Tcb

WBa

Cb

c

d

S

a

ARRA

KTCDrBAB

f

((l

h0

European Journal of Radiology 83 (2014) 1098–1105

Contents lists available at ScienceDirect

European Journal of Radiology

j ourna l h o mepage: www.elsev ier .com/ locate /e j rad

hree-dimensional contrast enhanced ultrasound score and dynamicontrast-enhanced magnetic resonance imaging score in evaluatingreast tumor angiogenesis: Correlation with biological factors

an-Ru Jiaa,1, Wei-Min Chaib,2, Lei Tanga,1, Yi Wanga,1, Xiao-Chun Fei c,3,ao-San Hand,4, Man Chena,∗

Department of Diagnostic Ultrasound, Rui Jin Hospital, School of Medicine, Shanghai Jiao Tong University, No. 197 Rui Jin 2nd Road, Shanghai 200025,hinaDepartment of Radiology, Rui Jin Hospital, School of Medicine, Shanghai Jiao Tong University, No. 197 Rui Jin 2nd Road, Shanghai 200025, ChinaDepartment of Pathology, Rui Jin Hospital, School of Medicine, Shanghai Jiao Tong University, No. 197 Rui Jin 2nd Road, Shanghai 200025, ChinaDepartment of Comprehensive Breast Health Center, Rui Jin Hospital, School of Medicine, Shanghai Jiao Tong University, No. 197 Rui Jin 2nd Road,hanghai 200025, China

r t i c l e i n f o

rticle history:eceived 24 November 2013eceived in revised form 8 March 2014ccepted 21 March 2014

eywords:hree-dimensional ultrasoundontrast-enhanced ultrasoundynamic contrast-enhanced magnetic

esonance imagingreast tumorngiogenesisiological factors

a b s t r a c t

Objective: To explore the clinical value of three-dimensional contrast enhanced ultrasound (3D-CEUS)and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) score systems in evaluatingbreast tumor angiogenesis by comparing their diagnostic efficacy and correlation with biological factors.Methods: 3D-CEUS was performed in 183 patients with breast tumors by Esaote Mylab90 with SonoVue(Bracco, Italy), DCE-MRI was performed on a dedicated breast magnetic resonance imaging (DBMRI)system (Aurora Dedicated Breast MRI Systems, USA) with a dedicated breast coil. 3D-CEUS and DCE-MRI score systems were created based on tumor perfusion and vascular characteristics. Microvesseldensity (MVD), vascular endothelial growth factor (VEGF) and matrix metalloproteinases (MMP-2, MMP-9) expression were measured by immunohistochemistry.Results: Pathological results showed 35 benign and 148 malignant breast tumors. MVD (P = 0.000, r = 0.76),VEGF (P = 0.000, r = 0.55), MMP-2 (P = 0.000, r = 0.39) and MMP-9 (P = 0.000, r = 0.41) expression were allsignificantly different between benignity and malignancy. Regarding 3D-CEUS 4 points as cutoff value,the sensitivity, specificity and accuracy were 85.1%, 94.3% and 86.9%, respectively, and correlated wellwith MVD (P = 0.000, r = 0.50), VEGF (P = 0.000, r = 0.50), MMP-2 (P = 0.000, r = 0.50) and MMP-9 (P = 0.000,r = 0.66). Taking DCE-MRI 5 points as cutoff value, the sensitivity, specificity and accuracy were 86.5%,

94.3% and 88.0%, respectively and also correlated well with MVD (P = 0.000, r = 0.52), VEGF (P = 0.000,r = 0.44), MMP-2 (P = 0.000, r = 0.42) and MMP-9 (P = 0.000, r = 0.35).Conclusions: 3D-CEUS score system displays inspiring diagnostic performance and good agreement withDCE-MRI scoring. Moreover, both score systems correlate well with MVD, VEGF, MMP-2 and MMP-9expression, and thus have great potentials in tumor angiogenesis evaluation.

∗ Corresponding author. Tel.: +86 021 64370045/+86 021 64600303;ax: +86 021 64333548.

E-mail addresses: jiawanru@126.com (W.-R. Jia), chai weimin@yahoo.com.cnW.-M. Chai), jessietang1003@163.com (L. Tang), xiatian.0602@163.comY. Wang), xcf0222@163.com (X.-C. Fei), hanbaosan@126.com (B.-S. Han),ucyjia1370@126.com (M. Chen).

1 Tel.: +86 021 64370045/+86 021 643700303; fax: +86 021 64333548.2 Tel.: +86 021 64370045/+86 021 64600314; fax:+86 021 64333548.3 Tel.: +86 021 64370045/+86 021 64662235; fax:+86 021 64333548.4 Tel.: +86 021 64370045/+86 021 64602203; fax: +86 021 64333548.

ttp://dx.doi.org/10.1016/j.ejrad.2014.03.027720-048X/© 2014 Elsevier Ireland Ltd. All rights reserved.

© 2014 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

Tumor angiogenesis played a central role in local tumorgrowth, invasion and distant metastasis, which was first con-firmed in breast cancer [1]. Neovasculature of malignant tumorwas entirely different from benign tumor both in morphol-ogy and hemodynamics. Therefore, the characteristics of tumorangiogenesis are of paramount importance in differentiation

between benign and malignant lesions. Currently, breast tumorangiogenesis could be assessed by contrast-enhanced ultrasound(CEUS), dynamic contrast-enhanced computed tomography(CT)and dynamic contrast-enhanced magnetic resonance imaging

al of R

(vttmigvatuihrlrsmptfaiStecdWlr

dsbm

2

2

beaiacw5sbwort0

2

2

fGf2o

W.-R. Jia et al. / European Journ

DCE-MRI), which are all able to quantify tumor perfusion, bloodolume and permeability in capillary level, and further displayumor characteristics in hemodynamics for their better spatial andemporal resolution [2]. Compared with CT or MR contrast agent,

icrobubble ultrasound contrast agents are truly blood pool imag-ng agent which would not diffuse into cell gap, and thereafterreatly improves the sensitivity of blood flow detection in lowelocity, accurately reflects breast tumor perfusion, and contributes

lot to breast tumor differentiation [3]. Combining the advan-ages of CEUS and 3D-US, three-dimensional contrast enhancedltrasound (3D-CEUS) can objectively evaluate tumor vascular-

ty by reconstruction of stereoscopic images. In particular, withigh resolution and bilateral breast imaging, breast MRI is supe-ior to other methods in detection of multi-center and multi-focalesion for various sequence post-processing and three-dimensionaleconstruction. Plenty of studies have demonstrated that the sen-itivity and accuracy of DCE-MRI were significantly higher thanammography or conventional US by differential diagnosis in mor-

hological and hemodynamic characteristics of enhanced breastumor, whereas the specificity is somewhat lower [4,5]. Based onormer results, we considered that for both 3D-CEUS and DCE-MRI,n objective score system is essential because just one character-stic cannot completely reflect the versatile vascular distribution.imultaneously, microvessel density (MVD) has been accepted ashe gold standard for evaluation of tumor angiogenesis [6]. Vascularndothelial growth factors (VEGF) have been found to be importantytokines in regulating endothelial cell proliferation and functionevelopment, which may also influence the expression of MVD [7],hile MMPs family has been proved to be actively involved in bio-

ogical changes of cancer, MMP-2, MMP-9 are especially closelyelated to tumor invasion and metastasis [8].

Therefore, our research aims at prospectively comparing theiagnostic efficacy and feasibility of 3D-CEUS score and DCE-MRIcore and their association with these biological factors in clinicalreast tumor application, thereafter providing a basis for assess-ent of breast tumor angiogenesis by 3D-CEUS and DCE-MRI.

. Materials and methods

.1. Study population

From May 2011 to February 2012, 183 consecutive patients withreast lesions treated by a multidisciplinary team approach werentered prospectively into a database in our institution, which waspproved by institutional review board committee, and patients’nformed consent was obtained. All of our investigation was inccordance with The Code of Ethics of the World Medical Asso-iation (Declaration of Helsinki). Among these 183 patients, 182ere female, one was male (age range, 20–76 years; mean age,

0.25 years). If a patient had multiple lesions, only the most con-picuous lesion was observed. All patients underwent preoperativeaseline US, 2D-CEUS, 3D-CEUS and MRI, and subsequently under-ent surgery (modified radical mastectomy or radical mastectomy

r conservative breast surgery) within one week. None of themeceived preoperative neoadjuvant chemotherapy or endocrineherapy. The mean diameter of the lesions was 2.6 cm ± 1.3 (range,.6–6.5 cm).

.2. Ultrasound examination and evaluation

.2.1. Ultrasound examinationConventional US, 2D-CEUS, and 3D-CEUS scanning were per-

ormed using the same ultrasound machine Mylab 90 (Esaote,

enoa, Italy). Conventional US and color Doppler US were per-

ormed by LA 532 transducer with a frequency of 13–4 MHz, whileD-CEUS was evaluated by LA 522 transducer with a frequencyf 9–3 MHz, the BL 433 volume transducer with a frequency of

adiology 83 (2014) 1098–1105 1099

15–9 MHz was used for 3D scanning. The contrast agent wasSonoVue (BR1, Bracco SpA, Milan, Italy), a sulfur hexafluoride-filledmicrobubble contrast agent. To avoid interobserver variability, allthe US scanning was performed by one radiologist with 5 yearsof experience in breast CEUS, and one year in breast 3D-CEUS. Allultrasound scanning were performed usually less than one weekbefore surgery.

First, conventional US, including color Doppler US and baseline3D scanning were carried out, respectively, to observe general fea-tures of breast tumors and to select the best tumor imaging inthe maximum plane, from which both the tumors and the nor-mal adjacent breast tissue could be observed. In the meantime,the appropriate volume angle was defined so that the whole lesionwould be included in the volume data without signal loss. Imagingparameter settings were optimized to ensure high quality imagesafter the target lesion was determined. Subsequently, 2D-CEUS wasperformed by LA 522 transducer with a frequency of 9–3 MHz.The scanner settings for CEUS were as follows: the selected planeincluded the lesion and its surrounding normal tissue if possible;range, 70 dB; the image depth was 3 or 4 cm. When the signalsfrom the microbubbles in the large vessels such as axillary veindisappeared, 3D-CEUS was initiated. The contrast agent SonoVuewas injected as the same dose and fashion with 2D-CEUS (2.4 mLof SonoVue as a bolus through an antecubital vein, followed by aflush of 5 mL of 0.9% saline). Ten seconds later, 3D-CEUS imageswere continuously obtained more than five times with the totaltime over 2 min. The imaging settings for 3D-CEUS was as follows:MI, 0.08–0.13; one focal zone; power output, 3–6%; dynamic range,40–60 dB; volume angle, 30–50◦ and the scanning route was consis-tent with that for 2D-CEUS. During the 3D scanning, the transducerwas kept in a stable position without movement and the patientwas asked to hold the breath for 5–10 s depending on the sizeof the volume data and the acquisition mode. Transparent modeand tomographic mode imaging (TMI) were mainly selected todepict neovascularization. After that, the reformatted CEUS imagesof three orthogonal planes were displayed on the screen. All thedata, including CEUS and 3D CEUS images were stored in the harddisk of the ultrasonography machine in the DICOM format for fur-ther analysis.

2.2.2. 3D-CEUS imaging analysisThe 3D-CEUS measurements were analyzed by 2 investigators

who did not perform US examinations and were blind to surgical,histological information and other imaging findings. After indepen-dent interpretations by the two observers, consensus was obtainedin conference, and if different assessment were assigned, a consen-sus reached after discussion. Based on our previously study [21]the characteristics of tumor vasculature were focused to assessas follows: (1) peripheral vessels (presence or absence), their dis-tribution (radial or not), courses (distorted or not) and dilateddegree (coarse or not), (2) penetrating vessels (presence or absence)and their courses(running inside tumor or towards center), (3)rim perfusion (presence or absence) and degree (thin/moderateor coarse) (4) intratumoral vessels (presence or absence) andtheir dilated degree (coarse or not). According to the correlationbetween the described diagnostic characteristics and the probabil-ity of malignancy, a score from 0 to 2 was given, respectively, foreach characteristic. The diagnostic and scoring criteria for 3D-CEUScharacteristics were summarized in Table 1.

2.3. MRI examination and evaluation

2.3.1. MRI examinationBreast MR imaging was performed on a dedicated breast mag-

netic resonance imaging (DBMRI) system (Aurora Dedicated BreastMRI Systems, USA) with a dedicated breast coil while the patients

1100 W.-R. Jia et al. / European Journal of Radiology 83 (2014) 1098–1105

Table 13D-CEUS scoring criterion in benign and malignant breast tumors.

3D-CEUS characteristics 0 1 2

Peripheral vesselsPresence and distribution Absent Present, not radial Present, radialDilated degree and courses Not coarse, not distorted Not coarse, distorted Coarse, distorted or not

wasmfG(stt

2

bARtsbttgpspmamalentrimcoi(rsAa

2

2

ytdate

Penetrating vessels Absent

Rim perfusion Absent

Intratumoral vessels Absent

ere in the prone position. After a localizer on the axial imagend coil calibration, a dynamic series of axial T1-weighted fat-uppression images (TR 29 ms, TE 4.8 ms, slice thickness 1.1 mm,atrix: 360 × 360 × 128, FOV 36 cm), including one pre- and

our post-contrast scans, were obtained. Gd-DTPA (Magnevist,ermany) was administered using a bolus intravenous injection

1.5 mL/s) at a dose of 0.2 mmol/kg body weight followed by a 20 mLaline solution flush. The dynamic images were acquired 90 s afterhe contrast media injection. The scan time was 3 min per scan, andhe total time was 12 min.

.3.2. MRI imaging analysisThe MRI images were prospectively interpreted by two off-sight

reast radiologists with more than 5 years’ experience. A dedicatedURORA CAD workstation was used for image post-processing.egion of interest (ROI) was placed within the tumor area withhe highest signal intensity enhancement, and time-signal inten-ity curve (TIC) was obtained by station software. Both readers werelinded to surgical, histological and other imaging findings beforehe evaluation. Each reader analyzed MRI characteristics includinghe maximum diameter, area, morphological features (shape, mar-in, distribution and internal enhancement), and the enhancementattern individually, then they jointly reviewed the controver-ial imaging and finally reached a consensus. The enhancementattern included homogeneous, heterogeneous and rim enhance-ent; TIC curve was classified into type I (persistent), II (plateau)

nd III (wash-out) in the light of Kuhl classification [9], but ifore than two types occurred, the type with highest probability

nd highest grades were selected after repeated multiple samp-ing; The initial signal increase was classified into <50% (mildnhancement), 50–100% (moderate enhancement), >100% (sig-ificant enhancement). Compared with the contralateral breast,he richness and dilated degree of peripheral vessels include notich, richer and richer and coarser, according to three-dimensionalmages reconstructed by multi-planar reconstruction (MPR) and

aximum intensity projection (MIP); The penetrating and encir-ling vessels were divided into thin and coarse penetrating vessels,n the basis of their presence and dilated degree in reconstructedmages (Fig. 1). All these indices were improved to criteria as followsTable 2): enhancement pattern, TIC curve, initial signal increase,ichness and dilated degree of peripheral vessels, penetrating ves-els and their dilated degree, relying on Fischer scoring system [10].

score of 0 to 2 points was given, respectively, for each indexccording to whether they are suspicious for malignancy.

.4. Histologic analysis

.4.1. Pathological analysisAll cases were specifically evaluated by one pathologist with 10

ears’ experience in breast diseases. The maximal histologic sec-ion of each mass was selected for comparison with US images. For

oubtful lesions, consensus was obtained between the radiologistnd pathologist by comparing CEUS images, MRI images and his-ologic slides in terms of mass localization and outlines and thenhancing and non-enhancing areas.

Running inside tumor Running towards centerThin/moderate CoarseThin/smooth Coarse

2.4.2. Immunohistochemical staining for CD34 MVD, VEGF,MMP-2 and MMP-9 expression

Microvessels were quantified by the counting procedure of Wei-dner et al. [11]. First each slide was examined at a low magnificationto identify 3 separate areas containing the greatest number ofmicrovessels (“hot spots”), then individual vessels from these hotspots were subsequently counted on a ×200 field. Finally ImagePro-plus 6.0 (IPP, Media Cybernetics, Silver Spring, Maryland, USA)were used to calculate the area of hot spots and integrated opticaldensity (IOD) of internal specific staining part semi-quantitatively,then taking three times the average of mean density (IOD/area) asMVD. It’s worth noting that all pictures should be taken under thesame condition.

VEGF, MMP-2 and MMP-9 expression were determined byimmunohistochemistry (IHC). Regarding membrane or cell plasmbrown products as positive value, and observing 10 high powerfield randomly, then counting 100 cells per field to evaluate ratio ofstained cells and staining intensity semi-quantitatively, finally forcomprehensive assessment. (1) The ratio of stained cells was scoredas follows: 0 = no stained cells in any microscopic field, 1 = 1–10%of tumor cells stained positively, 2 = 11–50% of tumor cells stainedpositively, and 3 = 51–100% of tumor cells stained positively. (2)Intensity of the stain was scored on the following scale: 0 = nostaining, 1 = mild staining, 2 = moderate staining, and 3 = intensestaining. (3) Finally the overall scores were assessed as follows:overall scores = scores of stained cell ratio plus scores of stainingintensity. The median VEGF, MMP-2 and MMP-9 staining scoreswas selected as cutoff values to categorize the tumors into lowly(0–3) – and highly (4–6) – expressed tumors (Fig. 2).

2.5. Statistical analysis

Histological diagnosis was used as golden standard for calcula-tion of sensitivity, specificity, accuracy, positive predict value (PPV)and negative predict value (NPV) of 3D-CEUS and DCE-MRI sco-ring system in differentiating malignant and benign breast lesions.Statistical analyses were performed with SPSS version 17.0 soft-ware for Windows (SPSS Inc, Chicago, IL). Spearman correlationcoefficients (r values) were calculated and classified as follows:lowly (|r|: 0.3–0.5), significantly (|r|: 0.5–0.8), and highly correlated(|r|: 0.8–1.0).�2 tests were used to compare variables and a valueof P < 0.05 was considered to be statistically significant.

3. Results

3.1. Histologic diagnosis

Tissue specimens of 183 breast tumors were obtained fromsurgical resection. Histological analysis showed 148 malignantones, including invasive ductal carcinoma (n = 108), mixed inva-sive carcinoma (n = 18), ductal carcinoma in situ (n = 8), invasivelobular carcinoma (n = 4), apocrine carcinoma (n = 4), invasive

micropapillary carcinoma (n = 2), metaplastic carcinoma (n = 2), andneuroendocrine carcinoma (n = 2), while 35 benign breast lesions,including fibroadenoma (n = 28), intraductal papilloma (n = 5), andphyllodes tumor (n = 2).

W.-R. Jia et al. / European Journal of Radiology 83 (2014) 1098–1105 1101

Fig. 1. DCE-MRI features after three-dimensional reconstruction: (a) compared with the contralateral breast, peripheral vessels (arrow) of tumor in the left breast are richerand coarser. Histological analysis showed invasive ductal carcinoma (b) coarse penetrating vessels with branches in the right breast (arrow). Histological analysis showedinvasive ductal carcinoma.

Table 2DCE-MRI scoring criteria in benign and malignant breast tumors.

DCE-MRI index 0 1 2

Enhancement pattern Homogeneous Heterogeneous Rim enhancementTIC curve Type I (persistent) Type II(plateau) Type III (wash-out)Initial signal increase Mild Moderate Significant

3a

w

F×

Peripheral vessels Not rich

Penetrating vessels None

.2. CD34 MVD, VEGF, MMP-2 and MMP-9 expression in benign

nd malignant breast tumors

For 183 benign and malignant lesions, the mean IOD of MVDas 8686.82 ± 4.32E3, those of benign was 3246.55 ± 1.64E3, and

ig. 2. Expression of biological factors: (a) high MVD distribution (CD34-immunostaining40); (c) high MMP-2 expression (IHC, original magnification ×100); (d) high MMP-9 exp

Richer Richer and coarserThin/moderate Coarse/Encircling

malignant was 9973.37 ± 3.70E3 (P = 0.000). The best cutoff value

of the mean IOD of MVD was 7001.67. High VEGF expression wasfound in 137 patients, while low expression in 46 patients. Besides,MMP-2 lowly expressed in 53 cases, and highly expressed in 130cases (P = 0.000), while low MMP-9 expression was in 31cases and

, original magnification ×200); (b) low VEGF expression (IHC, original magnificationression (IHC, original magnification ×100).

1102 W.-R. Jia et al. / European Journal of Radiology 83 (2014) 1098–1105

Table 3Diagnostic performance of MVD, VEGF, MMP-2 and MMP-9 Expression.

Sensitivity (%) Specificity (%) Accuracy (%) AUC 95% CI P value r

MVD 87.8 100.0 90.2 0.939 0.906–0.972 0.000 0.76

hoTfb

3b

sw(fdfticawEp

Fhac

VEGF 86.5 74.3 60.8

MMP-2 79.7 65.7 45.4

MMP-9 90.5 48.6 39.1

igh expression in 152 cases (P = 0.000). The diagnostic efficiencyf MVD, VEGF, MMP-2 and MMP-9 expression were delineated inable 3, from which we can concluded that all these biologicalactors were significantly different between benign and malignantreast tumors and correlated well with histology results.

.3. 3D-CEUS characteristics, scoring system and association withiological factors

Of all 183 breast tumors, 171 (93.4%) showed peripheral ves-els. Moreover, their distribution, courses and dilated degreeere all significantly different between benignity and malignancy

P = 0.000). Particularly, penetrating vessels (P = 0.006), rim per-usion (P = 0.018), coarse rim perfusion (P = 0.000) and dilatedegree of intratumoral vessels (P = 0.008) were all statistically dif-erent. 3D-CEUS scores of malignancy were significantly higherhan that of benignity of all 183 patients with breast tumor, accord-ng to characteristics of peripheral vessels (presence, distributingourses and dilated degree), penetrating vessels, rim perfusion

nd intratumoral vessels. The mean score of benign tumorsas 2.09 ± 1.067 and 5.77 ± 2.122 of malignant ones (P = 0.000).

xcept for 2 benign cases, all tumors with score higher than 4oints were proved to be malignant. The sensitivity, specificity,

ig. 3. Conventional US, 3D-CEUS and DCE-MRI characteristics in a 62-year-old womaypoechoic tumor with indistinct margin, irregular shape and rich vascularity in the left bround the tumor; (c) 3D-CEUS depicts peripheral vessels, which are distorted and coaarcinoma (H&E staining, original magnification ×100).

0.804 0.713–0.894 0.000 0 .550.727 0.628–0.827 0.000 0.390.696 0.628–0.827 0.000 0.41

accuracy, PPV and NPV of 4 points in differential diagnosis were85.1%, 94.3%, 86.9%, 98.4% and 60.0%. The correlation between 4points of 3D-CEUS scoring system and pathology was significant(P = 0.000, r = 0.68) (Table 4). Regarding 4 points of 3D-CEUS sco-ring system as the differential criterion, significant correlation wasfound with the best cutoff value of MVD. Additionally, it was alsostatistically associated with VEGF, MMP-2 and MMP-9 (Table 5).

3.4. DCE-MRI characteristics, scoring system and association withbiological factors

Among all 183 patients, 148 showed heterogeneous enhance-ment without significant difference between benign and malignantbreast tumors (P > 0.05). TIC curve was made up of type I of 21 cases,type II of 29 cases and type III of 133 cases, within which 94.7%(126/133) were malignant ones (P = 0.000). Moreover, initial sig-nal intensity was <50% in 14 patients, 50–100% in 135 cases and>100% in 34 cases, with significant difference between benignityand malignancy (P = 0.002). As for peripheral vessels, they were

not rich in 65 cases, richer in 34 cases and richer and coarser in84 cases (P = 0.000), compared to the contralateral breast. 73.8%(135/183) patients showed penetrating vessels, among whichthin or moderate in 59 patients, and furthermore, coarse and

n with invasive ductal carcinoma in the left breast. (a) conventional US shows areast; (b) DCE-MRI displays richer peripheral vessels and coarse penetrating vesselsrse and distributed radially; (d) histologic analysis demonstrates invasive ductal

W.-R. Jia et al. / European Journal of Radiology 83 (2014) 1098–1105 1103

Table 4Diagnostic performance of 3D-CEUS and DCE-MRI score.

Cutoff points Sensitivity (%) Specificity (%) Accuracy (%) NPV (%) PPV (%)

3D-CEUS score 4 points 85.1 94.3 86.9 98.4 60.0DCE-MRI score 5 points 86.5 94.3 88.0 98.5 62.3

Table 5Correlation between 3D-CEUS and DCE-MRI score with prognostic factors.

MVD VEGF MMP-2 MMP-9

P r P r P r P r

satrtDnwowcwDwas

4

vcHestovpaaciipmiwatlbcjgicio

C

3D-CEUS score 0.000 0.50 0.000

DCE-MRI score 0.000 0.52 0.000

urrounding the lesions in 76 cases. It is worth mentioning thatll these 76 cases with coarse penetrating vessels surrounding theumor were proved to be malignant (P = 0.000). As for DCE-MRI sco-ing system, the score of malignant tumors was obviously higherhan that of benign ones among all 183 patients with breast tumors.CE-MRI score of benign tumors were 2.71 ± 1.453, while malig-ant ones 6.79 ± 1.498, all cases with the score higher than 6 pointsere malignant (P = 0.000). Regarding 5 points of DCE-MRI as cut-

ff, the sensitivity, specificity, accuracy, PPV and NPV of 5 pointsere 86.5%, 94.3%, 88.0%, 98.5% and 62.3%, respectively. The asso-

iation between pathology and 5 points of DCE-MRI scoring systemas significant (P = 0.000, r = 0.70) (Table 4). Taking 5 points ofCE-MRI scoring system as criterion, it was significantly correlatedith MVD cutoff value, and lowly associated with VEGF, MMP-2

nd MMP-9. The comparison between 3D-CEUS and DCE-MRI washown in Table 5.

. Discussion

Angiogenesis is defined as the formation of new capillary bloodessels from existent microvessels by sprouting, and the cru-ial role it plays in tumor evolvement has been accepted [12].anahan hypothesized that the balance of inhibitors and induc-rs governs the angiogenic switch [13]. Malignant tumor cellsecrete inducers such as VEGF and elicit neovascularization to feedhe tumor. These new vessels are immature with features likebviously increased numbers, morphological changes, abnormalascular reticular structure, incomplete basal membrane and highermeability, which provide a pathophysiological foundation forngiogenesis evaluation of breast cancers. MVD has been accepteds the gold standard for evaluation of tumor angiogenesis [6]. Vas-ular endothelial growth factors (VEGF) have been found to bemportant cytokines in regulating endothelial cell proliferation andncreasing vascular permeability [7], While MMPs family has beenroved to be actively involved in biological changes such as inflam-ation, embryogenesis, angiogenesis, bone regeneration, tumor

nvasion and metastasis, from which MMP-2 and MMP-9 especiallyorked in breast cancer [8]. In our research, all MVD, VEGF, MMP-2

nd MMP-9 had significant differential diagnostic value in breastumors and correlated well with histological results. However, bio-ogical factors are more likely to be measured in preoperativeiopsy or postoperative resection by immunohistochemistry andan’t be performed repeatedly, especially for patients with neoad-uvant chemotherapy. Digital subtraction angiography (DSA) is theold standard for delineation of vessels in and around tumor, butts invasiveness limited the clinical application. Therefore, a moreonvenient, noninvasive and reproducible in vivo imaging modality

s in urgent need for evaluating tumor angiogenesis, chemotherapyutcome and prognosis.

Currently, breast tumor angiogenesis could be assessed by CEUS,T and DCE-MRI. Increased perfusion volume based on malignant

0.50 0.000 0.50 0.000 0.660.44 0.000 0.42 0.000 0.35

tumor angiogenesis produced fast and pronounced enhancementin all of these imaging modalities. Fischer et al. first demonstrateda comprehensive MRI score system of breast tumor according tomorphological and hemodynamic characteristics, including shape,margin, enhancement, initial signal intensity and TIC curve, andthe last three indices were proved to be significant in breast MRI[10]. In our research, we found heterogeneous enhancement in148 cases with no statistical difference, but initial signal inten-sity and TIC curve were different between benign and malignanttumors. Homogeneous enhancement usually suggested benignitywhile heterogeneous or rim enhancement indicated malignancy.TIC curve reflected the perfusion and clearance of breast tumorand directly described the contrast agent distribution in tissue.Kuhl et al. classified TIC curve and proposed that type I was sug-gestive of benignity, type III was indicative of malignancy [9].Compared to the contralateral breast, we found peripheral vesselswere obviously richer and coarser in malignant ones and all tumorswith coarse penetrating vessels were malignant. Therefore, basedon Fischer scoring, we created the new system, which preservedindices of enhancement, initial signal intensity and TIC curve,and added peripheral vessels (richness and dilated degree) andpenetrating vessels (dilated degree and courses) formed in recon-structed DCE-MRI images. We gave all indices 0–2 points accordingto previous studies and our results. It was reported that thesensitivity of DCE-MRI in diagnosis was 95–100%, while the speci-ficity was different with a range of 37–100% [4,5,14,15]. However,using three-dimensional AURORA CAD workstation and subtrac-tion images, we chose the lesion with signal enhancement andclearly showing the tortuous and coarse peripheral and penetrat-ing vessels with MIP, for this reason, our new score system laid anemphasis on malignant tumor vascular characteristics, overcamethe low specificity of the former studies, and thus improve its clin-ical value in breast cancer differentiation, staging and prognosis.

With the recent advent of second generation contrast agent,CEUS has become a more and more popular method for tumorangiogenesis evaluation. The correlation between CEUS and CT/MRIhas proved to be as high as 92% according to researches of approx-imately 100 thousand liver lesions [16]. In terms of correlationbetween MVD and VEGF, lots of studies have been carried out[17,7,18], Kim compared breast tumor vascularity evaluated byTc-99m MIBI scintimammography and power Doppler ultrasound(US) with MVD and reverse transcriptase-polymerase chain reac-tion (RT-PCR) of VEGF mRNA, then concluded that evaluation ofbreast cancer by power Doppler US with a microbubble contrastagent could predict tumor angiogenesis because power DopplerUS grading was well correlated with MVD (r = 0.552, P = 0.033), andMVD was also well associated with VEGF and VEGF mRNA level

at the same time. As for the diagnostic value in breast tumor,Ricci has found contrast sonographic enhancement pattern andperfusion curves correlated well with MRI with the sensitivity,specificity and accuracy of 69.2%, 66.7% and 68%, respectively

1104 W.-R. Jia et al. / European Journal of Radiology 83 (2014) 1098–1105

F omanh right

d trates

[eiowtidvtailidltc3g

iCtpbsdM3it3

ig. 4. Conventional US, 3D-CEUS and DCE-MRI characteristics in a 55-year-old wypoechoic tumor with circumscribed margin and intratumoral calcifications in theepicts coarse rim perfusion around the tumor; (d) histologic analysis also demons

19]. We have ever described 3D-CEUS perfusion feature, periph-ral vessels and intratumoral vessels of breast tumor and foundts great potentials in evaluating tumor angiogenesis [21]. Basedn all those results, we believed that an objective score systemas essential because just one feature cannot completely reflect

he versatile vascular distribution. Therefore, we picked out fivendices like peripheral vessels (distribution, courses and dilatedegree), penetrating vessels, rim enhancement and intratumoralessels, and gave them 0–2 points, respectively. Encouragingly,he scoring we created showed a high diagnostic performancend good agreement with DCE-MRI score. Combined with breastmaging reporting and data system (BI-RADS) of American col-ege of radiology (ACR), this 3D-CEUS score system can reveal thentratumoral and vascular biological characteristics, broaden theiagnostic value of BI-RADS, and make 3D-CEUS both a morpho-

ogical and a functional imaging modality [20]. Additionally, thesewo score systems were also well related to biological factors, espe-ially related to MVD, which strongly supported the view thatD-CEUS and DCE-MRI score can indirectly reflect tumor angio-enesis (Figs. 3 and 4).

Our study has some limitations. First, there were difficultiesn selecting patients who were willing to accept both 3D-EUS and DCE-MRI, for they were not only expensive but alsoime-consuming, so the patients were relatively small and largeopulation are needed for further studies. Second, it cannot belind for CEUS and 3D-CEUS because CEUS images were neces-ary for the investigator to determine the location and viewingirection of the ROI. Third, although both 3D-CEUS and DCE-RI can show breast tumor vasculature three-dimensionally,

D-CEUS images were less direct and less panoramic, thus all thendices were observed subjectively, so further studies were manda-ory for accumulating experience and improving application ofD-CEUS.

with invasive ductal carcinoma in the right breast. (a) Conventional US shows abreast; (b) DCE-MRI displays rim enhancement around the tumor; (c) 3D-CEUS also

invasive ductal carcinoma (H&E staining, original magnification ×100).

5. Conclusions

Our investigation indicated that both the 3D-CEUS and DCE-MRIscore systems we created had an inspiring diagnostic efficacy andcorrelated well with prognostic factors, especially with MVD, whichdemonstrated that these two modalities had great potentials inevaluating breast tumor angiogenesis, chemotherapy outcome andprognosis.

Conflict of interest statement

We declare that we have no financial and personal relationshipswith other people or organizations that can inappropriately influ-ence our work, there is no professional or other personal interestof any nature or kind in any product, service and/or company thatcould be construed as influencing the position presented in, or thereview of the manuscript entitled.

Acknowledgments

The work was supported in part by the National Natural Sci-ence Foundation of China (No. 81172078) and was also majorsubject of Huangpu District Health System in Shanghai Municipal(2013–2015).

References

[1] Weidner N, Semple JP, Welch WR, Folkman J. Tumor angiogenesis

and metastasis—correlation in invasive breast carcinoma. N Engl J Med1991;324(1):1–8.

[2] Teifke A, Behr O, Schmidt M, et al. Dynamic MR imaging of breast lesions: cor-relation with microvessel distribution pattern and histologic characteristics ofprognosis. Radiology 2006;239(2):351–60.

al of R

[

[

[

[

[

[

[

[

[

[

[

W.-R. Jia et al. / European Journ

[3] Stuhrmann M, Aronius R, Schietzel M. Tumor vascularity of breast lesions:potentials and limits of contrast-enhanced Doppler sonography. Am JRoentgenol 2000;175(6):1585–9.

[4] Lehman CD, Isaacs C, Schnall MD, et al. Cancer yield of mammography, MR, andUS in high-risk women: prospective multi-institution breast cancer screeningstudy. Radiology 2007;244(2):381–8.

[5] Kriege M, Brekelmans CT, Boetes C, et al. Efficacy of MRI and mammography forbreast-cancer screening in women with a familial or genetic predisposition. NEngl J Med 2004;351(5):427–37.

[6] Folkman J. Tumor angiogenesis: therapeutic implications. N Engl J Med1971;285(21):1182–6.

[7] Kim SW, Park SS, Ahn SJ, et al. Identification of angiogenesis in primarybreast carcinoma according to the image analysis. Breast Cancer Res Treat2002;74(2):121–9.

[8] Overall CM, Kleifeld O. Towards third generation matrix metalloproteinaseinhibitors for cancer therapy. Br J Cancer 2006;94:941–6.

[9] Kuhl CK, Mielcareck P, Klaschik S, et a1. Dynamic breast MR imaging: are sig-nal intensity course data useful for differential diagnosis of enhancing lesion?Radiology 1999;21l(1):101–10.

10] Fischer U, Kopka L, Grabbe E. Breast carcinoma: effect of preoperativecontrast-enhanced MR imaging on the therapeutic approach. Radiology

1999;213(3):881–8.

11] Weidner N, Barbarishi M, Gasparini G, et al. Microvascular density qualificationin breast carcinoma. Appl Immunohistochem 1995;9(2):75–9.

12] Battegay EJ. Angiogenesis: mechanistic insights, neovascular diseases, andtherapeutic prospects. J Mol Med 1995;73(7):333–46.

[

adiology 83 (2014) 1098–1105 1105

13] Hanahan D, Folkman J. Patterns and emerging mechanisms of theangiogenic switch during tumor angiogenesis. Cell 1996;86(3):353–64.

14] Brinck U, Fischer U, Korabiowska M, et al. The variability of fibroade-noma in contrast—enhanced dynamic MR mammography. Am J Roentgenol1997;168(5):1331–4.

15] Sinha S, Lucas-Quesada FA, DeBruhl ND, et al. Multifeature analysisof Gd-enhanced MR images of breast lesions. J Magn Reson Imaging1997;7(6):1016–26.

16] Harvey CJ, Albrecht T. Ultrasound of focal liver lesions. Eur Radiol2001;11(9):1578–93.

17] Ko EY, Lee SH, Kim HH, et al. Evaluation of tumor angiogenesis with a second-generation US contrast medium in a rat breast tumor model. Korean J Radiol2008;9(3):243–9.

18] Miles KA. Tumor angiogenesis and its relation to contrast enhancement oncomputed tomography: a review. Eur J Radiol 1999;30(3):198–205.

19] Ricci P, Cantisani V, Ballesio L, et al. Benign and malignant breast lesions:efficacy of real time contrast-enhanced ultrasound vs. magnetic resonanceimaging. Ultraschall Med 2007;28(1):57–62.

20] Shen WC, Chang RF, Moon WK. Computer aided classification system for breastultrasound based on Breast Imaging Reporting and Data System (BI-RADS).

Ultrasound Med Biol 2007;33(11):1688–98.

21] Chen M, Wang WP, Jia WR, et al. Three-dimensional contrast enhanced ultra-sound in the assessment of breast tumor angiogenesis: correlation withmicrovessel density and vascular endothelial growth factor expression. J Ultra-sound Med. in press.