MOLECULAR IMAGING

Development and validation of an intrinsic landmark-basedgating protocol applicable for functional and molecularultrasound imaging

Christoph Grouls & Max Hatting & Isabelle Tardy &

Jessica Bzyl & Georg Mühlenbruch &

Florian F. Behrendt & Tobias Penzkofer &

Christian Trautwein & Christiane Kuhl &Fabian Kiessling & Moritz Palmowski

Received: 18 November 2011 /Revised: 20 January 2012 /Accepted: 16 February 2012 /Published online: 30 March 2012# European Society of Radiology 2012

AbstractObjectives To implement a retrospective intrinsic landmark-based (ILB) gating protocol for contrast-enhanced ultra-sound (CEUS) and to compare its efficiency to non-gated,manually gated and extrinsically gated CEUS.Methods CEUS of the liver was performed in healthy mice(n05) and in NEMO knockout mice with dysplastic livers(n05). In healthy animals, first-pass kinetics of non-specific

microbubbles was recorded. Knockout mice were analysedregarding retention of VEGFR2-specific microbubbles. Forretrospective gating, a landmark which showed respiratorymovement was encircled as a region of interest (ROI). Duringinspiration, the signal intensity within the ROI altered, whichserved as gating signal. To evaluate the accuracy, non-gated,extrinsically gated and ILB-gated time-intensity curves werecreated. For each curve, descriptive parameters were calculatedand compared to the gold standard (manual frame-by-framegating).Results No significant differences in the variation of ILB- andextrinsically gated time-intensity curves from the gold standardwere observed. Non-gated data showed significantly highervariations. Also the variation of molecular ultrasound data wassignificantly lower for ILB-gated compared to non-gated data.Conclusion ILB gating is a robust and easy method to im-prove data accuracy in functional and molecular ultrasoundliver imaging. This technique can presumably be translated tocontrast-enhanced ultrasound examinations in humans.Key Points• Quantitative analysis of the uptake of contrast agentsduring ultrasound is complex.

• Intrinsic landmark-based gating (ILB) offers a simpleimplementable method for motion correction.

• Results using ILB-gating are comparable to extrinsicgating using external biomonitoring devices.

• Functional and molecular imaging of mobile organswill benefit from ILB gating.

Keywords Gating . Ultrasound . Contrast enhanced .

Functional imaging . Molecular imaging

C. Grouls : J. Bzyl : F. Kiessling :M. Palmowski (*)Department of Experimental Molecular Imaging,RWTH-Aachen University,Aachen, Germanye-mail: mpalmowski@ukaachen.de

C. Grouls : T. Penzkofer : C. Kuhl :M. PalmowskiDepartment of Diagnostic and Interventional Radiology,RWTH-Aachen University,Aachen, Germany

M. Hatting : C. TrautweinMedical Clinic II, University Hospital, RWTH-Aachen University,Aachen, Germany

I. TardyBracco Suisse SA,Geneva, Switzerland

G. MühlenbruchDepartment of Diagnostic and Interventional Neuroradiology,RWTH-Aachen University,Aachen, Germany

F. F. Behrendt :M. PalmowskiDepartment of Nuclear Medicine, RWTH-Aachen University,Aachen, Germany

Eur Radiol (2012) 22:1789–1796DOI 10.1007/s00330-012-2429-y

AbbreviationsROI Region of interestCEUS Contrast-enhanced ultrasoundILB-gating Intrinsic landmark-based gatingAUC Area under the curveVEGFR2 Vascular endothelial growth factor

receptor type 2

Introduction

Ultrasound plays a major role in preclinical cancer research,e.g. for the investigation of tumour biology [6, 8] or fortesting of drug efficiency in small animals [22]. In thiscontext, functional parameters of tumour vascularity or ex-pression of molecular markers (e.g. angiogenesis-relatedreceptors) can be assessed by contrast-enhanced ultrasound(CEUS) using non-specific or targeted contrast agents [11,15, 16]. Transgenic mouse models of spontaneously devel-oping tumours are of particular interest for preclinical re-search since they better display the biological reality ofprimary malignancies than tumours implanted under theskin [7, 9]. However, compared with subcutaneous tumours,many organs and especially the liver are subject to respira-tory motion. As a result, contrast agent kinetics are difficultto assess quantitatively.

One way to avoid artefacts from respiration is extrinsicgating by using external biomonitoring devices. However,these devices are not routinely available and due to theirhigh costs not affordable by all institutions. Besides ex-trinsic biomonitoring, intrinsic gating methods have suc-cessfully been applied in human imaging. One method ismanual frame-by-frame gating, based on the respiratorymovement of an intrinsic landmark (e.g. diaphragm), asdescribed by Averkiou et al. [1]. A more operator-independent but complex approach has been introducedby Renault and colleagues, who perform an independentcomponent analysis during examinations of the liver [20].In the search for a gating method that is easy to apply forsmall-animal imaging and/or patient studies, we decidedto perform a novel intrinsic landmark-based (ILB) gatingtechnique. ILB is based on an ultrasound landmark whichis subject to respiratory motion and detectable on theultrasound images throughout the entire cine loop.Software-based registration of the respiratory motion ena-bles the corresponding frames to be rejected from the cineloop automatically, generating a “post-processed” cineloop data without motion artefacts. So far, our methodhas not been described and has not been compared tonon-gated or extrinsically gated data.

Thus, the aim of our study was to develop and validate aneasily applicable and robust retrospective gating method forCEUS clips recorded with a clinical ultrasound system,

applicable for small-animal imaging and principally trans-latable for patient studies. Validation was done by calculat-ing the area under the curve (AUC), peak enhancement(functional ultrasound) and amount of target-bound specificmicrobubbles (molecular ultrasound) with different gatingmethods.

Materials and methods

Animal models

All animal experiments were approved by the governmen-tal review committee of animal care. Healthy nude mice(n05; CD-1, purchased from Charles River LaboratoriesInternational) were used to implement the gating protocolsand to assess first-pass kinetics with blood pool contrastagent (SonoVue, Bracco Suisse, Geneva, Switzerland).Hepatocyte-specific NEMO knockout mice (NEMOΔhepa)have been described previously [2, 3]. NEMO/IKKγ isthe regulatory subunit of the IKK complex, and deletionof NEMO in target cells blocks NF-kB activation aftere.g. TNF stimulation. NEMOΔhepa mice develop chronichepatitis after birth and show disease progression to fibro-sis and hepatocellular carcinomas (HCC). Therefore thissequence of events closely resembles the liver diseaseprogression as also seen in humans. NEMOΔhepa animalswere used to investigate the gating effect on a widely usedprotocol for molecular ultrasound imaging (using VEGFR2-specific microbubbles, BR55, Bracco Suisse).

Ultrasound imaging experiments

In all animals a two-dimensional contrast-enhanced ultra-sound examination of the liver was performed using aclinical ultrasound system (Acuson S2000, Siemens Health-care, Germany) equipped with contrast-specific pulse se-quencing (CADENCE). A multi-D matrix transduceroperating at a frequency of 4 MHz (9 L4 Transducer, Sie-mens Healthcare, Germany) was fixed on a railing systemabove the animal. The acoustic focus was centred at thelevel of the dorsal liver capsule. The imaging plane wasaligned in an axial view in the centre of the liver parenchy-ma containing the inferior vena cava and the abdominalaorta. Animals were kept anaesthetised with 2 % isofluranein room air (2 L/min). A tail vein catheter was placed forintravenous (i.v.) injection of contrast agents.

Functional imaging

Healthy mice (n05) were placed on a small pad connectedto a spirometer (Spirometer and PowerLab 4/30 fromADInstruments). Respiration was monitored and recorded.

1790 Eur Radiol (2012) 22:1789–1796

Contrast agent (SonoVue, Bracco Suisse, Geneva, Switzer-land) at a dosing of 5×106 microbubbles (MBs) suspendedin 50 μL saline solution was injected via tail vein followedby a flush of 50 μL saline solution. Starting with injection,cine loops were acquired with a frame rate of 13/s overa 16-s period.

Molecular imaging

For the molecular imaging protocol a different setting wasused: NEMO knockout mice (n05) with dysplastic liversreceived a bolus injection of target-specific ultrasound con-trast agents at a dosing of 5×106 MBs. The specific micro-bubbles are experimental contrast agents (BR55) developedby Bracco Suisse S.A. [19]. They were designed by conju-gating a VEGFR2-specific peptide to the amino group ofDSPE-PEG-2000-NH2, which was then incorporated intothe monolayer of lipids forming the shell of microbubbles[18]. The contrast enhancement was intermittently recordedfor 5 min with a frame rate of 13/s followed by a burst pulse.After the burst, the replenishment was recorded for 15 s.During postprocessing, mean signal intensity (SI) after thedestructive pulse was subtracted from the mean SI beforethe destructive pulse to determine the relative amount ofstationary microbubbles as described [21]. In these experi-ments, extrinsic gating was not performed due to technicalreasons (spirometer not available).

Postprocessing—software

During postprocessing, two software tools were used. First,we applied analysing software that creates time-intensitycurves from a region of interest (ROI)and exports the data(signal-intensity per frame) to a spreadsheet application (e.g.Microsoft Excel). We used custom-made software; however,in principle any ultrasound analysing software that is able tocreate time-intensity curves is suitable for this purpose.

For the gating process itself we used a standard spread-sheet application (Microsoft Excel®).

Retrospective intrinsic landmark-based (ILB) gating

Our retrospective intrinsic gating method was based ongating landmarks in ultrasound clips. Criteria for an areaserving as gating landmark were as follows: (1) an areawith movement due to respiration (either in the contrast-enhanced images or the B-mode images) and (2) norelevant enhancement after contrast media injection.The latter point is important because any enhancementmight influence the gating signal. Structures such as theskin or the liver capsule served as landmarks. For everyultrasound clip, one suitable gating ROI (gROI) aroundor in such a landmark was selected.

The mean signal intensity (SI) within the gROI wascalculated for each frame. Due to respiratory movement,SI of the gROI either dropped or increased (when a hyper-echoic or hypoechoic area moved into the gROI the signalincreased or decreased respectively, resulting in a down- orup-peak from the baseline SI). The baseline gating signalrepresents the respiratory rest during expiration whereas thepeaks represent respiratory movement during the inspira-tion. Consequently, frames providing a deviation of a user-defined baseline (defined by a manually chosen static gatingthreshold) could be ejected from the ultrasound data, result-ing in a gated dataset. We did this by using a digit filter inthe spreadsheet application. As control for the filter settingswe drew a graph from the gROI signal with and withoutusing the filter to illustrate the effect on the data and, incases of an unsuitable resulting graph (where peaks due torespiratory movement were still visible), to allow us to finda better gROI. The time the reader needed to perform thegating was measured starting with loading the stored ultra-sound clip into the analysing software. Time was stoppedwhen the gated time intensity curve of the analysed ROI wasplotted by the spreadsheet application.

Extrinsic gating using a biomonitoring device

To validate our method, we have chosen a commerciallyavailable extrinsic gating device, which is a standard ap-proach for rodent imaging. In healthy animals (n05) exam-ined with non-specific contrast agent (BR1), we recordedthe respiratory movement with a spirometer, as describedabove. The contrast intensity data were analysed offline inrandom order by an independent reader with 4 years ofexperience in molecular ultrasound imaging. Frames duringinspiration were excluded. This was done by importing thedata into the spreadsheet application and then manuallymarking the frames for inspiration and expiration. Theframes with breathing motion were excluded from the dataas was done for ILB gating.

Manual frame-by-frame gating

Manual frame-by-frame gating was performed as referencemethod. The ultrasound clips were examined frame-by-frame during post processing, and all frames showingmovement in any area of the images, except for move-ment due to blood flow or bowel movement, were notedand excluded from the data. We considered these cineclips, which had been manually frame-by-frame gated, asthe gold standard for respiratory gating due to the fact thatthis method is highly reliable when properly performed (inother studies, the method is used as a reference as well[12]). The time the reader needed to perform the gatingwas measured.

Eur Radiol (2012) 22:1789–1796 1791

Comparison of non-gated and gated time-intensity curves

For functional imaging using unspecific microbubbles(n05 animals), an ROI with a standard diameter(0.4 cm) was drawn in a comparable position in theliver tissue in each clip. Peak enhancement and areaunder the curve (AUC) were calculated as often appliedstandard parameters in CEUS [13, 17]. Differences inILB-gated, extrinsically gated and non-gated valuesfrom the gold standard (manual frame-to-frame gating)were calculated and expressed as percent variation from thegold standard.

Molecular ultrasound imaging with VEGFR2-specificmicrobubbles was performed as previously described [10].A standardised ROI was chosen as described for functionalimaging. Differences in ILB-gated and non-gated valuesfrom the gold standard were calculated and expressed aspercent variation from the gold standard.

Statistics

Statistical analysis was performed using GraphPad Prism(V5.01, GraphPad Software). Data are presented as means

with standard deviations. Differences between the peak en-hancement and AUC of the gold standard and those of theILB gated, extrinsically gated and non-gated methods werecalculated and compared using the Mann-Whitney test (un-paired, two-tailed). P values of <0.05 were considered assignificant.

Results

All animals could be examined successfully. No animal diedduring the examination.

Implementation of ILB gating

In all examinations a landmark that provides a movementpattern which correlates to respiration could be determined,and no ultrasound clip had to be excluded (illustrated inFig. 1). In Fig. 2, signal intensity of a gating ROI (gROI)and the corresponding non-gated and ILB-gated data areshown. gROIs ranged in maximum diameter from 1 to8 mm. Most frequently, these areas were located in or nextto the liver capsule or the skin.

Fig. 1 a–c Ultrasoundexamination. a Transversesection through a mouse liver(B-mode, 9 MHz). Vena cava(*), liver (**) and skin (***)are indicated. b and cCorresponding section indual-mode at 4 MHz prior tocontrast injection. b Section inthe contrast-specific mode. Arepresentative landmark usedfor intrinsic gating is encircledin red (white arrow). A repre-sentative ROI is placed in theliver (blue circle)

1792 Eur Radiol (2012) 22:1789–1796

Influence of gating on time-intensity curves: areaunder the curve

Figure 3 summarises the influence of respiratory gating ontime-intensity curves. The mean difference between ILB andframe-by-frame gated AUCs was 0.36±0.38 %. The meandifference of extrinsically gated examinations and frame-by-frame gated ones was 1.06±0.85 %. Results did not differsignificantly (P>0.05). However, the mean difference be-tween non-gated and frame-by-frame gated examinationswas 8.66±7.51 %. Values differed significantly from ILBgating (P<0.05) and extrinsic gating (P<0.05).

Influence of gating on time-intensity curves: peakenhancement

There was no difference between ILB-gated and frame-by-frame gated peak enhancement (Fig. 4). The mean differ-ence between extrinsically gated and frame-by-frame gatedexaminations was 3.5±5.5 %. The mean difference betweennon-gated and frame-by-frame gated examinations was21.5±33.9 %.

Influence of gating on time-intensity curves: retentionof VEGFR2-specific microbubbles

The retention of targeted contrast agents was determined ingated and non-gated CEUS clips for five mice and com-pared to the gold standard. Mean difference in the calculatedamount of bound targeted MB was 3.0±2.48 % for ILB-gated data and 11.25±7.635 % for non-gated data. Devia-tion from the gold standard was significantly lower (P<0.05) for ILB-gated data (Fig. 5). A representative exampleof the effect of gating on late enhancement is shown inFig. 6.

Examination time for frame-by-frame gating and ILB gating

The time to perform manual and ILB gating was measuredfor five functional ultrasound examinations and an addition-al five molecular ultrasound examinations with VEGFR2-specific microbubbles. The mean time needed to performour gold standard (manual frame-by-frame gating) for a 16 sultrasound clip in non-targeted contrast-enhanced ultra-sound was 20 min 48 s (SD 1 min 32 s). ILB gating took

Fig. 2 a–e Representative graphs of the different gating methods. aTime-intensity curve of an ROI around a landmark (gROI gating regionof interest). The drops in signal intensity represent the respiratorymovement during inspiration. A threshold (horizontal line) was de-fined manually to exclude frames from the cine loop. b Representativenon-gated time-intensity curve of an ROI within the liver. c–e Time-intensity curve of the same ROI after performing manual (gold stan-dard), intrinsic landmark-based (ILB) and extrinsic gating

�

Eur Radiol (2012) 22:1789–1796 1793

a mean time of 1 min 32 s (SD 11 s). For molecularultrasound analyses of the late enhancement the manualgating required 14 min 48 s (SD 2 min 50 s) per examina-tion, while only 1 min 19 s (SD 15 s) was needed for ILBgating. In both setups ILB gating was significantly faster(P<0.05).

Discussion

In this article, we describe an intrinsic landmark-based(ILB) gating technique for contrast-enhanced liver ultra-sound in freely breathing mice. Identification of a landmarkwhich is subject to respiratory movement resulted in the

ability to perform respiratory gating with a comparable oreven slightly higher accuracy than extrinsic gating using abiomonitoring device. The advantage of ILB-gated overnon-gated imaging was clearly demonstrated in contrast-enhanced liver imaging during first-pass perfusion analysisof non-targeted microbubbles and retention of VEGFR2-specific contrast agents. The utilisation of a standard clinicalultrasound system opens up the door for ILB-gated imagingnot only in mice but also in humans for functional andmolecular ultrasound studies.

Intrinsic landmark-based gating

The need for motion correction is a known fact for quanti-tative analysis of contrast-enhanced liver ultrasound. Due tothe long acquisition time necessary to assess all vascularphases, the simple use of apnoea phases alone is not feasi-ble, either in humans or in small animals. Thus, severalapproaches for motion correction have been applied [1,20]. As opposed to previous approaches, we decided toimplement an easily applicable and robust gating methodto review data acquired by a clinical ultrasound system. Byidentifying a landmark which is subject to respiratory mo-tion we derived a robust intrinsic gating signal that enabledus to automatically reject frames acquired during respiratorymotion. Implementation of ILB was technically easy toachieve: a custom-made post-processing tool for ultrasoundcine loops was used to generate time-intensity curves from aregion of interest (ROI). The time-intensity curve of a “gat-ing ROI” (which encompassed a moving landmark) wascopied to a standard spreadsheet application. A cut-off levelwas used to reject all frames of the cine-loop which wereacquired during the active part of the respiratory cycle.

In a clinical study including seven patients, Averkiou andcolleagues [1] used a similar approach: during expiration,they marked a reference position within the diaphragm andmanually rejected all frames where the diaphragm deviated

Fig. 3 Functional ultrasound imaging of the liver: analysis of the areaunder the curve. Differences of the areas under the curve of ILB-,extrinsically and non-gated time-intensity curves from the gold stan-dard (manual frame-by-frame gated time-intensity curves). ILB- andextrinsically gated time-intensity curves display smaller differences ascompared to the gold standard than non-gated data (P<0.05)

Fig. 4 Functional ultrasound imaging of the liver: analysis of the“peak enhancement”. Analysis of the “peak enhancement” from ILB-gated, extrinsically gated and non-gated time-intensity curves showsno significant differences as compared to the gold standard (manualframe-by-frame gated time-intensity curves). Nevertheless, in terms ofnon-significant differences, ILB gating shows no difference at allcompared to the gold standard, whereas the non-gated data show cleardifferences and much higher variances

Fig. 5 Molecular ultrasound imaging of the liver: analysis of the“microbubble retention”. Deviation from the gold standard is signifi-cantly lower in ILB-gated than in non-gated datasets concerning theretention of MB 5 min after contrast injection (P<0.05)

1794 Eur Radiol (2012) 22:1789–1796

from the reference position. In contrast to our study, they didnot acquire a gating signal but simply performed a manualframe-by-frame gating. We also used a manual frame-by-frame gating to obtain our “gold standard”. The results werevery smooth time-intensity curves without influence of res-piratory motion. But, as demonstrated, manual frame-by-frame gating is very time consuming. On the other hand,ILB gating required only a few steps of interaction: thereader has to find a suitable landmark ROI, copy that datato the spreadsheet application and then decide what thresh-old to use. Accordingly, we noted that ILB gating was

significantly quicker than manual gating. In studies handlingmany cases this will be a very relevant factor.

A comparable approach was proposed by Renault etal., who performed an independent component analysisto estimate respiratory motion [20]. Besides the changein signal intensity due to respiratory motion, this meth-od addresses (as an independent component) the changein signal intensity due to uptake of contrast agent. Inour animal study, we were able to identify landmarkswhich did not provide contrast enhancement such asparts of the liver capsule or the outer skin. Thus, therewas no need for an independent component analysis.Alternatively, ILB gating might be implemented by us-ing dual-mode ultrasound where the B-mode imagesshow little contrast enhancement and can be used forgating whereas the contrast-specific mode is used for lesioncharacterisation.

Extrinsic gating

Using our approach, we observed that ILB and extrinsicgating using a biomonitoring device provided statistical-ly comparable results compared to the gold standard. Tothe best of our knowledge, this is the first comparativestudy demonstrating the value of intrinsic gating in liverultrasound in mice. Bearing in mind the easy imple-mentation of intrinsic gating and the considerable costsof external gating devices, we propose ILB as thepreferable gating method. This holds true especiallywhen working with clinical ultrasound systems, whichrequire the utilisation of an external biomonitoring de-vice, most likely produced by a different company.Alternative options include dedicated small imagingsystems (e.g. high frequency ultrasound systems), whichoften provide an implemented biomonitoring device anda dedicated software solution for respiratory gating(both retrospective and prospective). In our institution,we also have such a high-frequency ultrasound system;however, we decided to use a clinical ultrasound systemin order to demonstrate that our method, which wasdeveloped for preclinical studies, has the potential tobe clinically translated. Moreover the clinical devicesprovide better sensitivity for phospholipid-shelled ultrasoundcontrast agents due to their excellent contrast-specific pulsesequences.

Another promising technique for motion correction ofultrasound studies is speckle tracking, which is predomi-nantly applied in echocardiography for quantification ofmyocardial deformation (e.g. for strain analysis) [4]. Anapplication for motion correction of liver ultrasound is fea-sible; however, it has not been described for contrast-enhanced liver studies to the best of our knowledge and willremain to be investigated in further studies.

Fig. 6 a–c Molecular ultrasound examination. a Time-intensity curveof an ROI around a landmark (gROI gating region of interest). Thehigh peak represents the destructive burst pulse. b Representative non-gated late enhancement curve of a molecular ultrasound examination.Signal intensity 5 min after contrast injection and after burst pulse. Toquantify the signal derived from targeted microbubbles, the mean SIafter the destructive pulse is subtracted from the mean SI before thepulse. c Late enhancement of the same ROI after ILB gating wasperformed

Eur Radiol (2012) 22:1789–1796 1795

Functional and molecular imaging

A crucial precondition for molecular imaging is the quanti-tative analysis of microbubble-derived signals within a le-sion. Previous studies on molecular ultrasound imaginghave been performed in subcutaneously implanted tumoursor organs not affected by respiratory movement [5, 14, 16,21]. As a result, it was not technically challenging to quan-tify the signal derived from microbubbles at their target:several minutes after injection a dynamic 2D dataset isacquired before and after a destructive pulse [5, 21]. Sub-traction of the mean signal intensity before and after thedestructive pulse reflects the amount of site-targeted bub-bles. In non-gated time-intensity curves of the liver, weobserved a considerable influence of motion artefacts onthe quantified signal, which was much lower in the ILB-gated datasets. These results demonstrate the benefit of agating protocol for molecular imaging when investigatingan organ subject to respiratory movement. This will be ofparticular interest for longitudinal studies where the expres-sion of disease-associated molecular markers is monitoredduring treatment to indicate response or non-response [14].

A limitation of our study was the fact that the extrinsicbiomonitoring device was not available during the molecularultrasound studies. Therefore, in the first part of our study, wevalidated our ILB-gating technique (against our gold stan-dard) and compared it to an established extrinsic gating pro-tocol. In the second part, we then investigated whether ILBgating provides a benefit for molecular imaging.

In conclusion, ILB gating is an easily implementablemethod for motion correction of ultrasound cine loops,which we propose as a favourable and less expensive alter-native to external biomonitoring devices. Its applicability toultrasound data acquired with a clinical ultrasound systemenables ILB-gated imaging not only in mice but also inhumans.

Acknowledgements This work was supported by the German Min-istry for Education and Research (BMBF), project “Virtual LiverConsortium”, number 0315743 and by the German Research Founda-tion (DFG) SFB TRR57.

References

1. Averkiou M, Lampaskis M, Kyriakopoulou K et al (2010) Quan-tification of tumor microvascularity with respiratory gated contrastenhanced ultrasound for monitoring therapy. Ultrasound Med Biol36:68–77

2. Beraza N, Malato Y, Sander LE et al (2009) Hepatocyte-specificNEMO deletion promotes NK/NKT cell- and TRAIL-dependentliver damage. J Exp Med 206:1727–1737

3. Bettermann K, Vucur M, Haybaeck J et al (2010) TAK1 sup-presses a NEMO-dependent but NF-kappaB-independent pathwayto liver cancer. Cancer Cell 17:481–496

4. Blessberger H, Binder T (2010) Non-invasive imaging. Two di-mensional speckle tracking echocardiography: basic principles.Heart 96:716–722

5. Bzyl J, Lederle W, Rix A et al (2011) Molecular and functionalultrasound imaging in differently aggressive breast cancer xeno-grafts using two novel ultrasound contrast agents (BR55 andBR38). Eur Radiol 21:1988–1995

6. Garbow JR, Ackerman JJ (2011) Imaging primary lung cancers inmice to study radiation biology: in regard to Kirsch et al. (Int JRadiat Oncol Biol Phys 2010;76:973–977). Int J Radiat Oncol BiolPhys 79:959, author reply 959

7. Hotz HG, Reber HA, Hotz B et al (2003) An orthotopic nudemouse model for evaluating pathophysiology and therapy of pan-creatic cancer. Pancreas 26:e89–e98

8. Kang BH, Siegelin MD, Plescia J et al (2010) Preclinical characteriza-tion of mitochondria-targeted small molecule hsp90 inhibitors, gami-trinibs, in advanced prostate cancer. Clin Cancer Res 16:4779–4788

9. Kerbel RS, Cornil I, Theodorescu D (1991) Importance of ortho-topic transplantation procedures in assessing the effects of trans-fected genes on human tumor growth and metastasis. CancerMetastasis Rev 10:201–215

10. Kiessling F, Gaetjens J, PalmowskiM (2011)Application ofmolecularultrasound for imaging integrin expression. Theranostics 1:127–134

11. Kiessling F, Huppert J, Palmowski M (2009) Functional and mo-lecular ultrasound imaging: concepts and contrast agents. CurrMed Chem 16:627–642

12. Mule S, Kachenoura N, Lucidarme O et al (2011) An automaticrespiratory gating method for the improvement of microcirculationevaluation: application to contrast-enhanced ultrasound studies offocal liver lesions. Phys Med Biol 56:5153–5165

13. Orlacchio A, Bolacchi F, Petrella MC et al (2011) Liver contrastenhanced ultrasound perfusion imaging in the evaluation of chronichepatitis C fibrosis: preliminary results. Ultrasound Med Biol 37:1–6

14. Palmowski M, Huppert J, Ladewig G et al (2008) Molecularprofiling of angiogenesis with targeted ultrasound imaging: earlyassessment of antiangiogenic therapy effects. Mol Cancer Ther7:101–109

15. Palmowski M, Lederle W, Gaetjens J et al (2010) Comparison ofconventional time-intensity curves vs. maximum intensity overtime for post-processing of dynamic contrast-enhanced ultrasound.Eur J Radiol 75:e149–e153

16. Palmowski M, Peschke P, Huppert J et al (2009) Molecular ultra-sound imaging of early vascular response in prostate tumors irra-diated with carbon ions. Neoplasia 11:856–863

17. Paprottka PM, Cyran CC, Zengel P et al (2010) Non-invasivecontrast enhanced ultrasound for quantitative assessment of tumormicrocirculation. Contrast mixed mode examination vs. only con-trast enhanced ultrasound examination. Clin Hemorheol Microcirc46:149–158

18. Pillai R, Marinelli ER, Fan H et al (2010) A phospholipid-PEG2000conjugate of a vascular endothelial growth factor receptor 2(VEGFR2)-targeting heterodimer peptide for contrast-enhanced ul-trasound imaging of angiogenesis. Bioconjug Chem 21:556–562

19. Pochon S, Tardy I, Bussat P et al (2010) BR55: a lipopeptide-basedVEGFR2-targeted ultrasound contrast agent for molecular imagingof angiogenesis. Invest Radiol 45:89–95

20. Renault G, Tranquart F, Perlbarg V, Bleuzen A, Herment A,Frouin F (2005) A posteriori respiratory gating in contrastultrasound for assessment of hepatic perfusion. Phys Med Biol50:4465–4480

21. Willmann JK, Paulmurugan R, Chen K et al (2008) US imaging oftumor angiogenesis with microbubbles targeted to vascular endothe-lial growth factor receptor type 2 in mice. Radiology 246:508–518

22. Willmann JK, van Bruggen N, Dinkelborg LM, Gambhir SS(2008) Molecular imaging in drug development. Nat Rev DrugDiscov 7:591–607

1796 Eur Radiol (2012) 22:1789–1796