Ultrasound in Med. & Biol., Vol. 32, No. 9, pp. 1307–1313, 2006Copyright © 2006 World Federation for Ultrasound in Medicine & Biology

Printed in the USA. All rights reserved0301-5629/06/$–see front matter

doi:10.1016/j.ultrasmedbio.2006.05.017

● Original Contribution

INTRAOPERATIVE BRAIN ULTRASOUND: A NEW APPROACH TOSTUDY FLOW DYNAMICS IN INTRACRANIAL ANEURYSMS

THILO HÖLSCHER,* JAVIER RODRIGUEZ-RODRIGUEZ,‡ WILKO G. WILKENING,§

JUAN C. LASHERAS,‡ and HOI SANG U†

*Department of Radiology, †Neurosurgery, and ‡Mechanical and Aerospace Engineering, University of CaliforniaSan Diego, San Diego, CA, USA; and §Ruhr Center of Excellence for Medical Engineering (KMR), Bochum,

Germany

(Received 18 October 2005, revised 5 May 2006, in final form 19 May 2006)

Abstract—The aim was to evaluate the potential of contrast-enhanced ultrasound to visualize the hemodynamicsin intracranial aneurysms during neurosurgical intervention and to quantify the ultrasound data using digitalparticle image velocimetry (DPIV) technique. Aneurysms were scanned through the intact dura mater, preclip-ping and again postclipping after closure of the dura. After intravenous injection of Optison™, angio-like viewsof the vascular tree surrounding the aneurysm, including the aneurysm sac, were obtained. Single ultrasoundcontrast agent microbubbles could be visualized in the aneurysm sac and the flow dynamics could be assessed invivo. Spatial and temporal distributions of the velocity in the aneurysm and in the parent vessels were measuredwith DPIV using the backscattered signals from the microbubbles. Subsequently, the fluid stresses, vorticity,circulation, etc., were calculated from the velocity fields. We demonstrate in this paper that intraoperativecontrast-enhanced ultrasound can be used to quantify the flow dynamics within an aneurysm. (E-mail:thoelscher@ucsd.edu) © 2006 World Federation for Ultrasound in Medicine & Biology.

Key Words: Intraoperative, Ultrasound contrast, Aneurysm, Phase inversion, Digital particle image velocimetry.

INTRODUCTION

The overall goal of flow dynamic studies in aneurysms isto predict the risk of aneurysm rupture. It is known thatfluid and wall stresses, as well as pressure fluctuations,affect aneurysm growth and the risk of rupture (Burlesonand Turitto 1996). To date, those flow dynamic param-eters cannot be extracted from in vivo imaging data. Tosimulate the in vivo situation, aneurysm casts are built tostudy the flow dynamics in a phantom flow facility(Gobin et al. 1994; Kerber et al. 1999; Strother 1995).Commonly, silicone casts of postmortem specimens orreconstructions of three-dimensional (3-D) computed to-mography (CT) or magnetic resonance (MR) data(Chung et al. 2004; Fahrig et al. 1999) are used toconduct flow dynamic studies. Optional approaches tocreate anatomically realistic human aneurysm models arederived from in vivo rotational angiography data (Stein-man et al. 2003). The techniques used to build these casts

Address correspondence to: Thilo Hoelscher, MD, Assistant Ad-junct Professor, University of California San Diego, Medical Center,

Department of Radiology, 212 West Dickinson St. B-412, San Diego,CA 92103-8756. E-mail: thoelscher@ucsd.edu

1307

are well described (Chong et al. 1994; Fahrig et al. 1999;Imbesi et al. 2003; Kato et al. 2001; Kerber and Heilman1983; Kerber et al. 1999; Salsac et al. 2004). However,using available clinical imaging data to create appropri-ate casts to study aneurysm hemodynamics is challeng-ing. Currently, the main limitations for making casts oraneurysm reproductions are the lack of informationabout wall thickness and elasticity, patient-related bloodpressure waveforms and flow pulsatility. Referring toSteinman et al. (2003), the latter can only be captured byimposing a representative periodic flow waveform at themodel inlet. Therefore, the flow dynamics rendered rep-resent only a “snapshot” of the hemodynamic stressesthat occur in a particular aneurysm. Moreover, as wasnoted by Imbesi and Kerber (2001), “the creation ofreplicas of human vascular abnormalities, particularlyaneurysms, is a tedious and technically difficult under-taking.”

Ultrasound is the only diagnostic tool that enablesreal-time assessment of hemodynamics, and its intraop-erative application during neurosurgical interventionshas been shown in several studies (Gronningsaeter et al.

1996; Woydt et al. 2001). The combination of advanced

1308 Ultrasound in Medicine and Biology Volume 32, Number 9, 2006

transducer technology, innovative imaging techniquesand the rapidly increasing knowledge about the acousticproperties of ultrasound contrast agents (UCA) encour-age the use during brain surgery (Kanno et al. 2005).

This paper describes first experiences of intraoper-ative contrast-enhanced ultrasound to visualize real-timehemodynamics of intracranial aneurysms and analysis ofthese data off-line, using a digital particle image velo-cimetry technique. The authors note that this report isstrictly method-oriented and the potential impact of con-trast-enhanced brain ultrasound on neurosurgical casemanagement in the operating room will be discussed in aseparate paper.

MATERIALS AND METHODS

Study population and procedureA total of six patients scheduled for neurosurgical

intervention of middle cerebral artery (MCA) aneurysmclipping were studied. Patients were enrolled after sign-ing a written informed consent according to the localInstitutional Review Board approval and preoperativeinformation about the risk-benefit ratio of the study wasobtained.

The first ultrasound study was performed intraoper-atively after skull trepanation through the intact duramater. The aim was to visualize the feeding artery of theaneurysm and the downstream vascular tree. The flowdynamics in the aneurysm sac were also assessed afterUCA injection and images were stored as DICOM movieclips on the machine. A second ultrasound study wasperformed after aneurysm clipping and closure of thedura mater to monitor the success of the surgical proce-dure and to exclude vessel stenosis or occlusion due toclip misplacement. Off-line, these results were comparedwith postclipping, intraoperative cerebral x-ray angiog-raphy.

Ultrasound imaging techniqueA Siemens Sonoline Antares™ ultrasound system

was used for the intraoperative studies. The ultrasoundmachine was equipped with an intraoperative linear arraytransducer (VF13-5SP) with a 25.9 � 2.5-mm footprint.The transducer has 128 elements and a maximum trans-mit frequency of 11.43 MHz in B-mode and 6.15 MHz intissue harmonic imaging (THI) mode. THI or widebandphase inversion imaging was used for all studies. Ac-cording to the official user manual for the Antares™system, the maximum surface temperature, assessed intissue-mimicking material at maximum output power(100%), is �41° (IEC 60601 to 2-37). At 100% outputpower, the peak-negative acoustic pressure is 1.97 MPaand the signal intensity (Ispta) is 266 mW/cm2. For the

contrast studies, the output power was decreased to 4 to

5% (mechanical index, MI: �0.4) and the maximumtransmit frequency was increased to (6.15 MHz). Theframe rate was in a range of 35 to 40 Hz.

Ultrasound contrast agent, administration and dosingFS 069 (Optison™) contains octafluoropropane gas

with a shell of human albumin (5%). The diameters ofthe microbubbles range between 2.0 and 4.5 �m. Opti-son™ is approved by the Food and Drug Administrationfor clinical use of endocardial border delineation. Pre-clinical studies of the open brain have shown that Opti-son™ does not alter the blood–brain barrier integrity incombination with ultrasound exposure levels used forclinical purposes (Hynynen et al. 2003). Optison™ wasadministered intravenously at rates of 0.05–0.1 mL perinjection. A maximum of four injections total (0.2 mL to0.4 mL) per patient was allowed, including pre- andpostclipping studies.

Digital particle image velocimetry (DPIV)DPIV measures particle displacements (Dx/Dy)

from statistical correlations of single pairs of images andenables computation of velocity fields. The fundamentalprinciple of DPIV is to measure the displacements (Dx/Dy) of fluid tracers from two consecutive images whileknowing the time delay between them. The displace-ments must be small enough so that Dx/Dt is a goodapproximation of the x-velocity. That means that thetrajectory must be nearly straight and the speed along thetrajectory should be nearly constant. Therefore, Dt, thetime between two consecutive images is chosen to besmall compared with the relevant flow scales of theexperiment. The DPIV technique is described in detail byCanton et al. (2005).

Ultrasound data analysis with DPIVThe ultrasound signals from the microbubbles are

used as fluid tracers. Individual frames are extractedfrom the ultrasound data and processed in pairs usingInsight™ (TSI Incorporated, St. Paul, MN, USA) soft-ware. This software identifies the fluid tracers and com-putes velocity fields from each pair of frames. The ve-locity fields are computed as means of all the tracerscontained in areas called “interrogation windows.” Thisreduces the noise that tracking the displacement of singletracers would imply. Therefore, the spatial resolutionwith which the velocity field is measured is limited bythe size of such windows. It has to be remarked that thisresolution is unrelated to that of the ultrasound image,which has to be higher in any case (at least to a level thatallows the detection of single tracers). Thus, it is desir-able to set the interrogation window size as small aspossible to maximize the spatial resolution. However, the

minimum size is limited by the requirement that it must

arench

Intraoperative brain ultrasound ● T. HÖLSCHER et al. 1309

contain a sufficient number of tracers. The velocity fieldmeasurements shown in this paper have been obtainedusing an interrogation window of 64 � 64 pixels andthen refined with subwindows of 16 � 16 pixels. Thisprovides a spatial resolution in the velocity field of 1.5mm � 1.5 mm (the diameter of the aneurysm shown inthe figures is about 15 mm). Finally, the velocity andother relevant fields (strain rate, vorticity, etc.) are dis-played with a MATLAB™ software tool (MathworksInc., Natick, MA, USA).

RESULTS

Online contrast effect pre-/postclippingContrast-enhanced angio-like views of the entire

vascular tree surrounding the aneurysm could be ob-tained. The flow velocities in the afferent and the down-stream vascular segments, as well as inside the aneurysmcould be recorded and assessed (Figs. 1 and 2). Afterclipping of the aneurysms, the same technique was usedto monitor successful positioning of the aneurysm clips.It provided a rapid and unequivocal means to determinepatency of the entire parent and effluent arterial complex.The results were in accordance with postclipping, intra-operative cerebral x-ray angiography. Acoustic signalsfrom single microbubbles could be visualized andtracked within each aneurysm, enabling the assessment

Fig. 1. Precontrast, the morphology of the aneurysm comater, posterior to two gyri. Maximum inner diameter oleft of the aneurysm as a hyperechogenic band (arrow) d

with p

of real-time flow dynamics (Fig. 3).

Off-line DPIV analysisSpatial and temporal distribution of the velocity in

the aneurysm as well as in the parent vessel could bemeasured, applying the DPIV algorithm to the backscat-tered signals from UCA microbubbles. Subsequently, thefluid stresses, vorticity, circulation, etc. were calculatedfrom the velocity measurements. Characteristic flow ve-locity fields within an aneurysm sac during an early anda medial systolic phase of a cardiac cycle are shown inFig. 4. In this particular case, measurements show thatthe flow inside the aneurysmal sac is dominated by thepresence of a large 3-D vortex. Although the circulation(integral of the vorticity flux over the cross-section) ismaximum at peak systole, the rotational motion persiststhroughout the whole cardiac cycle, indicating that theflow shear exerted on the dome never vanishes.

DISCUSSION

The risk of intracranial aneurysm rupture is mainlydefined by vascular wall fragility and flow dynamicchanges throughout the cardiac cycle. The latter has beenstudied extensively during recent years. Due to the lackof real-time data, in vitro pulsatile flow facilities havebeen established to mimic the in vivo situation of patho-logic flow patterns in the aneurysm sac. Steinman andcolleagues stated that high resolution 3-D medical imag-

visualized in an oblique view through the intact duraysm sac: 1.1 cm. The feeding artery can be seen on thetronger harmonic backscatter of erythrocytes comparedyma.

uld bef aneurue to s

ing and advanced image processing techniques reliably

e 3-D

1310 Ultrasound in Medicine and Biology Volume 32, Number 9, 2006

simulate pulsatile blood flow in anatomically realisticarterial geometries derived from in vivo imaging (Stein-man 2002; Steinman et al. 2003). To evaluate the dy-namic changes in the aneurysm sac, various strategies are

Fig. 2. (a) The upper part shows the aneurysm sac in thearrows), with the Doppler cursor placed in it. The lowerfeeding artery, with a peak systolic flow velocity of abouhowever, the Doppler gate has not been corrected. (b)atypical Doppler spectrum, representing a pathologic,

direction cannot be assessed properly because of th

available and can be differentiated into more descriptive

(Kerber et al. 1999) or computerized approaches (Salsacet al. 2004).

The current knowledge about flow dynamics inaneurysms is based exclusively on in vitro studies.

wer corner (dotted arrows) and the feeding vessel (solidf the image shows the regular Doppler spectrum of the/s. Note: the flow direction is toward the aneurysm sac;

er measurement within the aneurysm sac, showing anctional flow. Note: inside the aneurysm sac, the flowflow pattern and the reverse flow during diastole.

left lopart o

t 50 cmDopplbidire

Dynamic real-time information of pathologic flow pat-

g the p

Intraoperative brain ultrasound ● T. HÖLSCHER et al. 1311

terns within the aneurysm sac can be acquired usingcine-angiograms during the contrast injection. In prin-ciple, the dynamics of the in vivo contrast agent fillingcould be used to validate the computed in vitro slip-stream dynamics, generated in a correspondingflow model (Steinman et al. 2003). However, to date,

Fig. 3. (a) Postcontrast injection, acoustic signals from sthe aneurysm sac in the early systolic phase. (b) The same

flow of the microbubbles durin

cinegraphic image series can only be analyzed de-

scriptively by visually monitoring the contrast fillingeffect.

Ultrasound is known to be the only diagnostic toolable to analyze flow in vivo in real-time. Flow velocity,direction, pulsatility, resistance, etc. are flow dynamicparameters that can be assessed with almost any com-

icrobubbles could be visualized as white spots enteringplane as in (a), taken a couple of frames later. Turbulentostsystolic phase can be seen.

ingle mimage

mercially available ultrasound system nowadays. How-

dark b

1312 Ultrasound in Medicine and Biology Volume 32, Number 9, 2006

ever, complex flow dynamic characteristics such as flowvorticity, flow velocity and pressure fields, strain rate orshear stress, cannot be assessed. The introduction ofUCAs and the rapidly increasing knowledge of the spe-cific acoustic properties of UCA microbubbles led to thedevelopment of innovative techniques to image the in-tracranial macro-, as well as the micro-, vasculature(Eyding et al. 2005; Hölscher et al. 2005a; Hölscher et al.2005b). Using a contrast-specific imaging method, weare now able to visualize acoustic signals from singlemicrobubbles intraoperatively and to track them duringthe cardiac cycle within the aneurysm sac, the parentvessel and the downstream vascular tree. Moreover, thecontrast-specific imaging method enables the analysis ofthese data using advanced techniques to analyze flowdynamics such as DPIV, which, to date, have been usedexclusively for in vitro flow dynamic studies. The pos-sibility to study in vivo flow dynamic data are preferable

Fig. 4. (a) Off-line analysis of the dynamic ultrasound ddirection of flow, as well as the flow velocity during eathe length of the vector (length � velocity). (b) Paramepoints. Dark red: highest velocity (maximum 2.6 cm/s);

over any simulation because the individual flow charac-

teristics, especially the flow pulsatility, are an integralpart of each data assessment. The fundamental principleof DPIV analysis is to analyze the displacement of fluidtracers, which can be tracked throughout at least twoconsecutive images. The combination of harmonic im-aging, low UCA concentration (acoustic signals of singlemicrobubbles can be visualized) and low acoustic powerto avoid microbubble destruction enables using micro-bubbles as fluid tracers. Speckle tracking is not an optionin this case, because speckles are “noise phenomena”with a very short correlation length. The assessment ofspeckle displacement from two consecutive imageswould require a very high frame rate, which cannot beachieved under these particular insonation conditions.

A major limitation at this point is the 2-D dataacquisition. The flow dynamics could be visualized inreal-time in all cases by pivoting the transducer throughthe region-of-interest. For data acquisition in cine-loop, a

with DPIV. Single flow vectors (in white) represent theove) and medial (below) systolic phase, determined byimation of the flow velocity field at corresponding timelue: lowest velocity (minimum 0.0 cm/s).

atasetsrly (abtric an

dedicated scanning plane was chosen. The off-line anal-

Intraoperative brain ultrasound ● T. HÖLSCHER et al. 1313

ysis with DPIV was hampered in those cases where themicrobubbles were out of the scanning plane betweentwo pairs of images and could not be tracked appropri-ately. Therefore, the scanning plane has to be chosen tobe as parallel to the trajectories of the tracers as possible,to avoid them disappearing from the image plane be-tween two consecutive frames, thus reducing the accu-racy of the DPIV technique. Dynamic 3-D acquisition ofthe intraoperative ultrasound data would overcome thislimitation. However, an appropriate 3-D imaging tech-nique is not yet available for the system used for theintraoperative studies. A second limitation is that highspatial and temporal resolution imaging depends on hightransmit frequencies. Using a wideband phase inversiontechnique, a transmit frequency of 6.15 MHz was chosenin all cases, enabling the visualization of single micro-bubbles. Lower transmit frequencies, which are used fortranscranial imaging, for example, are currently not ableto achieve a comparable resolution (Hölscher et al.2005). Hence, acoustic signals from single microbubblescannot be visualized reliably, which currently excludesthe option to analyze contrast-enhanced transcranial ul-trasound data with DPIV. The possibility to analyze theultrasound radiofrequency data may provide a new op-portunity to overcome this limitation.

In summary, the present in vivo technique allows,for the first time, real-time measurement of the specificflow features associated with intracranial aneurysms andopens a valuable new avenue to study the flow featuresthat may be responsible for the enlargement rate, as wellas the eventual rupture, of the aneurysm. Moreover, theinformation obtained with the above-described techniquemay be used to validate experimental in vitro studies thatexist in the literature (Steinman et al. 2003; Stuhne andSteinman 2004).

REFERENCES

Burleson AC, Turitto VT. Identification of quantifiable hemodynamicfactors in the assessment of cerebral aneurysm behavior. On behalfof the Subcommittee on Biorheology of the Scientific and Stan-dardization Committee of the ISTH. Thromb Haemost 1996;76:118–123.

Canton G, Levy DI, Lasheras JC. Hemodynamic changes due to stentplacement in bifurcating intracranial aneurysms. J Neurosurg 2005;103:146–155.

Chong BW, Kerber CW, Buxton RB, Frank LR, Hesselink JR. Bloodflow dynamics in the vertebrobasilar system: Correlation of atransparent elastic model and MR angiography. AJNR Am J Neu-roradiol 1994;15:733–745.

Chung TS, Lee YJ, Kang WS, et al. Clinical and experimental inves-

tigation of pseudoaneurysm in the anterior communicating artery

area on 3-dimensional time-of-flight cerebral magnetic resonanceangiography. J Comput Assist Tomogr 2004;28:414–421.

Eyding J, Krogias C, Schollhammer M, et al. Contrast-enhanced ultra-sonic parametric perfusion imaging detects dysfunctional tissue atrisk in acute MCA stroke. J Cereb Blood Flow Metab 2006;26:576–582.

Fahrig R, Nikolov H, Fox AJ, Holdsworth DW. A three-dimensionalcerebrovascular flow phantom. Med Phys 1999;26:1589–1599.

Gobin YP, Counord JL, Flaud P, Duffaux J. In vitro study of haemo-dynamics in a giant saccular aneurysm model: Influence of flowdynamics in the parent vessel and effects of coil embolisation.Neuroradiology 1994;36:530–536.

Gronningsaeter A, Unsgard G, Ommedal S, Angelsen BA. Ultrasound-guided neurosurgery: A feasibility study in the 3-30 MHz fre-quency range. Br J Neurosurg 1996;10:161–168.

Hölscher T, Olson S, Lyden P, et al. Real time flow dynamics of a giantaneurysm. Neurology 2005;65:E8.

Hölscher T, Wilkening W, Draganski B, et al. Transcranial ultrasoundbrain perfusion assessment with a contrast agent-specific imagingmode. Results of a 2-center trial stroke. Stroke 2005a;36:2283–2285.

Hölscher T, Wilkening WG, Lyden PD, Mattrey RF. Transcranialultrasound angiography (tUSA): A new approach for contrast spe-cific imaging of intracranial arteries. Ultrasound Med Biol 2005b;31:1001–1006.

Hynynen K, McDannold N, Martin H, Jolesz FA, Vykhodtseva N. Thethreshold for brain damage in rabbits induced by bursts of ultra-sound in the presence of an ultrasound contrast agent (Optison).Ultrasound Med Biol 2003;29:473–481.

Imbesi SG, Kerber CW. Analysis of slipstream flow in a wide-neckedbasilar artery aneurysm: Evaluation of potential treatment regi-mens. AJNR Am J Neuroradiol 2001;22:721–724.

Imbesi SG, Knox K, Kerber CW. Aneurysm flow dynamics: Alter-ations of slipstream flow for neuroendovascular treatment withliquid embolic agents. AJNR Am J Neuroradiol 2003;24:2044–2049.

Kanno H, Ozawa Y, Sakata K, et al. Intraoperative power Dopplerultrasonography with a contrast-enhancing agent for intracranialtumors. J Neurosurg 2005;102:295–301.

Kato K, Ishiguchi T, Maruyama K, Naganawa S, Ishigaki T. Accuracyof plastic replica of aortic aneurysm using 3D-CT data for trans-luminal stent-grafting: experimental and clinical evaluation. J Com-put Assist Tomogr 2001;25:300–304.

Kerber CW, Heilman CB. Flow in experimental berry aneurysms:Method and model. AJNR Am J Neuroradiol 1983;4:374–377.

Kerber CW, Imbesi SG, Knox K. Flow dynamics in a lethal anteriorcommunicating artery aneurysm. AJNR Am J Neuroradiol 1999;20:2000–2003.

Salsac AV, Sparks SR, Lasheras JC. Hemodynamic changes occurringduring the progressive enlargement of abdominal aortic aneurysms.Ann Vasc Surg 2004;18:14–21.

Steinman DA. Image-based computational fluid dynamics modeling inrealistic arterial geometries. Ann Biomed Eng 2002;30:483–497.

Steinman DA, Milner JS, Norley CJ, Lownie SP, Holdsworth DW.Image-based computational simulation of flow dynamics in a giantintracranial aneurysm. AJNR Am J Neuroradiol 2003;24:559–566.

Strother CM. In vitro study of haemodynamics in a giant saccularaneurysm model: Influence of flow dynamics in the parent vesseland effects of coil embolisation. Neuroradiology 1995;37:159–161.

Stuhne GR, Steinman DA. Finite-element modeling of the hemody-namics of stented aneurysms. J Biomech Eng 2004;126:382–387.

Woydt M, Vince GH, Krauss J, Krone A, Soerensen N, Roosen K. Newultrasound techniques and their application in neurosurgical intra-

operative sonography. Neurol Res 2001;23:697–705.