An Intraoperative Brain Shift Monitor Using Shear Mode ...scholar.harvard.edu/files/sct/files/an_intra... · P. Jason White, PhD, Stephen Whalen, BS, Sai Chun Tang, PhD, Greg T. Clement,

An Intraoperative Brain Shift MonitorUsing Shear Mode TranscranialUltrasoundPreliminary Results

P. Jason White, PhD, Stephen Whalen, BS, Sai Chun Tang, PhD,Greg T. Clement, PhD, Ferenc Jolesz, MD, Alexandra J. Golby, MD

Objective. Various methods of intraoperative structural monitoring during neurosurgery are used tolocalize lesions after brain shift and to guide surgically introduced probes such as biopsy needles orstimulation electrodes. With its high temporal resolution, portability, and nonionizing mode of radia-tion, ultrasound has potential advantages over other existing imaging modalities for intraoperativemonitoring, yet ultrasound is rarely used during neurosurgery largely because of the craniotomyrequirement to achieve sufficiently useful signals. Methods. Prompted by results from recent studieson transcranial ultrasound, a prototype device that aims to use the shear mode of transcranial ultra-sound transmission for intraoperative monitoring was designed, constructed, and tested with 10human participants. Magnetic resonance images were then obtained with the device spatially regis-tered to the magnetic resonance imaging (MRI) reference coordinates. Peaks in both the ultrasoundand MRI signals were identified and analyzed for both spatial localization and signal-to-noise ratio(SNR). Results. The first results aimed toward validating the prototype device with MRI showed anexcellent correlation (n = 38; R2 = 0.9962) between the structural localization abilities of the twomodalities. In addition, the overall SNR of the ultrasound backscatter signals (n = 38; SNR = 25.4 ± 5.2dB, mean ± SD) was statistically equivalent to that of the MRI data (n = 38; SNR = 22.5 ± 4.8 dB).Conclusions. A statistically significant correlation of localized intracranial structures between intraop-erative transcranial ultrasound monitoring and MRI data was achieved with 10 human participants. Wehave shown and validated a prototype device incorporating transcranial shear mode ultrasound for clin-ical monitoring applications. Key words: brain shift; monitor; neurosurgery; transcranial; ultrasound.

AbbreviationsFOV, field of view; FSPGR, fast spoiled gradient echo;ICSF, insular cortex/sylvian fissure; ITUM, intraoperativetranscranial ultrasound monitor; MRI, magnetic reso-nance imaging; SE, spin echo; SNR, signal-to-noise ratio;TE, echo time; 3D, 3-dimensional; TR, repetition time

Intraoperative Ultrasound Monitoring forNeurosurgery

The use of preoperatively acquired images to guide neu-rosurgery has clear benefits in that it allows surgeons toplan and perform the initial steps of a surgical procedurewith information that is otherwise unobtainable fromdirect observation of the surgical field. During the courseof surgery, however, the images progressively becomemisrepresentations of the actual anatomy because ofintraoperative shifts and deformations of the brain.1–3

The incorporation of an intraoperative tool that can con-tinuously monitor brain shift and deformation couldpotentially solve this problem. Recent studies haveshown the benefits of intraoperative and perioperativeultrasound, especially as a real-time neuronavigation aidwhen used in conjunction with preoperative magnetic

© 2009 by the American Institute of Ultrasound in Medicine • J Ultrasound Med 2009; 28:191–203 • 0278-4297/09/$3.50

Technical Advance

Article includes CME test

CME

CME

Received July 29, 2008, from the Departmentsof Radiology (P.J.W., S.C.T., G.T.C., F.J.) andNeurosurgery (S.W., A.J.G.), Brigham and Women’sHospital, Harvard Medical School, Boston,Massachusetts USA. Revision requested September2, 2008. Revised manuscript accepted for publica-tion September 8, 2008.

This work was funded by National Institutes ofHealth grants P01CA067165, 1U41RR019703-01A2,R25 CA089017-06A2, and R21EB004353. DrClement is an original inventor on patent 7175599.He has no current financial interest in this intellec-tual property.

Address correspondence to P. Jason White, PhD,Department of Radiology, Brigham and Women’sHospital, Harvard Medical School, 221 LongwoodAve, EBRC 521, Boston, MA 02115 USA.

E-mail: [email protected]

jum282online.qxd:Layout 1 1/16/09 4:13 PM Page 191

resonance imaging (MRI) and computed tomog-raphy. Intraoperative ultrasound has been usedto successfully track brain shift,4–7 guide needlebiopsies,8 assist in brain mass resections,8–10

assist in the treatment of vascular malforma-tions,9,10 guide cystic evacuations,11 and guideendoscopic procedures.11 Hyperechoic struc-tures such as biopsy needles and most intra-parenchymal mass lesions can be localized inreal time and with sufficiently high spatial reso-lution. Vascular structures can also be readily andaccurately characterized with power Dopplerultrasound, which is useful not only for opera-tions involving vascular malformations but alsoto identify and track major vessels in the vicini-ty of targeted lesions.8,10 However, the cumber-some nature of existing devices, lack of anestablished registration method, and cranioto-my requirement for an acoustic window (whenthe craniotomy site is used for ultrasound guid-ance, the surgeon has to stop surgery, positionthe probe, image, and resume the procedureafter the probe has been removed from the sur-gical field) have limited the applicability ofintraoperative ultrasound as a standard tool.8–11

A new device, the intraoperative transcranialultrasound monitor (ITUM), operating tran-scranially as a real-time monitor during neuro-surgical procedures, was designed, constructed,and tested with a series of 10 healthy humanparticipants. The ITUM was developed to havethe capability to noninvasively assess, at a hightemporal resolution and a sufficient spatial res-olution, intracranial deformations of majorbrain structures. As a low-profile device mount-ed outside the surgical field, the ITUM wouldonly require a 1-time spatial registration andwould scan continuously without interferingwith surgical activities. In addition, becausethe device scans transcranially, a craniotomyacoustic window would no longer be requisite.

In this preliminary study to validate the ITUMwith a reference standard, the prototype devicewas first characterized with benchtop mea-surements and evaluated for safety. Then 10healthy adult participants were scanned byMRI in conjunction with the spatially regis-tered ITUM, and the results were analyzed tocorrelate anatomic structures as defined byboth modalities.

Transcranial UltrasoundSince the first attempts to use ultrasound formedical imaging, scientists have sought a way toimage brain tissue with sound waves.12 However,the transmission of ultrasound transcranially toproduce backscatter signals with sufficientlyuseful information has been a severe technicalobstacle for the field of neurosonography.Because the skull bone attenuates ultrasoundroughly in proportion to the frequency of the sig-nal, there exists a trade-off between the signal-to-noise ratio (SNR) and the resolution of traditionalB-mode transcranial ultrasound imaging. Inaddition, distortions caused by irregularities inthe skull’s shape, density, and sound speed col-lectively contribute to destroying the integrity ofan ultrasound beam.

However, it has been observed that under cer-tain conditions it is possible to propagate ultra-sound through the skull with reduced distortionand higher signal amplitudes by using highangles of incidence.13 Recent studies on shearmode transcranial ultrasound have suggestedthat ultrasound beams can be harnessed toimage through cranial bone by the intentionalinduction of a shear mode ultrasound wave with-in the skull bone.13–15 The method prevents skullsurface backscatter from interfering with echoesoriginating from intracranial regions of interestand makes diagnostic information easier to ana-lyze than traditional approaches to neurosonog-raphy. The method also minimizes refraction ofthe ultrasound wave as it passes through bone,making it possible to effectively focus the beamin tissues beyond the bone. In contrast, attemptsto use traditional ultrasound are plagued by diffi-culties resulting from refraction of the ultrasoundbeam.16

Both numeric and experimental investigationsindicate that ultrasound transmission enhance-ment at high angles of incidence is due to thebehavior of shear modes induced in the skullbone. When the angle of incidence is beyond theSnell critical angle for the longitudinal pressurewave, ultrasound propagation in the bonebecomes purely a result of a shear wave. Contraryto the common belief that transcranial shearmodes are of negligible application because of dis-tortion17 and attenuation,18 it has been shown thatthe conversion from a longitudinal wave (in the

192 J Ultrasound Med 2009; 28:191–203

Intraoperative Brain Shift Monitor Using Transcranial Ultrasound


epicranial tissue) to a shear wave (within the skullbone) and back to a longitudinal wave (in thebrain parenchyma) does not necessarily producea highly distorted or small-amplitude wave.13

In addition to using shear mode transcranialpropagation, the ITUM incorporates 3 otherapplication-specific design aspects: (1) a focusedultrasound transducer with a pressure field pro-file that selectively monitors specific intracranialregions; (2) a lowered effective ultrasound fre-quency (fc = 0.96 MHz) to overcome attenuationstemming from the skull bone; and (3) a lowultrasound duty cycle for time-extended appli-cations. This study shows application of the tran-scranial shear mode for ultrasound imaging inhuman participants.

Materials and Methods

Transducer Design and CharacterizationThe ITUM is composed of a lead zirconatetitanate piezoelectric transduction elementmounted in a polyvinyl chloride housing (Figure1). The source element is spherically focused witha 30-mm radius of curvature and a 25.4-mm-diameter circular aperture and is electricallypoled to operate in the thickness mode (f0 = 0.96MHz). The 30-mm radius of curvature was select-ed to specifically focus on the insular cor-tex/sylvian fissure (ICSF), which typically liesbetween 20 and 50 mm from the temporal acous-tic window. A backing layer consisting of tungstenpowder embedded within a polymer matrix isapplied to the nontransmitting face of the crys-tal to increase the radiation bandwidth. A poly-mer standoff layer at the transmitting aperture isdesigned to ensure shear angles of incidencebetween the ultrasound propagation vector andthe contact surface. The entire assembly is MRIcompatible, and 2 cylindrical MRI fiduciarymarkers are mounted on the transducer casing toprovide spatial registration with MRI data.

The spatial distribution and spectral content ofthe ITUM’s sonication field were characterizedby benchtop measurements. The ITUM wassubmerged in degassed and deionized waterwithin a rubber-padded tank at room temper-ature. As it was actuated with a broadbandpulser-receiver,19 the radiated field was rasterscanned by a pressure-sensitive polyvinyli-

dene difluoride hydrophone (aperture diameter,0.2 mm; Precision Acoustics Ltd, Dorchester,England) mounted to a computer-controlledpositioning system (Velmex, Inc, Bloomfield,NY). Propagated waveforms were recorded(10,000 points sampled over 100 microsecondswith 4× averaging) at a spatial resolution of 0.5mm throughout a 20 × 60-mm area coplanarwith the propagation axis of the ITUM, begin-ning 20 mm from the transducer aperture. Thegeometry of the standoff material prohibitedscanning of the field within 20 mm of the trans-ducer aperture.

To determine the spectral properties of theITUM’s performance, Fourier analysis was per-formed with MATLAB (The MathWorks, Natick,MA) on the recorded waveform at the focus ofthe radiated field. The subsequent reconstruc-tion of the spatial distribution of the ultrasoundintensity was based on the results of the spectralanalysis.

Transcranial Shear Mode Backscatter SignalAnalysisPrevious studies have established the enhance-ment of ultrasound propagation through theskull at high angles of incidence using a 1-waythrough-transmission method.13 For this study,measurements were performed to determinewhether similar results could be achieved byusing the ITUM under a pulse-receive backscat-ter analysis scenario.

J Ultrasound Med 2009; 28:191–203 193

White et al

Figure 1. Schematic of the prototype ITUM positioned againstthe skull.


The ITUM was again submerged in degassedand deionized water within a rubber-paddedtank at room temperature (Figure 2). The actuat-ing pulser-receiver in this and all subsequentexperiments was a customized system designedspecifically for transcranial ultrasound applica-tions.19 The system delivered controlled 3-cyclehigh-voltage pulses at a frequency of 1 MHz witha pulse repetition frequency of 6 Hz and thendetected and amplified the backscattered echoesreceived at the transducer face. A spherical steelbackscatter target (diameter = 3.1 mm) wasmounted via a thin rigid steel needle onto a com-puter-controlled positioning system (Velmex,Inc) and positioned within the propagation fieldof the ITUM such that the needle mount was par-allel to the propagation axis (Figure 2).

Then an ex vivo human skull fixed in 10%buffered formalin (formalin-fixed human skullshave been shown to have the same ultrasoundcharacteristics as nonfixed skull bone)20 wasmounted to a rotary positioning system andpositioned between the ITUM and the steeltarget such that ultrasound was propagatedthrough the thickness dimension of the lefttemporal bone (≈30 mm superior to the exter-nal auditory meatus). The right half of the skullwas excised for this experiment. The skull waspositioned such that the angle of incidencecould be varied from 0° to 35°; the rotation was

such that increasing the angle of incidencedirected the ultrasound beam path moretoward the posterior of the cranial vault (Figure2). A 50-mm distance between the ITUM andthe skull was geometrically necessary to achievethese angles of incidence. For each angle ofincidence between 0° and 35° in step sizes of 5°,a 20 × 20-mm scan (scan resolution = 0.5 mm)of the backscatter field was performed by rasterscanning the steel target and recording thebackscattered signal as received by the ITUM,pulser-receiver, and oscilloscope system(Tektronix, Inc, Beaverton, OR) at each gridlocation (Figure 2). The recorded waveformswere then analyzed with MATLAB to yieldbackscatter field reconstructions.

Validation With Human Participants and MRITen healthy adult participants were recruited formeasurements to validate ITUM results againstMRI data. The Partner’s Institutional ReviewBoard approved the protocol, and all partici-pants gave written informed consent. Safetystudies (benchtop and animal) had been per-formed to assess thermal effects, the mechani-cal index, and current leakage associated withthe ITUM. The system delivered controlled 3-cycle high-voltage pulses at a frequency of 1MHz with a pulse repetition frequency of 6 Hzand then detected and amplified backscatteredechoes received at the transducer face. Theseexcitation parameters (3-microsecond pulsetrains at 6 Hz gives a duty cycle of <0.002%) weredesigned to yield a sufficiently low time- averaged power for extended application in theoperating room. Even under the worst-casescenario of perfect energy absorption in thetissue, the applied time-averaged power ofapproximately 1.4 mW over 1 second wouldraise a 1-g tissue mass (specific heat ≈ 3500J/kg/°C) by 0.0004°C. The peak rarefactionalpressure for the ITUM was measured to be onthe order of 0.1 MPa in water.

The ITUM was mounted onto an MRI-compat-ible adjustable arm, which was then attached toan MRI head coil. Each participant was posi-tioned in a head-first supine position on thescanner bed, and the ITUM was placed in con-tact against the right temple (≈50 mm posteriorto the lateral canthus of the eye), aligned to pro-

194 J Ultrasound Med 2009; 28:191–203


Figure 2. Schematic of the experimental setup for investigating the effects of theangle of incidence on transcranial backscatter detection. In this view, the ultra-sound incident angle (θi) was increased by rotating the skull in a clockwise fashion,with the posterior aspect of the skull (P) moving toward the right and the anterioraspect of the skull (A) moving toward the left. The image is not drawn to scale.CH1 indicates channel 1; and CLK, clock.


ject in a posteromedial trajectory within thecranial vault (Figure 3). Ultrasound gel (ParkerLaboratories, Inc, Fairfield, NJ) was applied onboth the scalp and the ITUM contact surface toensure adequate ultrasound coupling.

The ITUM was then actuated with the parame-ters as specified above, and a time trace of thereceived backscattered signal was recorded(10,000 points sampled over 200 microsecondswith 8× averaging). The same sequence wasacquired without the participant (ITUM radiatinginto air) to be used as a baseline time trace. Theparticipants were then placed in the scanner (3-TMRI scanner; GE Healthcare, Milwaukee, WI),and a series of MRI scans were performed withthe ITUM in place (but not actuated). First, 3-dimensional (3D) fast spoiled gradient echo(FSPGR) scanning (repetition time [TR] = 7.5 mil-liseconds; echo time [TE] = 3 milliseconds; num-ber of excitations = 1; field of view [FOV] = 25.6 ×25.6 cm; slice thickness = 1 mm) was performed,taking care to include the ITUM fiduciary mark-ers within the FOV. The images from this scanwere used to plan the second scan, which was aT1-weighted spin echo (SE) coronosagittaloblique scan (TR = 650 milliseconds; TE = 10 mil-liseconds; number of excitations = 0.75; variableFOV; slice thickness = 4 mm) that was aligned to becoplanar with the propagation axis of the ITUM, asidentified by the fiduciary markers (Figure 3).

Signal Processing and Data AnalysisStatistical analysis was performed on the twosets of data (ITUM and MRI) to establish a cor-relation between the observed locations ofintracranial anatomic structures. First, the pathof ultrasound propagation was identified in theMRI oblique scan, and the pixel values along thispath were extracted for analysis. To make ameaningful comparison between the MRI pixelvalue and the ultrasound backscatter vector, theabsolute value of the first derivative with respectto the path was performed on the MRI data, ie,

(1)

where PV is the pixel value, and x is the path ofultrasound propagation. The rationale was that achange in structure composition (an anatomicboundary) is related to a change in the charac-teristic acoustic impedance, which is the sourceof a backscattered signal. A local peak in thisvalue should then correspond to a local peak inbackscattered ultrasound data.

Next, the recorded ITUM backscatter signal wassubtracted from the recorded baseline time trace,and the time segment of interest was extractedfor comparison with the processed MRI data.Sound speeds of 1562 m/s for the brainparenchyma21 and 1500 m/s for the skull bone

J Ultrasound Med 2009; 28:191–203 195

White et al

Figure 3. Diagram showing the registration of MR images with the ITUM in a healthy human participant. An axial section (left, 3DFSPGR; TR = 7.5 milliseconds; TE = 3.0 milliseconds; FOV = 25.6 × 25.6 cm; slice thickness = 1 mm) containing the fiduciary markers(FM1 and FM2) of the ITUM was used to prescribe the oblique section (center, T1-weighted SE; TR = 650 milliseconds; TE = 10 mil-liseconds; FOV = 24 × 24 cm; slide thickness = 3 mm) that was used for signal analysis. A second axial slice shows the location of theICSF (right, 3D FSPGR; TR = 7.5 milliseconds; TE = 3 milliseconds; FOV = 25.6 × 25.6 cm; slice thickness = 1 mm). The solid lines indi-cate the corresponding slices for both images (in the center diagram, line A corresponds to the image on the left, and line B corre-sponds to the image on the right); dotted line, the ultrasound propagation direction; and arrowheads, ICSF.


shear mode14 were used to accurately convert thetime sequences into propagation distances. TheITUM signal was processed according to

(2)

where I(t) is the ITUM backscatter data; IB(t) isthe ITUM baseline data; t is the time vector cor-responding to the recorded data; � represents across-correlation performed in Fourier space; His the Hilbert transform; and USNA represents thenonattenuated processed ultrasound signal.Cross-correlation of a gaussian-enveloped sinu-soid with the backscattered signal was per-formed in accordance with a method previouslydeveloped for improving ultrasound resolutionand signal strength through highly attenuatingmedia.22 A propagation distance–dependentattenuation factor was also applied to the result-ing backscatter A-line signal to compensate forsignal loss due to attenuation within the brain.This was done according to

(3)

where α is the attenuation coefficient (0.87dB/cm/MHz)21; f is the ultrasound frequency(0.96 MHz); d is the propagation distance; andUSNA is the nonattenuated result from Equation 2(US represents the processed ultrasound databoth here and in the figures).

Signal peaks in both the MRI and ultrasounddata were identified and analyzed for correlationin the localization of intracranial structures. Toexclude data that did not originate from intracra-nial structures, signal peaks between 0 and 20mm were not included in the analysis. Duringdata acquisition, a slight oscillation of approxi-mately 1 Hz in signal arrival time could beobserved for each participant but only for signalsoriginating from intracranial structures. Thisoscillatory motion was most likely due to car-diac-induced pulsatile motion and therefore wasalso used qualitatively to confirm that the signalwas of intracranial origin.

To compare the observed distance from thedevice from ITUM backscatter signal peaks andMRI pixel intensity gradients, linear regression

analysis was used (ie, calculation of the slope, y-intercept, and correlation coefficient). A pairedStudent t test confirmed zero bias between com-parisons. The Bland-Altman technique23 was alsoused to evaluate the agreement between the mea-surements, with the limits of agreement definedas the mean difference ± 1.96 SDs. Specific to thedesign of this study was the observation andlocalization of the ICSF (Figure 3), this being aprominent hyperechoic structure because of itsextensive cerebrospinal fluid–cortex structuralboundaries. The ICSF was clearly identifiable inall MRI scans by visual inspection. A signal peakfrom each of the ITUM and the MRI series wasidentified as originating from the ICSF, and a sep-arate statistical analysis was performed on this setof measurements.

Finally, SNR calculations and statistical analy-ses (participant mean, participant SD, overallmean, and overall SD) were performed indepen-dently on the series of ITUM and MRI data. TheSNR was defined as

(4)

where AS is the signal amplitude, and AN is themean noise amplitude. The mean noise ampli-tude, AN, was obtained by selecting a series oftemporospatial points beyond the signal andapplying the formula

(5)

where n is the number of points sampled; An isthe value of the point n as obtained by Equations1–3; and A

_is the mean values of all n points. The

overall statistical analyses were also indepen-dently performed on the SNR as measured fromsignals arising from the ICSF.

Results

Transducer Design and CharacterizationThe radiated pressure field of the prototypeITUM was spatially scanned. Fourier analysis ofthe recorded waveform at the focus of the radiat-ed pressure field revealed a peak at 0.96 MHzwith a –3-dB bandwidth of 260 kHz (Figure 4).

Σ

196 J Ultrasound Med 2009; 28:191–203



The intensity field reconstruction was then per-formed at this frequency, yielding a spatial full-width half-maximum (FWHM) intensity profileof 4.8 mm (orthogonal to the propagation vec-tor) at a distance of 30 mm from the aperture(Figure 4). A slight standing wave effect undula-tion in the intensity along the propagation direc-tion was observed in the focal area of the field.This pattern was most likely due to wave interac-tion with the ultrasound signal reflected from thehydrophone mount.

Transcranial Shear Mode Backscatter SignalAnalysisThe results from the transcranial backscatterfield scans corroborated earlier studies13,14 onthe enhancement of ultrasound transmissionthrough the skull bone at high angles of inci-dence. In our study, the reduced distortion of theshear mode was made evident by the detectionof a backscattered signal from a pointlike target.It was observed that the backscattered signalfrom a steel sphere as transmitted and receivedby the ITUM decreased in amplitude as an inter-vening ex vivo skull was rotated to increase theultrasound beam’s angle of incidence (Figure 5).This reduction in backscattered amplitude con-tinued as the angle of incidence was increase toapproximately 30°, at which point the signal wasobserved to increase in amplitude with anincreasing angle of incidence. The backscatteredamplitude continued to increase until the geom-etry of the experimental setup would not allowfurther rotation (θi = 35°). At this angle of inci-dence, the peak intensity was 84% of the originalpeak intensity at normal incidence (Figure 6). At its lowest (θi = 20°), the backscattered peakintensity was 4% of the original peak intensity atnormal incidence.

It was also observed that beyond a critical angleof incidence (≈30°), the backscattered signalfrom the target became less distorted and moreclearly isolated from spurious signals. A doublebackscattered signal as received by the ITUMfrom the target was evident when the skull wasaligned at normal incidence (Figure 5). As theangle of incidence was increased beyond 10°, thesignal that arrived at an earlier time (≈165microseconds) disappeared, leaving only a sin-gle signal (arriving at ≈177 microseconds), which

J Ultrasound Med 2009; 28:191–203 197

White et al

Figure 4. Experimental results from the benchtop characterization of the proto-type ITUM. A, Reconstructed intensity map of the ITUM’s radiation pattern. B,Same intensity radiation profile as depicted with 10% contour lines. C, Spectralimpulse response of the ITUM (the circle marks the peak at 0.96 MHz). D, Crosssection of the spatial intensity profile at 30 mm from the transducer aperture.

A

B

C

D


was then enhanced as the angle of incidence wasincreased beyond 30°. A qualitative assessmentof the scanned backscattered intensity fields(Figure 6) revealed an incident angle–dependentincrease in distortion of the spatial distributionof backscattered intensity as qi was increasedfrom 0° to 25°. The scanned fields for qi = 30° andθi = 35° were markedly less distorted than thoseat θi = 20° and θi = 25° and somewhat less dis-torted than the angles of incidence between 0°and 15°.

Validation With Human Participants and MRIThe ICSF was identifiable in both the MRI andITUM data for each of the 10 healthy adult par-ticipants (Figure 7). In addition to the ICSF, 28other structures (eg, outer cortex, superiortemporal sulcus, and lateral ventricle) wereidentified as correlative for both series of data,from as many as 7 intracranial structures for 1 participant to only 1 structure (the ICSF) for 1participant (Figure 8). Most of these structureswere in the vicinity of the ICSF (between 20 and60 mm from the scalp), but 5 backscattered sig-nals originating from structures beyond 80 mmwere identified on the MR image as the marginsof the ipsilateral lateral ventricles. Statisticalanalysis was performed for both the complete

set of signals arising from all intracranial struc-tures (n = 38) and those signals arising from theICSF (n = 10).

Linear regression analysis confirmed a highlevel of correlation between MRI localization andlocalization of the same structures with theITUM (slope = 1.02; y-intercept = –0.06; R2 =0.9962 for all structures; slope = 1.05; y-intercept= –0.2; R2 = 0.9922 for the ICSF). Confirmation ofa significant correlation between the two seriesof measurements was also achieved with theBland-Altman technique of analysis, using themean of the two measured values as the inde-pendent variable and the difference between thetwo measured values as the dependent variable.For the case of all identified structures, only 2measurements fell beyond ±1.96 SDs; for thecase of the ICSF measurements, no data pointsfell outside ±1.96 SDs.

The SNR for each signal peak was also mea-sured for both the MRI data and the ITUM sig-nals in 2 categories (Figure 9): all identifiablestructures (n = 38) and ICSF only (n = 10). Overall,the SNR levels of the ITUM signals were statisti-cally equivalent to those of the MRI signals. Themean SNRs for the ITUM backscattered signalswere 25.4 ± 5.2 dB for all structures and 28.5 ± 8.2dB for the ICSF. For the MRI signal peaks, themean SNRs were 22.5 ± 4.8 dB for all structuresand 25.3 ± 5.6 dB for the ICSF. As would beexpected, the mean SNR values for signals arisingfrom the ICSF were higher than the correspond-ing mean SNR values when all other structureswere considered. The lowest and highest SNRvalues obtained by MRI were 14.8 and 36 dB,respectively. The lowest and highest SNR valuesobtained by the ITUM were 14.4 and 38.6 dB.

Discussion

The clinical impetus for intraoperative naviga-tional aid stems from the need to visualizeanatomic structural changes and surgical toolplacements that are not readily observabledirectly from the surgical field. These includelocalization of lesions after brain shift,24 identifi-cation of acute iatrogenic conditions such asintracranial hemorrhage,25 and guidance of sur-gically introduced probes such as biopsy nee-dles26 and stimulation electrodes.27

198 J Ultrasound Med 2009; 28:191–203


Figure 5. Time-windowed backscatter signals from a 3.1-mm-diameter steelsphere as received by the ITUM after transcranial signal propagation. Results froma sequence of 8 ultrasound angles of incidence, ranging from normal (0°) to 35°,are shown.


Although MRI is effective in the spatial localiza-tion of such targets, it lacks efficacy as a real-timemonitor during surgery because real-time scan-ning cannot take place during surgical activities(eg, dural opening, lesion resection, and biopsy).The typical protocol for MRI-guided surgeryrequires cessation of surgical procedures andclearing of metallic tools from the surgical field tominimize image artifacts. The time added to theoverall duration of a procedure and to the timeduring which a patient is under general anesthe-sia increases the chance of surgical complica-tions. Also, standard MRI pulse sequences haveinsufficient time resolutions to monitor acutestatus changes during a procedure. Emergentsurgical complications that require immediateattention (eg, acute hemorrhage, aneurysm rup-ture, and excessive brain shift) occur on a timescale of seconds, whereas MRI can give feedbackonly at a temporal resolution of minutes. Becauseof these inherent MRI deficiencies, surgeons arerequired to weigh the benefit of scanning timewith the diagnostic potential before ordering anintraoperative scan. In addition, intraoperativeMRI is very resource intensive, and its availabilityis limited to a few dedicated centers. These short-comings of MRI-guided neurosurgery promptedthe design and investigation of the ITUM, whichhas the potential to be deployed in any operatingroom for intraoperative monitoring.

The discovery that a focused ultrasound beam,traveling as a shear wave through the skull, mayemerge less distorted than a longitudinal one13,14

became a key design aspect for incorporationinto the ITUM. In contrast with shear wave imag-ing in soft tissue, the ITUM was designed toincorporate a traditional compression wavebackscatter imaging modality with an inter-posed shear conversion within the osseous tis-sue of the skull bone. By directing a longitudinalwave onto the skull surface at angles beyond 30°,a generated shear wave travels through the thick-ness dimension of the skull bone in a directionnearly identical to the initial direction of propa-gation. When the shear wave reaches the interiorsurface of the skull, the energy is converted backinto a longitudinal wave. Backscatter signalsfrom intracranial structures undergo the samemode conversion as they travel along the samepath back to the transducer.

This backscatter detection mode of operationwas shown to be effective by benchtop experi-ments (see “Results,” “Transcranial Shear ModeBackscatter Signal Analysis”). At near-normalangles of ultrasound incidence, a doublebackscattered signal originating from the sametarget was observed, most likely due to the com-bined effects of modal bifurcation at the skullinterfaces. With a longitudinal mode phasespeed of approximately 2800 m/s14,20 and ashear mode phase speed of approximately 1500m/s,14 actuation of both modes at the skull sur-face (generation of only a longitudinal compo-nent with no shear component within an exvivo skull is not possible in an experimental set-ting) resulted in 2 detectable signals with differ-ent arrival times. The double backscattersignals obscured and distorted the definition ofthe single pointlike target. The signal stemming

J Ultrasound Med 2009; 28:191–203 199

White et al

Figure 6. Spatial distributions of backscatter intensities from a 3.1-mm-diametersteel sphere as received by the ITUM after transcranial signal propagation. A,Each plot represents the backscatter intensity measured on a 20 × 20-mm gridlying orthogonal to the propagation axis of the ITUM (10% contour lines, scan-ning resolution of 0.5 mm). The plots range from a normal incident angle (0°; topleft) to an incident angle of 35° (bottom right). B, Peak measured backscatteredintensity (normalized) for each of the fields in A.

A

B


from the longitudinal mode actuation decreasedas the ultrasound incident angle at the skullsurface was increased. After the critical angle

for longitudinal transmission was reached, asdefined by

(6)

where c1 is the sound speed of the first medium,and c2 is the longitudinal sound speed of the sec-ond medium (in the case of an ex vivo humanskull in water, c1 ≈ 1500 m/s, and c2 ≈ 2800 m/s;so θi ≈ 32°), it was observed that only 1 backscat-ter signal remained. This result gave strongsupport to the incorporation of shear modetranscranial ultrasound into the design of theITUM. Together with the lowered frequency ofoperation, the transcranial shear mode of oper-ation was a substantial contributor to the highSNR and the precision of localization exhibitedby the ITUM.

The high level of correlation between MRI andITUM localization of intracranial structures (R2 =0.9962) was shown predominantly by backscat-tered signals within the spatial range of 2 to 6 cmfrom the ITUM aperture. Because this studyfocused on signals originating from the ICSF, theobservation of 5 signals emanating from themargins of the lateral ventricles (beyond 8 cmfrom the aperture of the ITUM) was incidental.The inclusion of these data points in the overallanalysis may have reduced the correlationresults (this is most notable in the Bland-Altmanplot) or the SNR values, but a substantially highcorrelation was maintained nonetheless, and nosignificant difference was evident in the SNRanalysis. Future studies are being designed thatwill focus on localization and motion tracking ofthe lateral ventricles and other deeper-set struc-tures, including the midline falx, basal ganglia,and brain stem structures. It is anticipated thatsubarachnoid hemorrhage could also have aconsiderable impact on backscatter distortionand the SNR and, as such, may be investigatedfor use as a monitoring method for intraopera-tive subarachnoid hemorrhage.

200 J Ultrasound Med 2009; 28:191–203


Figure 7. A, Detail of an oblique MRI scan (T1-weighted SE; TR = 650 milliseconds;TE = 10 milliseconds; FOV = 12 × 12 cm; slice thickness = 3 mm) that has been reg-istered with the ITUM’s signal path (line) through the ICSF (arrow). B and C, Absolutevalues of the MRI pixel intensity gradient corresponding to the ITUM signal path (B)and the processed ITUM backscatter signal (C). Signal peaks corresponding to theICSF can be identified in both B and C at approximately 3.1 cm (arrows).

B

C

A


J Ultrasound Med 2009; 28:191–203 201

White et al

Figure 8. Results of the statistical analysis to validate the ability of the ITUM to localize intracranial structures against the referencestandard of MRI. A, Location of intracranial structures as measured by the ITUM plotted against the same measurement performedby MRI. B, Same as A but with analysis performed on only those signal peaks originating from the ICSF. In A and B, the solid line indi-cates linear regression; and dotted line, perfect correlation. C, Results of a Bland-Altman analysis comparing the mean of the twoseries of measurements (MRI and ultrasound) with the difference between the same measurements. D, Same as C but with Bland-Altman analysis performed on only those signal peaks originating from the ICSF. In C and D, the solid line indicates perfect correla-tion; and dotted line, ±1.96 SDs. In A–D, black squares indicate signal peaks originating from the ICSF; and white squares, signalpeaks originating from all other intracranial structures.

A

C

B

D


References

1. Dorward NL, Alberti O, Velani B, et al. Postimaging braindistortion: magnitude, correlates, and impact on neuron-avigation. J Neurosurg 1998; 88:656–662.

2. Hill DL, Maurer CR Jr, Maciunas RJ, Barwise JA, FitzpatrickJM, Wang MY. Measurement of intraoperative brain sur-face deformation under a craniotomy. Neurosurgery 1998;43:514–528.

3. Roberts DW, Hartov A, Kennedy FE, Miga MI, Paulsen KD.Intraoperative brain shift and deformation: a quantitativeanalysis of cortical displacement in 28 cases. Neurosurgery1998; 43:749–760.

4. Comeau RM, Sadikot AF, Fenster A, Peters TM.Intraoperative ultrasound for guidance and tissue shift cor-rection in image-guided neurosurgery. Med Phys 2000;27:787–800.

5. Coenen VA, Krings T, Weidemann J, et al. Sequential visu-alization of brain and fiber tract deformation duringintracranial surgery with three-dimensional ultrasound: anapproach to evaluate the effect of brain shift. Neurosurgery2005; 56(suppl):133–141.

6. Lindner D, Trantakis C, Renner C, et al. Application of intra-operative 3D ultrasound during navigated tumor resection.Minim Invasive Neurosurg 2006; 49:197–202.

202 J Ultrasound Med 2009; 28:191–203


Figure 9. Signal-to-noise ratios for the identified signal peaks in both the MRI (A and C) and ITUM (B and D) data. A and B, Signal-to-noise ratios for all identified signal peaks. C and D, Signal-to-noise ratios for signal peaks identified as originating from the ICSF.In A–D, the solid line indicates the mean value; dotted line, SDs; squares, MRI data points; triangles, ultrasound data points; blacksquares and triangles, signal peaks originating from the ICSF, and white squares and triangles, signal peaks originating from all otherintracranial structures.

A

C

B

D


7. Letteboer MMJ, Willems PWA, Viergever MA, Niessen WJ.Brain shift estimation in image-guided neurosurgery using3-D ultrasound. IEEE Trans Biomed Eng 2005; 52:268–276.

8. Sosna J, Barth MM, Kruskal JB, Kane RA. Intraoperativesonography for neurosurgery. J Ultrasound Med 2005;24:1671–1682.

9. Tirakotai W, Miller D, Heinze S, Benes L, Bertalanffy H, SureU. A novel platform for image-guided ultrasound.Neurosurgery 2006; 58:710–718.

10. Unsgaard G, Rygh OM, Selbekk T, et al. Intra-operative 3Dultrasound in neurosurgery. Acta Neurochir (Wien) 2006;148:235–253.

11. Strowitzki M, Kiefer M, Steudel WI. A new method ofultrasonic guidance of neuroendoscopic procedures: tech-nical note. J Neurosurg 2002; 96:628–632.

12. Dussik KT. On the possibility of using ultrasound waves asa diagnostic aid. Neurol Psychiatry 1942; 174:153–168.

13. Clement GT, White PJ, Hynynen K. Enhanced ultrasoundtransmission through the human skull using shear modeconversion. J Acoust Soc Am 2004; 115:1356–1364.

14. White PJ, Clement GT, Hynynen K. Longitudinal and shearmode ultrasound propagation in human skull bone.Ultrasound Med Biol 2006; 32:1085–1096.

15. Pichardo S, Hynynen K. Treatment of near-skull brain tissuewith a focused device using shear-mode conversion: anumerical study. Phys Med Biol 2007; 52:7313–7332.

16. White DN, Clark JM, Chesebrough JN, White MN,Campbell JK. Effect of skull in degrading the display ofechoencephalographic B and C scans. J Acoust Soc Am1968; 44:1339–1345.

17. Clement GT, Sun J, Hynynen K. The role of internal reflec-tion in transskull phase distortion. Ultrasonics 2001; 39:109–113.

18. Hayner M, Hynynen K. Numerical analysis of ultrasonictransmission and absorption of oblique plane wavesthrough the human skull. J Acoust Soc Am 2001; 110:3319–3330.

19. Tang SC, Clement GT, Hynynen K. A computer-controlledultrasound pulser-receiver system for transkull fluid detec-tion using a shear wave transmission technique. IEEE TransUltrason Ferroelectr Freq Control 2007; 54:1772–1783.

20. Fry FJ, Barger JE. Acoustical properties of the human skull.J Acoust Soc Am 1978; 63:1576–1590.

21. Kremkau FW, Barnes RW, McGraw CP. Ultrasonic attenua-tion and propagation speed in normal human brain. J Acoust Soc Am 1981; 70:29–38.

22. Clement GT. Spectral image reconstruction for transcranialultrasound measurement. Phys Med Biol 2005; 50:5557–5572.

23. Bland JM, Altman DG. Statistical methods for assessingagreement between two methods of clinical measure-ment. Lancet 1986; 1:307–310.

24. Nabavi A, Black PM, Gering DT, et al. Serial intraoperativemagnetic resonance imaging of brain shift. Neurosurgery2001; 48:787–798.

25. Morandi X, Haegelen C, Henaux PL, Riffaud L. Brain shift iscentral to the pathogenesis of intracerebral haemorrhageremote from the site of the initial neurosurgical procedure.Med Hypotheses 2006; 67:856–859.

26. Rubin JM, Dohrmann GJ. Intraoperative neurosurgicalultrasound in the localization and characterization ofintracranial masses. Radiology 1983; 148:519–524.

27. Moringlane JR, Fuss G, Becker G. Peroperative transcranialsonography for electrode placement into the targeted sub-thalamic nucleus of patients with Parkinson disease: tech-nical note. Surg Neurol 2005; 63:66–69.

J Ultrasound Med 2009; 28:191–203 203

White et al


Documents

An Intraoperative Brain Shift Monitor Using Shear Mode ...scholar.harvard.edu/files/sct/files/an_intra... · P. Jason White, PhD, Stephen Whalen, BS, Sai Chun Tang, PhD, Greg T. Clement,