RiM

ESU

a

ARRA

KBRMSM

1

brto

((

h0

European Journal of Radiology 83 (2014) 942–950

Contents lists available at ScienceDirect

European Journal of Radiology

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

eal-time MRI navigated US: Role in diagnosis and guided biopsy ofncidental breast lesions and axillary lymph nodes detected on breast

RI but not on second look US

lena Pastor Pons1, Francisco Miras Azcón1, María Culianez Casas ∗,alvador Martínez Meca1, José Luis García Espona1

niversity Hospital Virgen de las Nieves, Servicio de Radiodiagnóstico Hospital Materno Infantil, Avenida de las Fuerzas Armadas s/n, 18004 Granada, Spain

r t i c l e i n f o

rticle history:eceived 6 August 2013eceived in revised form 24 February 2014ccepted 1 March 2014

eywords:reast cancereal-time ultrasoundagnetic resonance imaging fusion

econd look ultrasoundagnetic resonance navigated ultrasound

a b s t r a c t

Objectives: To prospectively evaluate the accuracy of real-time ultrasound combined with supine-MRIusing volume navigation technique (RtMR-US) in diagnosis and biopsy of incidental breast lesions (ILSM)and axillary lymph nodes (LNSM) suspicious of malignancy on contrast enhanced magnetic resonanceimaging (CE-MRI).Materials and methods: Five hundred and seventy-seven women were examined using breast CE-MRI.Those with incidental breast lesions not identified after second-look ultrasound (US) were recruited forRtMR-US. Biopsy was performed in ILSM. Breast lesions were categorized with BI-RADS system and Fisher’exact test. Axillary lymph nodes morphology was described. To assess efficacy of RtMR-US, diagnosticaccuracy, sensitivity, specificity, detection rate and Kappa index of conventional-US and RtMR-US werecalculated.Results: Forty-three lesions were detected on CE-MRI before navigation. Eighteen were carcinomas and25 ILSM. Of these, 21 underwent a RtMR-US. Detection rate on RtMR-US (90.7%) was higher than onconventional-US (43%) (p < 0.001). Agreement between both techniques was low (k = 0.138). Twenty ILSMand 2 LNSM were biopsied. Sixty-five percent were benign (100% of BI-RADS3 and 56% of BI-RADS4-5). Diagnostic performance of RtMR-US identifying malignant nodules for overall lesions and for thesubgroup of ILSM was respectively: sensitivity 96.3% and 100%, specificity 18.8% and 30.7%, positivepredictive value 66.7% and 43.7%, negative predictive value 75% and 100%. In addition RtMR-US enabled

biopsy of 2 metastatic lymph nodes.Conclusions: Real time-US with supine-MRI using a volume navigation technique increases the detectionof ILSM. RtMR-US may be used to detect occult breast carcinomas and to assess cancer extension, pre-venting unnecessary MRI-guided biopsies and sentinel lymph node biopsies. Incidental lesions BI-RADS3 non-detected on conventional-US are probably benign.

. Introduction

Contrast enhanced magnetic resonance imaging (CE-MRI) of thereast has become an essential technique in the preoperative loco

egional staging of patients with breast cancer, in the evaluation ofheir response to neoadjuvant chemotherapy and also in the studyf an unknown primary malignancy [1–3]. Breast MRI has high

∗ Corresponding author. Tel.: +34 958020463; fax: +34 958020653.E-mail addresses: elenapastorpons@gmail.com (E.P. Pons), frmiaz00@gmail.com

F.M. Azcón), mariacc1980@gmail.com (M.C. Casas), isalvaa@hotmail.comS.M. Meca), gespona@hotmail.com (J.L.G. Espona).

1 Tel.: +34 958020463.

ttp://dx.doi.org/10.1016/j.ejrad.2014.03.006720-048X/© 2014 Elsevier Ireland Ltd. All rights reserved.

© 2014 Elsevier Ireland Ltd. All rights reserved.

sensitivity, ranging from 94% to 100% for the detection of breastcancer and also for other lesions that are occult in other imagingtechniques such as mammography or ultrasound (US). HoweverMRI is limited by its variable specificity (E), ranging from 20%to 100%, and low positive predictive value (PPV) [4–6]. Inciden-tal lesions previously not detected by other imaging techniquesare found in 6–34% of patients with breast cancer who undergopreoperative breast MRI [7]. Moreover in 10–34% of patients, addi-tional foci of mammographically occult cancer are found on MRI[8].

Second look US followed by US-guided biopsy should be thefirst choice. However, despite the fact that it is a reliable, safe andaccurate method for identifying and characterizing incidental MRIfindings, it has been reported that 16–29% of these lesions will not

nal of Radiology 83 (2014) 942–950 943

bblihua

pUupdUtaplh

sleioa

uIr

2

Cbce

2

amdbwaww

(lclt1tesIs(e

was finished the soft gel capsules were removed and the skin markswere covered with a transparent dressing. When it was necessarythe soft gel capsules were placed over both breasts (Fig. 2).

E.P. Pons et al. / European Jour

e detected on conventional imaging (mammography and whole-reast US) [9]. Therefore MRI-guided biopsy is needed to sample

esions that are not seen on second look US. It is being used withncreasing frequency and it has been reported to be very accurate,owever it is not widely available, the sample is not performednder real time visualization, and it requires costly use of MR timend personnel.

To overcome these limitations, technological development hasrovided a third tool to sample incidental breast MRI lesions, RtMR-S, also known as breast US with volume navigation techniquesing CE-MRI as reference. This is a new technique in which areviously acquired CE-MRI volume is co-registered and displayeduring live US examination. It combines the advantages of MRI andS, is cheaper and can be easily performed within a reasonable

ime [10]. Moreover, real time visualization of lesions is possiblellowing real-time MRI navigated US guided biopsy of them. Thisrocedure has already been used in other anatomic areas such as

iver or prostate with excellent results and prior reports in breastave shown its accuracy and feasibility [2,3,10–14].

Axillary lymph node status is an extremely important progno-tic factor in the assessment of new breast cancer patients. Sentinelymph node biopsy (SLNB), that is considered the optimal strat-gy for axillary staging, can be avoided if lymph node metastasiss documented presurgically [15,16]. US-guided core needle biopsyf axillary lymph nodes in breast cancer patients can yield a highccuracy rate with no significant complications [17,18].

As far as we know this is one of the first reports referred on these of real-time MRI navigated US guided biopsy to characterize

LSM and LNSM. The aim of this study is to describe our preliminaryesults using this technique in clinical settings.

. Materials and methods

This is a descriptive study based on a prospective case series.ases were consecutively selected following our institutionalreast cancer management protocol, approved by the multidis-iplinary oncology committee on breast cancer and the hospitalthical committee.

.1. Participants

From May 2011 through January 2013, 577 women underwent prone CE-MRI with dedicated coil at our institution. The fourain indications for this were: (a) staging of a new breast cancer

iagnosis made by percutaneous US guided biopsy or stereotaxiciopsy (464 patients; 80.4%). (b) Control of breast cancer treatedith chemotherapy (54 patients; 9.4%). (c) Breast cancer screening

nd follow-up in high risk women (56 patients; 9.7%). (d) Womenith metastatic disease from unknown primary malignancy andith negative mammography and breast US (3 patients; 0.5%).

ILSM were detected on prone CE-MRI in 148 of 577 women25.7%). We considered ILSM on CE-MRI: (a) all BIRADS 4 or 5esions and (b) BIRADS 3 lesions when the multidisciplinary breastancer committee considered biopsy better than follow-up. Fol-owing our institutional multidisciplinary protocol on breast cancerhese women were firstly evaluated with second look US. In 21 of48 (14.2%) women this technique was negative for the evalua-ion of ILSM detected at CE-MRI, therefore they were prospectivelyvaluated using RtMR-US (Fig. 1) and core needle biopsy. Inclu-ion criteria for RtMR-US were: (a) patients with breast cancer and

LSM detected on staging or control breast MRI and (b) negativeecond look US for the detection of ILSM. Exclusion criteria were:a) patients with non-operable breast cancer, regardless of dis-ase stage; (b) patients with limitations for MRI procedure (such as

Fig. 1. Study flowchart shows included patients.

non-compatible MRI devices, physical limitations or claustropho-bia) and (c) positive second look US for the detection of ILSM.

These 21 patients were the participants in this study (meanage: 53.3 years; age range: 30–70 years) and informed consentwas obtained from all of them. Imaging characteristics of lesionsdetected on prone CE-MRI and RtMR-US, axillary lymph node statusand their histological correlation after ILSM biopsy were analyzed.

2.2. Supine breast MRI

Before navigation, CE-MRI was acquired in supine position, withupper extremities extended over the head. A cardiac MRI phasedarray coil was used covering both breasts. In order to avoid breastcompression three dedicated mattress were used, one placed inthe intermammary line and the others on each side of the thoraxso the coil was parallel to surface and compression was as lowest aspossible. Before MRI exam, three soft gel capsules of d-� tocopherol(Vitamin E, 400 UI; Chiesi S.A, Barcelona, Spain) were placed overthree lineal skin marks, without compression, and fixed with coatedsurgical tape. The three marks were drawn at 3, 9 and 12 o’clockradial to the nipple using an indelible skin marker. When CE-MRI

Fig. 2. Preparation of the patient before supine breast MRI. Vitamin E soft gel cap-sules (white arrows) used for US-MRI co-registration placed over lineal skin marks(black arrow) on the patient’ right breast.

944 E.P. Pons et al. / European Journal of Radiology 83 (2014) 942–950

F hite

e after

Gsepswfm1fv

2

RmsMafia

psm

ap6

ig. 3. US and RtMR-US system. (a) Electromagnetic sensors on the tip of the probe (wlectromagnetic transmitter, sensors and the navigation system. (d) RtMR-US exam

MRI exam was performed using 1.5 T MR scanner (HDX,E Healthcare, Milwaukee, WI, United States). Prenavigationupine CE-MRI protocol included a 3D T1 fast spoiled gradient-cho sequence with spectral fat suppression (LAVA) in axiallane (TR/TE = 4.3/2.0; flip angle = 12◦; field of view = 38 mm;pacing = 0; slice thickness = 1.5 mm; matrix = 350 × 350; band-idth = 50.0; acquisition time = 5′25′′) and one pre-contrast and

our post-contrast phases (80s between each phase), after auto-ated intravenous contrast administration of gadobutrol (Gadovist

mmol/ml; Bayer Pharma AG, Berlin, Germany) 0.1 mml/kg, 2 ml/s,ollowed by a flush of 20 ml of saline solution using an antecubitalein or dorsal metacarpal vein.

.3. US and real-time MRI navigated US

Initial breast US using conventional B-mode and posteriortMR-US were performed with a commercially available US equip-ent that contains a magnetic tracking system and a special

oftware for real-time volume navigation (LOGIC E9, GE Healthcare,ilwaukee, WI, United States). A conventional 6–15 MHz matrix

rray linear transducer with geometry allowing to observe a wideeld-of-view was used. Frame rate over 30 frames/s. Color Dopplernd elastosonography were also available during navigation exam.

Before RtMR-US, breast MR images were evaluated on a post-rocessing console, choosing the arterial vascular phase, which wasaved in a universal serial bus (USB) and uploaded in the US equip-ent.

RtMR-US exam was performed in two steps. The first one was to

ssemble the components of the device. The equipment was com-osed of: (a) two electromagnetic sensors attached to the tip of the–15 MHz transducer; (b) a portable electromagnetic transmitter

arrows). (b) Electromagnetic transmitter (black arrow). (c) Connection unit between co-registration.

that was positioned near the subject under examination. Both partswere connected to a third one, a position-sensing unit, installedin the US platform that allowed tracking probe position and ori-entation within the electromagnetic field. The second step was toco-register live US images to the previously uploaded MRI volumedata by coupling four reference points, the three skin markers andthe nipple. At the beginning both US and MRI images were dis-played side-by-side and activated alternately on the US equipmentscreen. So, during the co-registration process, live US or MR multi-planar reconstructed images were alternatively shown as the sametime the operator moved the transducer on the skin. Co-registrationwas done putting the middle of the transducer long axis on the skinmarker and selecting the same point in the active US image. Afterthat, it was frozen and only MR multiplanar reconstructed imageswere active to look for the hyperintensitiy of the vitamin E capsule.When it was found and selected the process was completed. Afterfour repetitions, one for each reference point, the co-registrationwas completed and both images tracked together, with the live USon the left screen and the MR on the right, as a road-map (Fig. 3).Then it was time to search at the live US the ILSM and LNSM iden-tified on the MR study.

2.4. Imaging findings

In the 21 participants all the lesions detected on prone andsupine CE-MRI and on RtMR-US were classified using BIRADSguideline [19]. Attending the shape, lesions were classified as nod-

ules and as enhancing areas at CE-MRI or echogenicity distortionareas at US. Axillary lymph nodes were also evaluated in thesewomen to rule out malignancy. Imaging criteria of axillary lymphnodes to be considered suspicious of malignancy were cortical

E.P. Pons et al. / European Journal of Radiology 83 (2014) 942–950 945

Table 1BI-RADS categorization and detected lesions on each imaging technique.

BI-RADS Prono CE-MRI US using conventional B-mode RtMRI-US

6 15 15 18ILSM 28 3 21

4–5 20 3 163 8 0 5

Not visualized 25 44–5 – 17 13 8 3

Total lesions 43 43 43

C BI-RADS 6 and 28 incidental lesions suspicious of malignancy (ILSM). Breast US usingc RI plus 3 more lesions, previously categorized as ILSM, on second look US), but did noti he 25 not detected on second look US.

th

2

aslaLgcliaf

2

uaucoawbSw

3

3

RCtrm(tUio(pit

Table 2Clinicopathological information of patients and characteristics of lesions on CE-MRI.

Clinicopathological information of patients N = 21

Age (mean, range) 54 (30–70)Indications for prone CE-MRI

BCa staging 13 (62%)BC follow-up 5 (23.8%)Unknown primary BC 3 (14.2%)

Breast sizeNormotrophic 16 (76.1%)Hipertrophic 3 (14.2%)Hipoplasia 2 (9.5%)

Lesion localizationRight breast 8 (38%)Bilateral 7 (33.5%)Left breast 6 (28.5%)Multicentric 3 (14.2%)Extensiveb 3 (14.2%)Multifocal 1 (5%)

Lesions on CE-MRI N = 43

Measures (mm): mean,median, range

18.1; 13; 4–110

Morphology 43Nodule 28 (65.1%)Enhancing area 13 (30.2%)

Histologic results of BI-RADS 6 15IDC 14 (93.3%)ILC 1 (6.6%)

the 21 patients. This represents 14% (21/148) of all the patientswho underwent a second look US after prone CE-MRI. Initial breastUS, previous to prone CE-MRI, identified 15 BI-RADS 6 lesions. Onthe second look US 3 of the 28 ILSM detected on prone CE-MRI

Table 3Correlation between BI-RADS categorization of ILSM detected on CE-MRI and histo-logic results as benign or malignant lesions.

Malignant Benign N

BI-RADS 4–5 9 8 17BI-RADS 3 0 8 8

ontrast enhanced resonance magnetic imaging (CE-MRI) detected 43 lesions: 15onventional B- mode detected 18 BI-RADS 6 (15 lesions on initial US before CE-Mdentify the others 25 ILSM. RtMRI-US detected 18 BI-RADS 6 and 21 ILSM among t

hickening (≥3 mm), round shape and diminished or absent fattyilium [20].

.5. RtMR-US guided biopsy

Core needle biopsy was necessary to take a sample of tissue fromll ILSM detected on RtMR-US and from all axillary lymph nodesuspicious of malignancy not detected on both initial and secondook US. Biopsy was performed using a single use spring-loadedutomatic biopsy device with double shoot mechanism (ACECUT,eleman S.L, Valencia, Spain), 14-gauge × 22 mm for ILSM and 16-auge × 11 mm for LNSM. After biopsy was performed a titaniumoil (Leleman S.L, Valencia, Spain) was inserted in the middle of theesion. Lesions were classified as benign or malignant. Those yield-ng high-risk, as lobulillar carcinoma in situ (LCIS), were classifieds benign. When biopsy could not be performed the lesions wereollowed up every 6 months by MRI.

.6. Data analysis

The mean, median, range, minimum and maximum values weresed to describe quantitative data. To describe qualitative data thebsolute frequency and percentages were calculated in the pop-lation of the study. To evaluate the diagnostic yield of RtMR-USategorical data were summarized in crosstabs for the calculationf sensitivity, specificity, positive predictive value (PPV) and neg-tive predictive value (NPV). Differences between categorical dataere calculated using �2 test or Fisher’ exact test. The concordance

etween different imaging techniques was evaluated using � index.tatistical significance was defined as p < 0.05. Statistical analysisas performed using IBM SPSS statistics 19.0.

. Results

.1. Imaging findings on CE-MRI

During the period of study 21 consecutive women underwent atMR-US, that corresponded to 3.6% (21/577) of those with proneE-MRI. Clinic-pathologic information of these patients and charac-eristics of lesions on prone CE-MRI are shown in Table 1. CE-MRIevealed a total of 43 lesions in these 21 women. Regarding theorphologic features 65% (28/43) of these were nodules and 35%

15/43) were enhancing areas (Figs. 4–6). BI-RADS categories ofhe detected lesions on each imaging technique (CE-MRI; RtMR-S and US using B mode, both initial and second look US) are listed

n Table 2. Among the 43 lesions 15 were BI-RADS 6 (14 patients,ne with initial multicentric cancer) and the others 28 were ILSM

8 lesions BI-RADS 3, 20 lesions BI-RADS 4–5). All ILSM detected onrone breast MRI were also found on supine breast MRI, although

n the last one the margins of lesions were less defined and some-imes with changes in their measures and location in the breast

a BC: breast cancer.b More than expected on initial breast US.

quadrant (Fig. 5). Table 3 shows accuracy of CE-MRI for BI-RADScategorization of ILSM attending to their likelihood of malignancyand correlation with histological results, with a sensibility of 100%,specificity of 50%, PPV of 52.9% and NPV of 100%.

3.2. Imaging findings on US using conventional B-mode

Second look US using conventional B-mode was performed in

N 9 16 25

CE-MRI accuracy of BI-RADS categorization of ILSM attending to their likelihoodof malignancy. Sensibility 100% (9/9), specificity 50% (8/16), PPV 52.9% (9/17), NPV100% (8/8).

946 E.P. Pons et al. / European Journal of Radiology 83 (2014) 942–950

F le on

t e samc

arvl3u

d1

3

d4

FbdU

ig. 4. RtMR-US, US and supine CE-MRI displayed side by side. (a) Poor defined noduhe RtMRI-US exam showing blue score of the nodule, no strain is measured. (c) Thorrelate a point in both MRI and US images.

nd categorized as BI-RADS 4–5 were visualized and biopsied withesult of IDC. Therefore, both initial and second look US using con-entional B mode, detected 11% of ILSM (3/28) and 42% of all theesions (18/43). The other 25 ILSM (18 BI-RADS 4–5 and 7 BI-RADS) remained non-detected after second look US and were studiedsing RtMR-US (Table 3).

Eight of 21 (38%) patients had LNSM on CE-MRI. Six of 8 wereetected on US, 5 during the initial axillary exam, before MRI, and

at second look US, and all of them were biopsied.

.3. Imaging findings on RtMR-US

Among the 43 lesions studied by RtMR-US, 39 (90.7%) wereetected; 100% (15/15) of BI-RADS 6, 94.1% (17/18) of BI-RADS–5 and 57% (4/7) of BI-RADS 3 (Table 2). RtMR-US sensibility was

ig. 5. Woman with known IDC in union of outer quadrants of the right breast. (a and biopsied area is pointed with the white arrow (a), two ILSM (arrows) as enhancing areas

) RtMRI-US, BIRADS 6 lesion in supine CE-MRI and real time US (c); correlation betweenS (white arrows) (d). After biopsy multicentric IDC was confirmed.

B-mode US images. (b) Elastosonography evaluation of the lesion performed duringe nodule at supine CE-MRI. A marker (cross 1) within the scan area (white square)

higher than US using B mode (p < 0.01). Agreement between bothtechniques was low � = 0.138 (95% CI 0.004–0.271).

Regarding the morphology of the lesions 82.1% (32/39) werenodules and 17.9% (7/39) were echogenicity distortion areas (in 3cases nodules were found within the echogenicity distortion areas).Lesions were more frequently seen as nodules on RtMR-US com-pared to CE-MRI, Fischer exact test p = 0.03. There were significantdifferences between both techniques (p = 0.018), agreement waslow, � = 0.34 (95% CI 0.04–0.65). Coinciding with CE-MRI findings,3 of 21 (14%) patients showed bilateral and multicentric lesions. Inother 3 among the 21 women, nodules or echogenicity distortionareas were more extensive than previously seen on initial imaging

mammography and US using B mode (Fig. 5 and Table 1).

In 4 of 43 cases (9.3%) RtMR-US did not correlate with initialbreast CE-MRI (false negatives): 3 benign and 1 malignant ILSM.3 were nodules under 10 mm and localized in the inferior-outer

) Staging prone CE-MRI, extension and enhancing BIRADS 6 lesion. Its previouslyin lower and inner quadrant of the same breast, categorized as BIRADS 5 (b). (c and

the two ILSM at supine CE-MRI and two echogenicity distortion areas at real time

E.P. Pons et al. / European Journal of Radiology 83 (2014) 942–950 947

F reast

U ion wc RI ima

qbnmRaT(h

i

3

Iag

o(nwmm

Ma

are listed in Table 3. Sensitivity of CE-MRI for detection of malignantlesions was 100% (9/9) and specificity 50% (8/16). 64% (16/25) ofILSM were benign, including 100% (7/7) of BI-RADS 3 and 50% (9/18)of BI-RADS 4–5. Detection rate was 68% (17/25).

Table 4Correlation between BIRADS categorization of ILSM biopsied by RtMRI-US and his-tologic results.

BI-RADS 4–5 BI-RADS 3 N

Malignant lesions 7 0IDC 6

7ILC 1Benign lesions 9 4

Fibrocystic disease 4 1

13Fibroadenoma 2 1Intramammary lymph node 1 2

ig. 6. 70 years old woman with metastasic lymph node and initially unknown bS. (a) 5 mm nodule on RtMR-US (white arrow), location and morphology correlatompression during the biopsy procedure the scan area shown as a road map on M

uadrant of hypertrophic breasts: one BI-RADS 4 nodule wasiopsied using stereotaxic guide, revealing an IDC; one BI-RADS 3odule was reclassified as BI-RADS 2 as it was detected on previousammography with no evidence of change for four years; one BI-

ADS 3 nodule detected concomitant with a multicentric IDC withnatomopathological result of fibroadenoma after mastectomy.he last false negative case was pathological enhancement areaBI-RADS 3) detected on the follow-up MRI of a LCIS. This lesionas not changed in control studies for 18 months.

Two LNSM located depth and adjacent to the chest wall weredentified only by RtMR-US (Fig. 7).

.4. RtMR-US core needle guided biopsy and histological results

RtMR-US core needle guided biopsy was performed in 20 of 21LSM (Figs. 6 and 7). Histological results and BI-RADS correlationre summarized in Table 4. 2 LNSM were also biopsied by RtMR-USuided procedure (Fig. 7).

Sixty-five percent (13/20) of lesions were benign; 100% (4/4)f BI-RADS 3 and 56.2% (9/16) of BI-RADS 4–5. Thirty-five percent7/20) were malignant lesions, all categorized as BI-RADS 4–5. Theon-biopsied ILSM was a BI-RADS 5 nodule detected in a patientith two BI-RADS 6 multicentric lesions. Due to mastectomy wasandatory the ILSM biopsy was considered unnecessary. After

astectomy an IDC was confirmed.In terms of indications for the CE-MRI: (a) 13 patients had CE-

RI for breast cancer staging: in 6 patients ILSM were malignantnd in 7 were benign (1 multicentric cancer, 1 multifocal cancer

cancer. An ILSM was revealed on prone CE-MRI but not detected on second lookith supine CE-MRI (arrow head). (b) RtMR-US guided biopsy of the nodule. Due toges is shifted.

and 1 cancer more extensive than previously expected) (Fig. 5).(b) Five patients underwent a CE-MRI for breast cancer follow-up:4 benign lesions and 1 bilateral and multicentric carcinoma. (c)Three patients with metastatic disease and unknown breast can-cer: 3 ILSM were detected on CE-MRI and correlated and biopsiedby RtMR-US. The histological results were 1 LCIS and 2 DCIS, bothunder 10 mm (Fig. 6).

Correlation between BI-RADS categorization of ILSM detectedon CE-MRI and histological results as benign or malignant lesions

Fibrosis 1 -Hemorraghic cyst 1

N 16 5 20

Sensibility 100% (7/7), specificity 30.7% (4/13), PPV 43.7%, NPV 100%.

948 E.P. Pons et al. / European Journal of Radiology 83 (2014) 942–950

F e betwa sis wa

in96

4

tioAOwubr

sisowwUNcUsm

Cibi(

TCa

D7

ig. 7. (a) RtMR-US shows correlation in the location of a deep axillary lymph nodrrowhead). (b) US guided biopsy of the lymph node (white arrow). Breast metasta

Correlation between all the lesions detected on CE-MRI anddentified on RtMR-US and histological results as benign or malig-ant lesions are shown in Table 5. Detection rate was 91% (39/43).% (4/43) were FN. Sensitivity was 96.3%, specificity 18.8%, PPV6.7% and NPV 75%.

. Discussion

To evaluate the feasibility, accuracy and reproducibility of realime US combined with supine-MRI in identifying and character-zing breast lesions and suspicious axillary lymph nodes detectedn CE-MRI, we analyzed 21 consecutives patients with 43 lesions.mong these lesions, 18 were BI-RADS 6 whereas 25 were ILSM.ur data showed that these 21 patients were only the 3.6% of allomen who underwent a CE-MRI, and only the 14.2% of those whonderwent a second look US (Fig. 1). The percentage of incidentalreast lesions identified on MRI and detected on second look USeported by other authors range between 30% and 71% [3,21–23].

The most common indication of CE-MRI among the women whoubsequently were studied using RtMR-US was breast cancer stag-ng, 62% (13/21 cases). The second indication was breast cancercreening and follow-up in high risk women, 24% (5/21 cases). Thether 14% (3/21 cases) corresponded to the 100% of those womenith metastatic disease from unknown primary malignancy andith negative mammography and breast US. In these cases RtMR-S allowed us to detect and characterize 3 breast cancers (Fig. 6).one of the women that underwent a CE-MRI for control of breastancer treated with chemotherapy needed evaluation with RtMR-S. We suggest that RtMR-US may have an important role in the

tudy of women with metastatic disease from unknown primaryalignancy (on mammography and US).On supine CE-MRI we identified all lesions detected on prone

E-MRI with dedicated coil. However, imaging quality was lower

n the first modality. We attribute that to the imaging protocol,reath and cardiac motion artifacts, the absence of a breast ded-

cated coil, and mainly to the non-expanded position of breastsFig. 5). On one hand some authors prefer to perform only supine

able 5orrelation between overall lesions detected on CE-MRI and identified on RtMRI-USnd histologic results as benign or malignant lesions.

Malignant Benign N

RtMR-US + 26 13 39RtMR-US − 1 3 4N 27 16 43

etection rate 90.7% (39/43); sensibility: 96.3%; specificity: 18.8%; PPV: 66.7% NPV:5%. False negatives 9% (4/43).

een live US (white arrow) and volume navigation technique using CE-MRI (whites confirmed.

CE-MRI before volume navigation. Nevertheless, it is known to beless accurate for breast cancer diagnosis than prone CE-MRI, it is notan established method and has to be more carefully evaluated [2,3].On the other hand other authors combine US with prone CE-MRI.However in these cases correlation is poor [13,22]. Prone positionin CE-MRI allow us to acquire higher quality images. Indeed it isthe standard MRI technique for breast cancer evaluation and hasproved to be more accurate than conventional US alone to definetumor extension. However due to the fact that RtMR-US is per-formed with the patient in supine position we think it could lead todisagreement in lesions location. Moreover our main objective withRtMR-US is to obtain tissue samples of breast lesions that other-wise would be impossible to achieve. Under our consideration evenalthough we obtained good correlation between prone CE-MRI andUS, it would be hardly difficult to perform a breast biopsy becauseof patient position, being even worst for an axillary lymph nodebiopsy. Thereby we think better to perform an additional supineCE-MRI prior to navigation when it is indicated. Besides, regardingbreast surgery, supine CE-MRI has been reported to have bettercorrelation with tumor extension [10,24–26].

The imaging characteristics of the 43 lesions were evaluatedboth on CE-MRI and RtMR-US. Morphologically, most of them wereidentified as nodules in both techniques. Nevertheless, nodulesdetection rate was higher on RtMR-US, 82% (instead of echogeni-city distortion areas), than on CE-MRI, 65% (instead of enhancingareas), with significance statistical difference (p < 0.005). Previousstudies have also published a higher detection rate of nodules onUS [22,27,28].

Regarding lesions size RtMR-US allowed us to determine tumorextension in 7 cases that were underestimated in the initial imagingstudy: 3 multicentric cancer, 3 extensive carcinomas and 1 multi-focal carcinoma. These findings modified the treatment of thesewomen (Figs. 4 and 5).

Second look US detected 85.8% (127/148) of ILSM. The additionaluse of RtMR-US allowed us to increase this rate to 97.2% (144/148),identifying ILSM in 17 of 21 patients with negative second look US.Similar detection rates of lesions on RtMR-US has been previouslyreported, ranging from 83% to 100% [2,3,14].

Comparing the lesion detection rates of RtMR-US versus USusing conventional B mode, the first one showed a higher number ofdetected lesions with significance statistical differences (p < 0.005)and a low Kappa index � = 0.138 (95% CI 0.004–0.27), Table 2. RtMR-US identified 91% (39/43) of all the lesions and 84% (21/25) of ILSM.US using conventional B-mode identified 42% (18/43) of all the

MRI detected lesions and 11% (3/28) of ILSM, similar rates to thatreported by Nakano et al. [3]. Using RtMR-US, 100% of BI-RADS 6lesions and 94% of BI-RADS 4–5 lesions were detected, whereasonly 62.5% (5/8) of BI-RADS 3 were identified (Table 2). Moreover

nal of

6a4ba

tvvlal

h(NngrpwtIsRr

ocspcafibwtfdo

cnbslvoRMFumtbUaRoasc

Upi

[

[

[

[

[

[

[

[

[

E.P. Pons et al. / European Jour

4% (16/25) of the total of ILSM were non-malignant lesions, andmong them 50% were classified as BI-RADS 3 and 50% as BI-RADS–5 on CE-MRI. All BI-RADS 3 lesions and 47% of BI-RADS 4–5 wereenign. However, 36% of ILSM detected on CE-MRI were malignantnd this is the reason why biopsy is mandatory.

Table 3 shows CE-MRI accuracy for the BI-RADS categorization ofhe 25 ILSM studied by RtMR-US. Sensitivity and positive predictivealues were both of 100%, but specificity and negative predictivealues were lower, 50% and 52.9% respectively. These results, simi-ar to those previously published by other authors, show the limitedccuracy of CE-MRI for the characterization of breast incidentalesions.

Core needle RtMR-US guided biopsy allowed us to achieve theistological diagnosis in 80% (20/25) of ILSM detected on CE-MRITable 4). Our results are similar to that previously published byakano et al. [3]. The non-biopsied lesions were 4 false negativesot detected on RtMR-US. However, none of them needed MRuided biopsy due to final diagnosis was achieved by mammog-aphy or follow-up. The last non-biopsied lesion was detected in aatient with multicentric carcinoma, therefore core needle biopsyas considered unnecessary. We found that in our series of cases

otal accuracy of RtMR-US for BI-RADS categorization of biopsiedLSM attending to their likelihood of malignancy was 55%, withensitivity and negative predictive value both of 100%. In contrasttMR-US specificity and positive predictive value (30.7% and 43.7%espectively) was low for BI-RADS categorization.

Additionally CE-MRI may show LNSM not previously identifiedn initial US, and also non- detectable on second look US. In theseases US core needle guided biopsy cannot be performed. In oureries 8 of 21 patients (38%) had LNSM. Core needle biopsy waserformed in all of them. In 6 cases (75%) it was positive for breastancer. Using RtMR-US was possible to perform the biopsy in twodditional patients with deep lymph nodes not previously identi-ed on US using B-mode (Fig. 7). Metastatic disease was detectedy sentinel lymph node biopsy only in one of the other 15 patientsith negative axillary evaluation. Thereby, beyond the identifica-

ion of the primary breast tumor, RtMR-US has an additional valueor the detection of suspicious axillary lymph nodes otherwise non-etectable using other imaging techniques, improving the selectionf patients for sentinel lymph node biopsy.

MRI guided biopsy is being used with increasing frequency forharacterization of lesions, but it is expensive, time-consuming andot widely available, so second look US with core needle guidediopsy should be the first option However correlation betweenecond look US and CE-MRI is variable (23–89%) [14,27–30]. Radio-ogists must perform second look US based only on mentallyisualized MRI, there is no real time correlation and no evidencef whether the lesion has been accurately detected and biopsied.tMR-US is an additional technique that allows the evaluation ofRI detected lesions that are not identified on second look US.

urthermore RtMR-US provides a new safe method for the eval-ation and guided biopsy of axillary lymph nodes suspicious ofalignancy. Compared with MRI guided biopsy, this navigation

echnique is cheaper, more comfortable for the patient and cane performed within a reasonable time with good results. RtMR-S increases the diagnostic value of second look US using B modend can be used to perform a guided biopsy as well. TherebytMR-US is useful not only to characterize breast lesions detectedn CE-MRI and not seen on second look US, but also to identifynd select the patients that truly need MRI guided biopsy andentinel lymph node biopsy, reducing unnecessary surgical pro-edures.

Attending to our institutional protocol a limitation for RtMR-S is that an additional supine MRI must be performed after therone MRI and before the navigation technique. The main lim-

tation of this study is the small size of the patient population.

[

Radiology 83 (2014) 942–950 949

Therefore, further studies with a large patient database arerequired. In addition, multicentric studies are needed to evaluatethe variability of these results among different institutions.

In conclusion, RtMR-US combines the advantages of US and MRIand improves the detection rate of ILSM, avoiding unnecessaryMRI-guided biopsies and also unnecessary SLNB. This new tech-nique may be useful in centers which do not have the capabilityof performing a MR core biopsy. In dedicated breast centers, sincethe patient is undergoing a second MR with contrast, biopsy underMR guidance has become a common practice although it is moreexpensive and requires more time and personnel. RtMR-US is anaccurate method to establish tumor extension and may be used toidentify occult breast carcinomas. Our results suggest that most ofILSM not detected on US using B mode and categorized as BI-RADS3 on MRI are probably benign.

Conflict of interest

None declared.

References

[1] Mann RM, Kuhl CK, Kinkel K, et al. Breast MRI: guidelines from the EuropeanSociety of Breast Imaging. Eur Radiol 2008;18:1307–18.

[2] Nakano S, Yoshida M, Fujii K, et al. Fusion of MRI and sonography image forbreast cancer evaluation using real-time virtual sonography with magneticnavigation: first experience. Jpn J Clin Oncol 2009;39:552–9.

[3] Nakano S, Kousaka J, Fujii K, et al. Impact of real-time virtual sonography, acoordinated sonography and MRI system that uses an image fusion technique,on the sonographic evaluation of MRI-detected lesions of the breast in second-look sonography. Breast Cancer Res Treat 2012;134:1179–88.

[4] Sardanelli F, Podo F, D’Agnolo G, et al. Multicenter comparative multimodalitysurveillance of women at genetic-familial high risk for breast cancer (HIBCRITstudy): interim results. Radiology 2007;242:698–715.

[5] Lee SG, Orel SG, Woo IJ, et al. MR imaging screening of the contralateral breastin patients with newly diagnosed breast cancer: preliminary results. Radiology2003;226:773–8.

[6] Liberman L, Morris EA, Dershaw DD, et al. MR imaging of the ipsilateralbreast in women with percutaneously proven breast cancer. Am J Roentgenol2003;180:901–10.

[7] Houssami N, Ciatto S, Macaskill P, et al. Accuracy and surgical impact ofmagnetic resonance imaging in breast cancer staging: systematic review andmeta-analysis in detection of multifocal and multicentric cancer. J Clin Oncol2008;26:3248–58.

[8] Harms SE, Flamig DP, Hesley KL, et al. MR imaging of the breast with rotat-ing delivery of excitation off resonance: clinical experience with pathologiccorrelation. Radiology 1993;187:493–501.

[9] Teifke A, Lehr HA, Vomweg TW, et al. Outcome analysis and rational manage-ment of enhancing lesions incidentally detected on contrast-enhanced MRI ofthe breast. Am J Roentgenol 2003;181:655–62.

10] Fausto A, Rizzatto G, Preziosa A, et al. A new method to combine contrast-enhanced magnetic resonance imaging during live ultrasound of the breastusing volume navigation technique: a study for evaluating feasibility, accuracyand reproducibility in healthy volunteers. Eur J Radiol 2012;81:e332–7.

11] Kunishi Y, Numata K, Morimoto M, et al. Efficacy of fusion imaging combiningsonography and hepatobiliary phase MRI with Gd-EOB-DTPA to detect smallhepatocellular carcinoma. Am J Roentgenol 2012;198:106–14.

12] Rud E, Baco E, Eggesbo HB. MRI and ultrasound-guided prostate biopsy usingsoft image fusion. Anticancer Res 2012;32(8):3383–9.

13] Chang JM, Han W, Moon HG, et al. Evaluation of tumor extent in breast cancerpatients using real-time MR navigated ultrasound: preliminary study. Eur JRadiol 2012;81:3208–15.

14] Nakano S, Yoshida M, Fujii K, et al. Real-time virtual sonography, a coordinatedsonography and MRI system that uses magnetic navigation, improves the sono-graphic identification of enhancing lesions on breast MRI. Ultrasound Med Biol2012;38:42–9.

15] Rahbar H, Partridge SC, Javid SH, et al. Imaging axillary lymph nodes in patientswith newly diagnosed breast cancer. Curr Probl Diagn Radiol 2012;41:149–58.

16] Shigekawa T, Sugitani I, Takeuchi H, et al. Axillary ultrasound examination isuseful for selecting patients optimally suited for sentinel lymph node biopsyafter primary systemic chemotherapy. Am J Surg 2012;204:487–93.

17] Piniero A, Gimenez J, Vidal-Sicart S, et al. Selective sentinel lymph node biopsyand primary systemic therapy in breast cancer. Tumori 2010;96:17–23.

18] Pinero A, Gimenez J, Merck B, et al. Consensus meeting on sentinel lymph node

biopsy in breast cancer, Spanish society of mastology and breast disease. CirEsp 2007;82:146–9.

19] American College of Radiology (ACR). BI-RADS®-4th edition. In: ACR breastimaging reporting and data system, breast imaging Atlas. Reston, VA: AmericanCollege of Radiology; 2003.

9 nal of

[

[

[

[

[

[

[

[

[

50 E.P. Pons et al. / European Jour

20] Huo L, Yan K, Zhang H, et al. Morphology-based criteria for ultrasonic assess-ment of axillary lymph node status in primary breast cancer. Zhonghua Yi XueZa Zhi 2012;92:672–4.

21] Abe H, Schmidt RA, Shah RN, et al. MR-directed (second-look) ultrasoundexamination for breast lesions detected initially on MRI: MR and sonographicfindings. Am J Roentgenol 2010;194:370–7.

22] Carbognin G, Girardi V, Brandalise A, et al. MR-guided vacuum-assisted breast biopsy in the management of incidental enhancinglesions detected by breast MR imaging. Radiol Med 2011;116:876–85.

23] Kim TH, Kang DK, Jung YS, et al. Contralateral enhancing lesions on magneticresonance imaging in patients with breast cancer: role of second-look sonog-

raphy and imaging findings of synchronous contralateral cancer. J UltrasoundMed 2012;31:903–13.

24] Alderliesten T, Loo C, Paape A, et al. On the feasibility of MRI-guided navi-gation to demarcate breast cancer for breast-conserving surgery. Med Phys2010;37:2617–26.

[

[

Radiology 83 (2014) 942–950

25] Nakamura R, Nagashima T, Sakakibara M, et al. Breast-conserving surgery usingsupine magnetic resonance imaging in breast cancer patients receiving neoad-juvant chemotherapy. Breast 2008;17:245–51.

26] Sakakibara M, Nagashima T, Sangai T, et al. Breast-conserving surgeryusing projection and reproduction techniques of surgical-position breast MRIin patients with ductal carcinoma in situ of the breast. J Am Coll Surg2008;207:62–8.

27] Demartini WB, Eby PR, Peacock S, et al. Utility of targeted sonography for breastlesions that were suspicious on MRI. Am J Roentgenol 2009;192:1128–34.

28] Meissnitzer M, Dershaw DD, Lee CH, et al. Targeted ultrasound of the breast inwomen with abnormal MRI findings for whom biopsy has been recommended.Am J Roentgenol 2009;193:1025–9.

29] Peters NH, Borel Rinkes IH, Zuithoff NP, et al. Meta-analysis of MR imaging inthe diagnosis of breast lesions. Radiology 2008;246:116–24.

30] Carbognin G, Girardi V, Calciolari C, et al. Utility of second-look ultrasound in themanagement of incidental enhancing lesions detected by breast MR imaging.Radiol Med 2010;115:1234–45.