 7 T.Microbubbles induced ΔR2⁎ maps were observed to be similar tothose caused by MION, a standard intravascular contrast agent,reflecting the blood distribution in tissue vasculature (Wu et al.,2003, 2004c, d). ΔR2⁎ induced by A-MB (0.2 mL of ∼4% volumefraction) and SonoVue® (0.2 mL of ∼3.5% volume fraction) wasshown to be comparable to that by 0.6 mg Fe/kg MION. Theseresults indicate that microbubbles could be used as a blood pool MRcontrast agent in brain. With the increasing availability of high-fieldMRI systems in both clinical and research setting, gas-filledmicrobubbles offer the promise as a viable and unique contrastagent since their MR susceptibility effect increases with B0

2

(Dharmakumar et al., 2002).Approximately linear relationship was observed between ΔR2⁎

and microbubble volume fractions, consistent with the microbubblesusceptibility effect in a theoretical study (Dharmakumar et al.,2002). Such relationship was also observed in other microbubblephantom studies (Alexander et al., 1996; Ueguchi et al., 2006; Wonget al., 2004). Because MR susceptibility effect scales up withmicrobubble size for the same volume fraction (Alexander et al.,1996; Dharmakumar et al., 2002), one would expect SonoVue®

microbubbles to produce weaker susceptibility effect than A-MB dueto their smaller diameters. However, two types of microbubblesexhibited similar susceptibility effects both in phantoms and in vivo.This might result from the increased susceptibility effect caused bySonoVue® microbubble aggregation (as observed in the lightmicrographs). On the other hand, microbubbles larger than 10 μmcould be filtered by lung capillary bed (Conhaim and Rodenkirch,1997), which might lead to the decreased susceptibility effectobserved in vivo for A-MB.

For a 250 g rat (8% v/w blood plasma) with 0.2 mL microbubblesuspension injection, the in vivo volume fraction of microbubbles inthe plasma will be diluted by a factor of ∼100 when the steady-stateconcentration is reached. Assuming 5% blood volume fraction in brain(Kwong et al., 1995), microbubble volume fraction will be furtherreduced by a factor of ∼20. A comparison of the ΔR2⁎ of microbubblesin uniform suspension (234 s−1 for 4% A-MB and 185 s−1 for 3.5%SonoVue® as estimated from Fig. 4) and in vivo ΔR2⁎ observations inbrain (2.49 s−1 for 0.2 mL 4% A-MB and 2.41 s−1 for 0.2 mL 3.5%SonoVue®) reveals that the microbubble susceptibility effects weremuch stronger in vivo than in phantom suspension. This likely arisesfrom the fact that, for a given concentration, a MR susceptibilitycontrast agent produces much stronger effect when it is partitioned intissue microvasculature thanwhen uniformly distributed in a solution(Bjornerud et al., 2002). In addition, microbubbles, with size compar-able to that of the capillaries, may produce stronger ΔR2⁎ effects thansmaller-sized susceptibility particles for a given volume fraction(Weisskoff et al.,1994). Other factors, such as increases inmicrobubblesize due to intravascularmicrobubble growth (Wong et al., 2004), mayalso contribute to the strong susceptibility effect in vivo.

Time-to-peak for microbubbles in rat brain was found to be longerthan that for MION. This is largely expected as microbubbles, with sizecomparable to that of red blood cells, flow slower than blood plasmawhile MION nanoparticles flow together with plasma. In one of theanimals studied, the T2⁎-weighted signals after microbubble injectiondid not return to the preinjection baseline, probably caused bymicrobubble trapping in local tissue vasculature.

Microbubble susceptibility effect is relatively weak in comparisonwith other intravascular MR susceptibility contrast agents. TheSonoVue® dosage used in current study, 0.2 mL of ∼3.5% volumefraction (equivalent to 0.8 mL/kg of ∼3.5% volume fraction), washigher than that used in clinical ultrasound imaging (0.08 mL/kg of0.8% volume fraction) though it was much lower than the tolerabledosage in animals (20 mL/kg of 0.8% volume fraction) (EuropeanMedicines Agency, 2001). Microbubble-induced ΔR2⁎ depends on themicrobubbles' radius, volume fraction, overall magnetic susceptibilitydifference between the microbubble and blood plasma (Δχ), and B0.Without increasing B0, it is possible to increase the ΔR2⁎ by optimizingthe bubble size, core gas type, or modifying themagnetic properties ofmicrobubble shells as suggested by an earlier theoretical study(Dharmakumar et al., 2002). However, such development remainsto be experimentally demonstrated in the future.

With the unique cavitation and sonoporation characteristics,microbubbles offer other potentially valuable and exciting applica-tions such as MRI guidance of microbubble-based gene and drugdelivery in brain. The relatively short in vivo lifetime of microbubblesmay pose challenges in some of these therapeutic applications.Nevertheless, the microbubble fabrication technology is advancing.For example, substantially increased in vivo lifetime has beendemonstrated recently using surfactant molecules with multiphasemixing technique (Dressaire et al., 2008). Furthermore, moleculartargeting capability can be achieved by microbubble surface mod-ification (Stieger et al., 2008).

Conclusions

In this study, we investigated the feasibility of gas-filled micro-bubbles as an intravascular MR susceptibility contrast agent in brain at7 T. Considerable susceptibility induced changes were observed andcharacterized in rat brain using the custom-made albumin-coatedmicrobubbles and a commercially available clinical ultrasoundmicrobubble contrast agent. The results indicate that microbubblescan serve as a unique intravascular MR contrast agent in vivo at highfield. Such capability has the potential to lead to real-time MRIguidance in various microbubble-based drug delivery and therapeuticapplications in the central nervous system.

664 J.S. Cheung et al. / NeuroImage 46 (2009) 658–664

Acknowledgments

We thank Dr. Joseph C.K. Leung of the Department of Medicine andDr. Ke Xia Cai at the Laboratory of Biomedical Imaging and SignalProcessing of the University of Hong Kong for assistance. This workwas supported by the Hong Kong Research Grant Council (CERG HKU7642/06M).

References

Alexander, A.L., McCreery, T.T., Barrette, T.R., Gmitro, A.F., Unger, E.C.,1996. Microbubblesas novel pressure-sensitive MR contrast agents. Magn. Reson. Med. 35, 801–806.

Bekeredjian, R., Chen, S., Frenkel, P.A., Grayburn, P.A., Shohet, R.V., 2003. Ultrasound-targeted microbubble destruction can repeatedly direct highly specific plasmidexpression to the heart. Circulation 108, 1022–1026.

Belliveau, J.W., Rosen, B.R., Kantor, H.L., Rzedzian, R.R., Kennedy, D.N., McKinstry, R.C.,Vevea, J.M., Cohen, M.S., Pykett, I.L., Brady, T.J., 1990. Functional cerebral imaging bysusceptibility-contrast NMR. Magn. Reson. Med. 14, 538–546.

Bjornerud, A., Johansson, L.O., Briley-Saebo, K., Ahlstrom, H.K., 2002. Assessment of T1and T2⁎ effects in vivo and ex vivo using iron oxide nanoparticles in steady state—dependence on blood volume and water exchange. Magn. Reson. Med. 47, 461–471.

Bouakaz, A., Versluis, M., de Jong, N., 2005. High-speed optical observations of contrastagent destruction. Ultrasound Med. Biol. 31, 391–399.

Cerny, D., Mills, G.J., Westkaemper, P.J., 1990. Continuous Sonication Method forPreparation Protein Encapculated Microbubbles. United States.

Conhaim, R.L., Rodenkirch, L.A., 1997. Estimated functional diameter of alveolar septalmicrovessels at the zone I–II border. Microcirculation 4, 51–59.

Culp, W.C., Porter, T.R., Lowery, J., Xie, F., Roberson, P.K., Marky, L., 2004. Intracranial clotlysis with intravenous microbubbles and transcranial ultrasound in swine. Stroke35, 2407–2411.

Daffertshofer, M., Hennerici, M., 2003. Ultrasound in the treatment of ischaemic stroke.Lancet Neurology 2, 283–290.

De Guio, F., Benoit-Cattin, H., Davenel, A., 2008. Signal Decay Due to Susceptibility-Induced Intravoxel Dephasing on Multiple Air-Filled Cylinders: MRI Simulationsand Experiments. Magma 21, 261–271.

Dharmakumar, R., Plewes, D.B., Wright, G.A., 2002. On the parameters affecting the sensi-tivity ofMRmeasures of pressurewithmicrobubbles.Magn. Reson. Med. 47, 264–273.

Dressaire, E., Bee, R., Bell, D.C., Lips, A., Stone, H.A., 2008. Interfacial polygonalnanopatterning of stable microbubbles. Science 320, 1198–1201.

European Medicines Agency, 2001. SonoVue®. London.Hauff, P., Seemann, S., Reszka, R., Schultze-Mosgau, M., Reinhardt, M., Buzasi, T., Plath, T.,

Rosewicz, S., Schirner, M., 2005. Evaluation of gas-filled microparticles andsonoporation as gene delivery system: feasibility study in rodent tumor models.Radiology 236, 572–578.

Kaneko, Y., Maruyama, T., Takegami, K., Watanabe, T., Mitsui, H., Hanajiri, K., Nagawa, H.,Matsumoto, Y., 2005. Use of a microbubble agent to increase the effects of highintensity focused ultrasound on liver tissue. Eur. Radiol. 15, 1415–1420.

Kwong, K.K., Chesler, D.A., Weisskoff, R.M., Donahue, K.M., Davis, T.L., Ostergaard, L.,Campbell, T.A., Rosen, B.R., 1995. MR perfusion studies with T1-weighted echoplanar imaging. Magn. Reson. Med. 34, 878–887.

Liu, Y., Miyoshi, H., Nakamura, M., 2006. Encapsulated ultrasound microbubbles:therapeutic application in drug/gene delivery. J. Control Release 114, 89–99.

Mehier-Humbert, S., Bettinger, T., Yan, F., Guy, R.H., 2005. Plasma membrane porationinduced by ultrasound exposure: implication for drug delivery. J. Control Release104, 213–222.

Moseley, M.E., Wendland, M.F., Rampil, I., Barnhart, J., 1991. Microbubbles: a novel MRsusceptibility contrast agent. Proceedings of the 10th Annual Meeting of theISMRM. San Francisco, California, USA, p. 1020.

Pickens, D.R., 1992. Perfusion/diffusion quantitation with magnetic resonance imaging.Invest. Radiol. 27 (Suppl 2), S12–S17.

Rempp, K.A., Brix, G., Wenz, F., Becker, C.R., Guckel, F., Lorenz, W.J., 1994. Quantificationof regional cerebral blood flow and volume with dynamic susceptibility contrast-enhanced MR imaging. Radiology 193, 637–641.

Rosen, B.R., Belliveau, J.W., Buchbinder, B.R., McKinstry, R.C., Porkka, L.M., Kennedy, D.N.,Neuder, M.S., Fisel, C.R., Aronen, H.J., Kwong, K.K., 1991. Contrast agents and cerebralhemodynamics. Magn. Reson. Med. 19, 285–292.

Sheikov, N., McDannold, N., Vykhodtseva, N., Jolesz, F., Hynynen, K., 2004. Cellularmechanisms of the blood–brain barrier opening induced by ultrasound in presenceof microbubbles. Ultrasound Med. Biol. 30, 979–989.

Shimamura, M., Sato, N., Taniyama, Y., Yamamoto, S., Endoh, M., Kurinami, H., Aoki, M.,Ogihara, T., Kaneda, Y., Morishita, R., 2004. Development of efficient plasmid DNAtransfer into adult rat central nervous system using microbubble-enhancedultrasound. Gene Ther. 11, 1532–1539.

Simonsen, C.Z., Ostergaard, L., Vestergaard-Poulsen, P., Rohl, L., Bjornerud, A., Gyldensted,C., 1999. CBF and CBV measurements by USPIO bolus tracking: reproducibility andcomparison with Gd-based values. J. Magn. Reson. Imaging 9, 342–347.

Stieger, S.M., Dayton, P.A., Borden, M.A., Caskey, C.F., Griffey, S.M.,Wisner, E.R., Ferrara, K.W., 2008. Imaging of angiogenesis using Cadence contrast pulse sequencing andtargeted contrast agents. Contrast Media Mol. Imaging 3, 9–18.

Taniyama, Y., Tachibana, K., Hiraoka, K., Namba, T., Yamasaki, K., Hashiya, N., Aoki, M.,Ogihara, T., Yasufumi, K., Morishita, R., 2002. Local delivery of plasmid DNA into ratcarotid artery using ultrasound. Circulation 105, 1233–1239.

Ueguchi, T., Tanaka, Y., Hamada, S., Kawamoto, R., Ogata, Y., Matsumoto, M., Nakamura,H., Johkoh, T., 2006. Air microbubbles as MR susceptibility contrast agent at1.5 Tesla. Magn. Reson. Med. Sci 5, 147–150.

Unger, E.C., Hersh, E., Vannan, M., Matsunaga, T.O., McCreery, T., 2001. Local drug andgene delivery through microbubbles. Prog. Cardiovasc. Dis. 44, 45–54.

van Bruggen, N., Busch, E., Palmer, J.T., Williams, S.P., de Crespigny, A.J., 1998. High-resolution functional magnetic resonance imaging of the rat brain: mappingchanges in cerebral blood volume using iron oxide contrast media. J. Cereb. BloodFlow Metab. 18, 1178–1183.

Weisskoff, R.M., Zuo, C.S., Boxerman, J.L., Rosen, B.R.,1994.Microscopic susceptibility variationand transverse relaxation: theory and experiment. Magn. Reson. Med. 31, 601–610.

Wong, K.K., Huang, I., Kim, Y.R., Tang, H., Yang, E.S., Kwong, K.K., Wu, E.X., 2004. In vivostudy of microbubbles as an MR susceptibility contrast agent. Magn. Reson. Med.52, 445–452.

Woods, R.P., Grafton, S.T., Holmes, C.J., Cherry, S.R., Mazziotta, J.C., 1998. Automatedimage registration: I. General methods and intrasubject, intramodality validation.J. Comput. Assist. Tomogr. 22, 139–152.

Wu, E.X., Wong, K.K., Andrassy, M., Tang, H., 2003. High-resolution in vivo CBVmappingwith MRI in wild-type mice. Magn. Reson. Med. 49, 765–770.

Wu, E.X., Tang, H., Asai, T., Yan, S.D., 2004a. Regional cerebral blood volumereduction in transgenic mutant APP (V717F, K670N/M671L) mice. Neurosci. Lett.365, 223–227.

Wu, E.X., Tang, H., Jensen, J.H., 2004b. Applications of ultrasmall superparamagnetic ironoxide contrast agents in the MR study of animal models. NMR Biomed. 17, 478–483.

Wu, E.X., Tang, H., Jensen, J.H., 2004c. High-resolution MR imaging of mouse brainmicrovasculature using the relaxation rate shift index Q. NMR Biomed. 17, 507–512.

Wu, E.X., Tang, H., Wong, K.K., 2004d. Mapping cyclic change of regional myocardialblood volume using steady-state susceptibility effect of iron oxide nanoparticles. J.Magn. Reson. Imaging 19, 50–58.

Wu, J., Pepe, J., Rincon, M., 2006. Sonoporation, anti-cancer drug and antibody deliveryusing ultrasound. Ultrasonics 44 (Suppl 1), e21–e25.