Breast cancer staging: Combined digital breast tomosynthesis and automated breast ultrasound versus magnetic resonance imaging

Accepted Manuscript Title: Breast cancer staging: combined Digital breast tomosynthesis and Automated breast ultrasound versus Magnetic resonance imaging Authors: Rossano Girometti, Ludmila Tomkova, Lorenzo Cereser, Chiara Zuiani PII: DOI: Reference:

S0720-048X(18)30306-1 https://doi.org/10.1016/j.ejrad.2018.09.002 EURR 8295

To appear in:

European Journal of Radiology

Received date: Revised date: Accepted date:

30-4-2018 5-7-2018 3-9-2018

Please cite this article as: Girometti R, Tomkova L, Cereser L, Zuiani C, Breast cancer staging: combined Digital breast tomosynthesis and Automated breast ultrasound versus Magnetic resonance imaging, European Journal of Radiology (2018), https://doi.org/10.1016/j.ejrad.2018.09.002 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Concise title: Breast cancer staging: combined Digital breast tomosynthesis and Automated breast ultrasound versus Magnetic resonance imaging. Informative title: Breast cancer staging: combined Digital breast tomosynthesis and Automated breast ultrasound versus Magnetic resonance imaging. Rossano Girometti ([email protected])

Lorenzo Cereser ([email protected]) Chiara Zuiani ([email protected])

IP T

Ludmila Tomkova ([email protected])

SC R

All the Authors have the following affiliation: Institute of Radiology, Department of

Medicine, University of Udine – University Hospital „S. Maria della Misericordia“, p.le S. Maria della Misericordia n. 15 – 33100 Udine (UD), Italy, tel. +39 0432559231, fax +39

U

0432559867

N

Corresponding Author: Ludmila Tomkova ([email protected])

A

Declaration of interests: none.

M

ABSTRACT

Purpose: To investigate whether combined Digital breast tomosynthesis and Automated

staging breast cancer.

ED

breast volume scanner (DBT-ABVS) are comparable to Magnetic resonance imaging (MRI) in

Methods: We retrospectively included seventy-three patients with histologically proven breast

PT

cancer who underwent preoperative DBT, ABVS and 1.5T MRI in the period July 2015–July 2016. Two radiologists in consensus recorded the number, site and Breast imaging-reporting

CC E

and data system (BI-RADS) category of breast findings during two independent reading strategies, i.e. DBT-ABVS vs. MRI. Using histology or 1-year follow up as the standard of reference, we calculated the accuracy for cancer of both imaging strategies. Bland-Altman analysis was used to evaluate the agreement between MRI vs. DBT or ABVS in cancer size

A

assessment.

Results: Patients showed a total of 160 lesions (108 malignant and 52 benign). Malignant lesions were unifocal, multifocal, multicentric and biltateral in 53, 15, 4 and 1 cases, respectively. Diagnostic accuracy of DBT-ABVS vs. MRI was comparable for all cancers (90.0% [95%C.I. 84.3-94.2] vs. 93.8% [95%C.I. 88.8-97.0], respectively). DBT-ABVS showed lower sensitivity and positive predictive values for additional disease (76.5% [95%C.I. 58.889.3] vs. 91.7% [95%C.I. 84.6-96.1], and 78.8% [95%C.I. 61.0-91.0] vs 93.4% [95%C.I. 86.91

97.3], respectively). Compared to MRI, ABVS+DBT missed 6 lesions, including two invasive cancers and one extensive intravascular invasion associated to ductal carcinoma in situ. Bland-Altman analysis showed ABVS to agree with MRI at a higher extent than DBT in assessing cancer size. Conclusions: Though less performing than MRI, DBT-ABVS showed acceptable diagnostic accuracy in staging breast cancer. This strategy might be used if MRI is unavailable or

IP T

unfeasible. Keywords: Breast cancer; Digital breast tomosynthesis; Automated breast volume scanner;

A

CC E

PT

ED

M

A

N

U

SC R

Magnetic resonance imaging

2

INTRODUCTION Magnetic resonance imaging (MRI) is the most sensitive tool to achieve local staging of breast cancer (BC), with a significant impact on the type and extent of surgical treatment [1, 2, 3]. However, the use of MRI in the preoperative setting is still under debate, given limited availability, costs, and controversies on the low specificity reported in previous works [2-4]. Additional concerns are related to contrast medium safety issues, including the risk for adverse reactions (hypersensitivity/allergy-like reactions or nephrogenic systemic fibrosis), as well as

IP T

ever-increasing caution related to gadolinium retention in some clinical scenarios [5]. Strict indications to staging MRI are limited to invasive lobular carcinoma (ILC), patients at high risk

for breast cancer, patients <60 years of age with a discrepancy >1 cm in size between digital

SC R

mammography (DM) and ultrasound (US) (with expected impact on treatment decision), and patients eligible for partial breast irradiation [1].

In this light, local staging of BC is frequently achieved using US coupled with DM and/or digital

U

breast tomosynthesis (DBT). In a recent study, combined evaluation with DM, DBT and US was found to approach the sensitivity of MRI [6]. However, DBT is promising as a stand-alone

N

complement to US, providing more accurate assessment of BC size and morphology than DM

A

[7, 8], as well as reliable tumor volume assessment [9]. This is in line with the expected capability of DBT to overcome tissue superimposition effects and achieve better lesions

M

conspicuity [10].

ED

Automated breast volume scanner (ABVS) has been introduced as a novel ultrasound-based modality to provide standardized scans of the breasts and a high-resolution volumetric set of images readable at any time from the acquisition [11]. ABVS is an intensive matter for

PT

research, given potential advantages in terms of reduced operator-dependency, use of reformatted images to achieve better lesion definition, and maximized cost-effectiveness of the diagnostic process [12]. The role for ABVS in staging BC has been poorly assessed, with a

CC E

few works showing that this technique outperforms US in assessing BC size (with comparable assessment of cancer features) [13] or guiding breast conservative surgery in patients with ductal carcinoma in situ (DCIS) [14].

A

As far as we know, no previous studies investigated whether combining DBT and ABVS (DBTABVS) is an accurate strategy to stage BC, as one might expect from the potential advantages of those techniques over DM and US, respectively. We hypothesized that DBT-ABVS might represent a reliable tool to stage BC in clinical practice if approximating MRI in identifying cancer foci and assessing lesions size. If yes, DBT-ABVS might be used in those scenarios in which MRI is unavailable or contraindicated (e.g., because of an increased risk of adverse reactions to contrast medium administration). 3

The purpose of this study was twofold: (1) to compare the accuracy in detecting breast cancer of two different staging strategies, namely DBT-ABVS vs. MRI; (2) to compare DBT-ABVS vs. MRI in the assessment of BC size. MATERIALS AND METHODS Study population and standards of reference Our Institutional Review Board approved the study protocol. Informed consent acquisition was

IP T

waived given the retrospective design.

By performing a search in our institutional database, we identified all patients operated for

breast cancer who underwent the following cancer workup examinations in the period July

SC R

2015-July 2016: DBT, ABVS, and staging MRI. In accordance to the policy of our tertiary referral center, DBT was performed before biopsy as a diagnostic complement to DM, as supported by literature [7-10].

Since this study is a retrospective branch of an ethics

U

committee-approved larger prospective trial investigating the role for ABVS in the clinical setting, ABVS was performed in all women in the pre-biopsy phase. Examinations were

After

identifying

one

hundred-six

subjects,

we

excluded

cases

with

A

surgery.

N

performed bilaterally, within an overall period of 1 month from the first exam and breast

unreliable/unavailable correlation with postoperative histology (8 patients referred to

M

neoadjuvant chemotherapy and 1 patient operated elsewhere), monolateral examinations (13 patients with monolateral DBT, and 7 patients with monolateral ABVS), and technically limited

ED

examinations (4 patients with low-quality ABVS). Final population included 73 women (mean age: 55.9 years, SD: ± 10.6 years; range 34-81 years).

PT

The standard of reference for malignant lesions was histological examination after mastectomy (40 monolateral and 1 bilateral procedures) or breast-conservative surgery (32 procedures). Findings assessed as benign were confirmed with histological examination or pathological

CC E

examination in 29 cases, and with a mammography and/or sonography-based follow-up of at least one year in 23 lesions (range 12-18 months, median time 15 months). Histological or pathological analysis was performed by one of two breast pathologists (25 and 5 years of

A

experience, respectively). DBT and MRI protocols DBT was performed with one of two different systems, namely AW5000 3D (Hologic Selenia Dimensions; Hologic, Bedford, MA, USA) (system 1) or Giotto TOMO (Internazionale Medico Scientifica IMS, Bologna, Italy) (system 2), using the following acquisition parameters: 1) for system 1, low-dose protocol, 15 projections on a ±15° angular range, stationary a-Se digital detector, sensitive area 24x29 cm, filtered back projection / iterative contrast reconstructions 4

2) for system 2, low-dose protocol, 13 projections on a ±20° angular range, with stationary aSe digital detector, sensitive area 24x30 cm, iterative reconstruction algorithm. Regardless of the equipment, patients underwent DBT in both cranio-caudal and medio-lateral-oblique views. No additional views were required. MRI examinations were carried on one of two different 1.5T systems, i.e. (1) Magnetom Avanto, or (2) Magnetom Aera (both Siemens Medical Solutions, Erlangen, Germany), using a bilateral 4-channel or 16-channel breast coil, respectively. Study sequences and acquisition

IP T

parameters are reported in Tab. 1. The diffusion-weighted imaging (DWI) sequence implemented fat saturation with the SPAIR (spectral selection attenuated inversion recovery) technique and parallel imaging (sensitivity encoding algorithm with an acceleration factor of 2).

SC R

The dynamic contrast enhancement study was performed after T2-weighted imaging and DWI using one pre-contrast acquisition followed by five consecutive scans after intravenous administration of 0.1 mmol/kg of gadobenate dimeglumine (MultiHance, Bracco Imaging, Milan, Italy) injected at a 2 ml/sec rate by a remote control system (Optistar Elite, Mallinckrodt,

U

United Kingdom).

N

ABVS protocol

A

The ABVS equipment was the Acuson S2000 ABVS US system (Siemens Medical Solutions,

M

Inc., Pleasonton, CA, USA), mounting a 5–14MHz linear transducer (Siemens 14L5BV) on a flexible mechanical arm. Reference frequencies were set in accordance with cup sizes, with

ED

cup A = 11MHz, cup B = 10MHz, cup C = 9 MHz, cup D = 9 MHz and cup D+ = 9 MHz. During the examination, the patient laid in the supine position with the side under examination placed overhead. The transducer was covered with a replaceable membrane to achieve

PT

adequate contact with the skin, and then applied to each breast with a gentle compression. Each breast was scanned with anterior-posterior, lateral and medial views,. Additional views

CC E

were acquired if better anatomical coverage of larger breasts was needed. For the purpose of correct orientation and image reformation, the nipple was used as a reference point in each scan view. Scan parameters for each view were as follows: field of view 15.4x16.8 cm, scan depth up to 6 cm, maximum tissue volume 1.552.3 cm3, image thickness 0.5 mm, transducer

A

speed 16.8 cm/min, reconstructed slice thickness 0.5-1 mm, acquisition time 60 sec. Scans were not aborted after having achieved full anatomic coverage. The total scan time for a 3views bilateral breast examination was 6 min. On average, the time for a bilateral examination with no additional views was 15 min (3-views bilateral scan time plus patient’s preparation/positioning time and probe positioning time).

5

Image analysis A study coordinator organized reading sessions involving two radiologists in consensus (3 and 10 years of experience in breast imaging, respectively). Reading sessions were set to show independently (1) DBT-ABVS images only vs. (2) MRI images only, and separated by a “washout” period of one month to avoid recall bias. Although readers were aware that patients had cancer, they were blinded to final diagnosis of breast findings (benign vs. malignant), as well as to the type and site of the index lesion (defined as the lesion prompting the staging process

IP T

in the original clinical setting). Readings were performed on the following workstations: (1)

Suite Estensa (Ebit AET, Esaote, Genova, Italy) for DBT and MRI images; (2) Syngo

Ultrasound Breast AnalysisTM (sUSBA) (Siemens Healthcare, Erlangen, Germany) for ABVS

SC R

images. During DBT-ABVS staging, both DBT and ABVS examinations were available at the same time, with no definite order of analysis. DBT images included source thin slices of the

breast, as well as 2D synthetic reconstructed images. ABVS images were displayed on the

U

native acquisition plane, as well as reformatted transverse, coronal and sagittal planes. For each imaging technique, radiologists identified breast findings, recording their site (side

N

and quadrant), size (mm), and Breast imaging reporting and data system (BI-RADS) category

A

[15]. For the purpose of analysis, final BI-RADS category in the DBT-ABVS strategy was the one with the most suspicious score between the two techniques. Multifocality was defined as

M

two or more separate malignant lesions in the same quadrant of the breast, whereas multicentricity was defined as two or more separate malignancies occupying more than one

ED

quadrant of the same breast. Bilateral tumor was defined as two or more malignant lesions in both breasts. Size was measured in the slice showing largest lesions diameter, including reformatted images in the case of ABVS and MRI. To reflect clinical practice, radiologists were

PT

free to access images and reports of ultrasonography second-look examinations to solve doubtful MRI cases.

CC E

Data analysis

After readings completion, findings of both staging strategies were dichotomized as benign (BI-RADS 1-3) vs. malignant (BI-RADS 4-5) and matched with the standard of reference to

A

calculate sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy, together with 95% C.I. We used the McNemar test to evaluate whether the two staging strategies differed at a significant extent in the attribution of benign vs. malignant BI-RADS categories. Analysis was performed on a per-lesions basis, and stratified for all lesions and additional lesions (i.e., lesions found in addition compared to the index lesions that prompted the staging process).

6

After having verified non-normal distribution of the data with the Shapiro-Wilk test, we compared lesions’ size provided by DBT, ABVS and MRI with the Kruskall-Wallis test. Pairwise comparisons were performed with the Wilcoxon test. The agreement between DBT and ABVS vs. MRI in determining lesion’s size was assessed with Bland-Altman analysis, which provided pairwise average difference between the methods and 95% limits of agreement (LOA). LOA represent the range of agreement within which 95% of the differences between one measurement and the other lie [16]. Given non-normal distribution of the differences in size

IP T

[16], analysis was applied on logarithmically-transformed data [16] for all pairwise comparisons. Size analysis was targeted to lesions proven to be malignant on histology.

Alfa level was set 0.05. In size assessment, we used the Bonferroni correction to account for

SC R

multiple comparisons (referring p = 0.05/3 = 0.016). Analysis was run with a commercially available tool (MedCalc version 14.8.1, MariaKerke, Belgium). RESULTS

U

Study population

N

Seventy-three included patients showed a total of 108 malignant lesions on histological

A

analysis (per-patient mean 1.45), including 74/108 index lesions (68.5%) and 34/108 non index lesions (31.5%). Lesions were unifocal, multifocal, multicentric, and bilateral in 53/73 (72.6%),

M

15/73 (20.5%), 4/73 (5.5%), and 1/73 (1.4%) patients, respectively. Lesions type included 55/108 invasive ductal carcinomas (IDC) (50.9%), 19/108 IDCs with ductal carcinoma in situ

ED

(DCIS) (17.6%), 1/108 IDC with micropapillary intracystic carcinoma (0.9%), 19/108 pure DCIS (17.6%), 1/108 DCIS with IDC (0.9%), 7/108 invasive lobular carcinomas (ILC) (6.5%) and

PT

6/108 invasive ductulo-lobular carcinomas (5.6%). Overall, we observed a total of 52 benign lesions. Findings were visible on both DBT-ABVS in 40/52 cases, on DBT-ABVS only in 7/52 cases, and on MRI only in 5/52 cases, respectively.

CC E

Twenty-three of 52 findings (44.2%) showed no changes on an average follow-period of 15 months (range 12-18), and were predominantly assessed as fibroadenomas. Of the remaining 29/52 findings, benignity was confirmed by pathological examination in 12/29 cases included in the surgical specimen, and by US-guided or DBT-guided biopsy in the remaining 17/29

A

cases. Findings included fibroadenoma in 3/52 cases (5.8%), fibrosis in 1/52 case (1.9%), cyst in 2/52 cases (3.9%), florid adenosis in 2/52 cases (3.9%), fibrous-cystic changes in 7/52 (13.5%), papilloma 1/52 case (1.9%), scleroelastotic lesion 1/52 case (1.9%), atypical ductal hyperplasia 1/52 case (1.9%), and 1 benign lymph node in 1/52 cases (1.9%). The remaining 10/52 cases (19.2%) included normal breast tissue (with pathological confirmation of fibroglandular tissue in 9/10 cases showing non-mass enhancement on MRI, and biopsy with

7

1-year follow-up confirmation in 1/10 cases showing non mass enhancement on MRI and architectural distortion on DBT). Accuracy in detecting cancer DBT-ABVS identified a total of 147 findings, classifying them as BI-RADS 2-3 in 41/147 cases (27.9%; 95%C.I. 2,3 - 2,6) and BI-RADS 4-5 in 106/147 cases (72.1%; 4,5 - 4,7 95%C.I.), respectively. MRI found a total of 151 findings, classifying them as BI-RADS 2-3 in 41/151 cases (27.1%; 95%C.I. 2,3 - 2,6) and BI-RADS 4-5 in 110/151 cases (72.9%; 95%C.I.

IP T

4.5 – 4.7), respectively. There was no significant difference in BI-RADS assignments of lesions identified by both staging strategies (n = 141) (p>0.05).

SC R

Diagnostic accuracy for breast cancer of both staging strategies is shown in Tab. 2. DBT-ABVS and MRI missed 9 and 4 lesions, respectively, as detailed in Tab. 3. Exemplificative cases from our series are illustrated in Fig. 1-3.

U

Assessment of lesions size

N

There was no significant difference (p>0.016) in the size of malignant lesions as assessed by DBT (mean 19.8 mm, median 16 mm, range 5-92 mm), ABVS (mean 17.8 mm, median 15

A

mm, range 4-56 mm), and MRI (mean 18.5 mm, median 16 mm, range 4-87 mm).

M

Results of Bland-Altman analysis are shown in Fig. 4. Given logarithmic transformation (see the Materials and Methods), LOA were calculated as dimensionless antilogarithms of the limits

ED

around the mean (mean – 2SD and mean + 2SD) shown in Fig. 4. E.g., LOA for DBT vs. MRI resulted 0.45 to 2.24, meaning that for 95% of cases, the size provided by DBT was 0.45 to 2.24 times the MRI size (which in turn means from 55% below to 124% above) [16]. LOA for

CC E

DISCUSSION

PT

ABVS vs. MRI resulted 0.62 to 1.38 (variation from 38% below to 38% above the MRI size).

Previous studies showed that adding DBT [6] or ABVS [17] to DM and US improves the preoperative assessment of BC, with 97.7-97.1% sensitivity and 82.8-95.2% specificity for cancer detection, respectively. To our knowledge, our study is the first to report that the direct

A

combination of DBT and ABVS achieves comparable sensitivity (91.7%) and specificity (86.5%) in the same clinical setting. Our results are somewhat expected on the base of previous results showing DBT to outperform DM in cancer detection (sensitivity/specificity 93100%/64-75% vs. 72.4-88.0%/46.4-60.0%, respectively) [18, 19] or inter-reader agreement [20]. Moreover, ABVS was found comparable to US in terms of sensitivity (77.8-100% vs. 62.5100%, respectively), specificity (80.5-97.8% vs. 82.5-96.7%, respectively) [21-23], and reproducibility of sonographic features [13]. Of note, DBT-ABVS showed quite larger specificity

8

(86.5%) than the one provided in a reference study using combined analogic mammography and US (23%) [24], thus approaching MRI (88.5%). Though indirectly, our findings support the hypothesis that DBT-ABVS can replace DM and US, with the potential of combining at best the advantages of the new-generation techniques, including the avoidance of tissue superimposition and availability of a 3D set of sonographic images for image review and communication with the referring clinician [10, 12, 22, 25-27]. Further studies should investigate whether DBT-ABVS can maximize cost-effectiveness, speed and reproducibility of

IP T

preoperative cancer assessment. As expected, MRI was the most reliable tool to detect BC in our series (96.3% sensitivity,

88.5% specificity, 94.5% PPV, and 92.0% NPV, respectively) [6, 28]. Although DBT-ABVS

SC R

approached the overall accuracy of MRI (90.0% vs 93.8%, respectively), stratified analysis on

35/108 non-index lesions showed lower sensitivity (76.5% vs. 91.2%, respectively) and PPV (78.8% vs. 83.8%, respectively). While detecting most multicentric (3/4) or bilateral (1/1) lesions, DBT-ABVS missed two invasive cancers (one 6 mm index IDC with DCIS, and one 8

U

mm multifocal IDC, respectively), as well as 1 multicentric and 3 multifocal DCISs up to 54

N

mm. Despite the retrospective design of the study makes difficult to assess the impact on patients’ management, one might reasonably assume that invasive and/or larger lesions

A

missed by DBT-ABVS might have impaired the choice of the most appropriate operative plan.

M

However, the sensitivity for additional disease (76.5%) we observed for DBT-ABVS was notably larger than the one previously reported using analogic mammography (37% and 18%

ED

for multifocal and multicentric disease, respectively) and US (41% and 9% for multifocal an multicentric disease, respectively) in a direct comparison with MRI [29]. Our findings suggest that the performance of DBT-ABVS is suboptimal, but accurate enough to provide reliable

PT

preoperative assessment when MRI is unavailable or unfeasible (e.g., because of increase risk of adverse reactions to contrast medium), with lesser risk of missing significant additional

CC E

disease compared to conventional imaging with DM and US. This hypothesis is in accordance with previous results observed by Krammer et al. [30], who found increased detection of multifocal and multicentric disease in dense breasts using DBT (63.2%) compared to DM (21.1%). Similarly, Huang et al. [14] showed that cancer detection rate of ABVS (82.1%) was

A

significantly higher than US in patients with DCIS referred to breast conservative surgery. Reliable assessment of cancer size is of pivotal importance in defining the T stage, and in turn determining treatment choice and prognosis [13]. We were not able to match imaging with histological cancer size on a lesion-by-lesion basis, because not all cancers were visible on both DBT and ABVS. Since MRI is the most accurate imaging tool for measuring breast cancer [8, 24, 31, 32], we used MRI size as a surrogate standard of reference, having in mind the risk of overestimation due to large DCIS components, atypia or fibrosis [31]. ABVS agreed with 9

MRI at a greater extent than DBT, with an expected discrepancy compared to the MRI size from 38% below to 38% above vs. 55% below to 124% above, respectively (e.g., it is easy to calculate that a cancer showing 20 mm size on MRI might be expected from 20-7.6=12.4 mm to 20+7.6=27.6 mm on ABVS, given 7.6 mm being 38% of 20 mm). Our results are in line with those by Schmachtenberg et al. [33], who observed high correlation between ABVS and MRI size (r=0.89). Other works on ABVS showed acceptable agreement with histology, leading to better size assessment than US, even if with a tendency to slight underestimation [13, 34]. A

IP T

recent study by Lagendijk et al. [35] showed that ABVS agrees significantly with histology even in terms of tumor volume (intraclass correlation coefficient 0.78), leading to slight

overestimation in larger tumors. Despite some Authors showed up to 70% concordance or

SC R

r=0.90 correlation with histology for DBT [8, 36, 37], as far as we know no previous studies

provided the absolute magnitude of agreement with MRI using the proper instrument of BlandAltman analysis [16]. We observed disappointing LOA of 0.45 to 2.24, which is in line with previously reported values ranging ±10 mm to ±20 mm when comparing DBT vs. pathology

U

[38]. Factors related to the disappointing performance of DBT in assessing cancer size include breast density and histological type (LOA up to ±41 mm in the case of ILC) [38]. Overall, our

N

results showed acceptable agreement between DBT-ABVS and MRI in assessing cancer size.

A

ABVS should be used as the reference tool to measure BC in the combined imaging strategy.

M

We have to acknowledge some study limitations. First, we investigated cancer detection in the context of preoperative assessment, thus carrying the risk of detection bias (i.e.,

ED

overestimation of sensitivity and PPV) given a scenario with 100% cancer prevalence. One might assume this limitation is further emphasized by the fact that readers were unblinded to images and reports of second-look ultrasound. Further studies should evaluate different clinical

PT

settings (e.g., screening or symptomatic women) to validate DBT-ABVS. Second, our series included a few cancers with lobular component (6.4% ILC and 5.5% invasive ductulo-lobular

CC E

carcinomas), thus reflecting a population with lower prevalence of multifocal and multicentric disease [39], and a weaker indication to staging MRI [1]. However, the higher number of IDC or DCIS in our series reflects the true prevalence of cancer histological types [40], and in turn, the clinical scenario to which DBT-ABVS should be applied as an alternative to DM and US.

A

Third, the follow-up to assess benign lesions was relatively short (median time 15 months). Our results should be confirmed by studies with prolonged follow-up to exclude any falsenegative cancer. Finally, one might argue this study mirrors clinical practice of a tertiary referral center breast imaging unit, with potential limited generalizability of the results. We believe that further works should assess inter-observer agreement of DBT-ABVS across different centers and/or radiologists with different experience.

10

In conclusion, DBT-ABVS approached the accuracy of MRI in the preoperative assessment of breast cancer, though DBT-ABVS showed lower sensitivity and PPV for additional disease. Assumed that MRI is the most reliable tool for breast cancer staging, DBT-ABVS might represent a better alternative to DM and US when MRI is not feasible and/or available. Compared to MRI, ABVS provided closer lesion sizes than DBT, suggesting that cancer should

DISCLOUSURES None of the Authors have any form of conflicts of interest to be disclosed.

IP T

be measured on ABVS-visible lesions in the combined imaging strategy.

SC R

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

A

CC E

PT

ED

M

A

N

U

All authors have approved the final article.

11

REFERENCES [1] Sardanelli F., Boetes C., Borisch B. et al, Magnetic resonance imaging of the breast: Recommendations from the EUSOMA working group, European Journal of Cancer. 46 (2010) 1296–1316. https://doi.org/10.1016/j.ejca.2010.02.015. [2] Preibsch H., Blumenstock G., Oberlechner E. et al, Preoperative breast MR Imaging in patients with primary breast cancer has the potential to decrease the rate of repeated

IP T

surgeries, Eur J Radiol. 94 (2017) 148-153. https://doi.org/10.1016/j.ejrad.2017.06.020. [3] Leithner D., Wengert G.J., Helbich T.H. et al (2017). Clinical role of breast MRI now and going forward, ClinRadiol, https://doi.org/10.1016/j.crad.2017.10.021.

and

mammography

for

breast

cancer,

J

Can

Res

https://doi.org/10.4103/jcrt.JCRT_678_17.

SC R

[4] Zhang Y., Ren H., Meta-analysis of diagnostic accuracy of magnetic resonance imaging Ther.

13

(2017)

862-868.

U

[5] Thomsen H.S. et al, ESUR Guidelines on Contrast Agents v 10.0, http://www.esur-

N

cm.org/index.php/en/, last access 04th July 2018.

A

[6] Mariscotti G., Nehmat H., Durando M. et al, Accuracy of Mammography, Digital Breast Tomosynthesis, Ultrasound and MR Imaging in Preoperative Assessment of Breast Cancer,

M

Anticancer Research. 34 (2014) 1219-1226.

[7] Helal M.H., Mansour S.M., Zaglol M., Salaleldin L.A., Nada O.M., Haggag M.A., Staging of

ED

breast cancer and the advanced applications of digital mammogram: what the physician needs to know?, Br J Radiol. 90 (2017) 20160717. https://doi.org/10.1259/bjr.20160717.

PT

[8] Luparia A., Mariscotti G., Durando M. et al, Accuracy of tumour size assessment in the preoperative staging of breast cancer: comparison of digital mammography, tomosynthesis,

CC E

ultrasound and MRI, Radiol Med. 118 (2013) 1119-1136. https://doi.org/10.1007/s11547-0130941-z.

[9] Pöhlmann S.T.L., Lim Y.Y., Harkness E., Pritchard S., Taylor C.J., Astley S.M., Threedimensional segmentation of breast masses from digital breast tomosynthesis images, J Med

A

Imaging (Bellingham). 4 (2017) 034007. https://doi.org/10.1117/1.JMI.4.3.034007. [10] Vedantham S., Karellas A., Vijayaraghavan G.R., Kopans D.B., Digital Breast Tomosynthesis:

State

of

the

Art,

Radiology.

277

(2015)

663–684.

https://doi.org/10.1148/radiol.2015141303. [11] Kaplan S.S., Automated whole breast ultrasound, Radiol Clin N Am. 52 (2014) 539–546. https://doi.org/10.1016/j.rcl.2014.01.002. 12

[12] Zanotel M., Bednarova I., Londero V. et al, Automated breast ultrasound: basic principles and

emerging

clinical

applications,

Radiol

Med.

123

(2018)

1-12.

https://doi.org/10.1007/s11547-017-0805-z. [13] Girometti R., Zanotel M., Londero V., Linda A., Lorenzon M., Zuiani C., Automated breast volume scanner (ABVS) in assessing breast cancer size: A comparison with conventional ultrasound and magnetic resonance imaging, Eur Radiol. 28 (2018) 1000-1008.

IP T

https://doi.org/10.1007/s00330-017-5074-7. [14] Huang A., Zhu L., Tan Y. et al, Evaluation of automated breast volume scanner for breast

conservation surgery in ductal carcinoma in situ, Oncol Lett. 12 (2016) 2481–2486.

SC R

https://doi.org/10.3892/ol.2016.4924.

[15] D’Orsi C.J., Sickles E.A., Mendelson E.B. et al, ACR BI-RADS® Atlas, Breast Imaging Reporting and Data System, American College of Radiology, Reston, 2013.

U

[16] Giavarina D., Understanding Bland Altman analysis, Biochem Med. 25 (2015) 141–151.

N

https://doi.org/10.11613/BM.2015.015.

[17] Wang Z.L., Xw J.H., Li J.L., Huang Y., Tang J., Comparison of automated breast volume

M

https://doi.org/10.1007/s11547-012-0836-4.

A

scanning to hand-held ultrasound and mammography, Radiol med. 117 (2012) 1287-1293.

[18] Roganovic D., Djilas D., Vujnovic S., Pavic D., Stojanov D., Breast MRI, digital

breast

cancer,

ED

mammography and breast tomosynthesis: Comparison of three methods for early detection of Bons

J

Basic

Med

Sci.

15

(2015)

64-68.

PT

https://doi.org/10.17305/bjbms.2015.616. [19] Gur D., Abrams G.S., Chough D.M. et al, Digital Breast Tomosynthesis: Observer

CC E

Performance Study, AJR. 193 (2009) 586–591. https://doi.org/10.2214/AJR.08.2031. [20] Chan H.P., Helvie M.A., Hadjiiski L. et al (2017). Characterization of Breast Masses in Digital Breast Tomosynthesis and Digital Mammograms: An Observer Performance Study. Acad Radiol, https://doi.org/10.1016/j.acra.2017.04.016.

A

[21] Lin X., Wang J., Han F., Fu J., Li A., Analysis of eighty-one cases with breast lesions using automated breast volume scanner and comparison with handheld ultrasound, European Journal of Radiology. 81 (2012) 873-878. https://doi.org/10.1016/j.ejrad2011.02.038. [22] Choi W.J., Cha J.H., Kim H.H. et al, Comparison of Automated Breast Volume Scanning and Hand-Held Ultrasound in the Detection of Breast Cancer: an Analysis of 5,566 Patient Evaluations.

Asian

Pac

J

Cancer

Prev.

15

(2014)

9101-9105.

https://doi.org/10.7314/APJCP.2014.15.21.9101. 13

[23] Wang H.Y., Jiang Y.X., Zhu Q.L. et al, Differentiation of benign and malignant breast lesions: A comparison between automatically generated breast volume scans and handheld ultrasound examinations, European Journal of

Radiology. 81 (2012) 3190–3200.

https://doi.org/10.1016/j.ejrad.2012.01.034. [24] Berg W.A., Gutierrez L., Ness Aiver M.S. et al, Diagnostic Accuracy of Mammography, Clinical Examination, US, and MR Imaging in Preoperative Assessment of Breast Cancer,

IP T

Radiology. 233 (2004) 830–849. https://doi.org/10.1148/radiol.2333031484. [25] Chang J.M., Cha J.H., Park J.S., et al, Automated breast ultrasound system (ABUS):

reproducibility of mass localization, size measurement, and characterization on serial

SC R

examinations, Acta Radiol. 56 (2015) 1163-1170. https://doi.org/10.1177/0284185114551565. [26] Golatta M., Franz D., Harcos A. et al, Interobserver realiability of automated breast volume scanner (ABVS) interpretation and agreement of ABVS findings with hand held breast ultrasound (HHUS), mammography and pathology results, European Journal of Radiology. 82

U

(2013) 332-336. https://doi.org/10.1016/j.ejrad.2013.03.005.

N

[27] Chudgar A.V., Conant E.F., Weinstein S.P. et al, Assessment of disease extent on

mammography

alone,

Clinical

Radiology.

72

(2017)

573-579.

M

https://doi.org/10.1016/j.crad.2017.02.013.

A

contrast-enhanced MRI in breast cancer detected at digital breast tomosynthesis versus digital

[28] Mercier J., Kwiatkowski F., Abrial C. et al, The role of tomosynthesis in breast cancer

ED

staging in 75 patients, Diagnostic and Interventional Imaging. 96 (2015) 27-35. https://doi.org/10.1016/j.diii.2014.06.010.

breast

PT

[29] Schelfout K., VanGoethemb M., Kersschotc E. et al, Contrast-enhanced MR imaging of lesions

and

effect

on

treatment,

EJSO.

30

(2004)

501–507.

CC E

https://doi.org/10.1016/j.ejso.2004.02.003. [30] Krammer J., Stepniewski K., Kaiser C.G. et al, Value of Additional Digital Breast Tomosynthesis for Preoperative Staging of Breast Cancer in Dense Breasts, Anticancer Res.

A

37 (2017) 5255-5261. https://doi.org/10.21873/anticanres.11950 [31] Gruber I.V., Rueckert M., Kagan K.O. et al, Measurement of tumour size with mammography, sonography and magnetic resonance imaging as compared to histological tumour

size

in

primary

breast

cancer,

BMC

Cancer.

13

(2013)

328.

https://doi.org/10.1186/1471-2407-13-328.

14

[32] Wasif N., Garreau J., Terando A., Kirsch D., Mund D.F., Giuliano A.E., MRI versus ultrasonography and mammography for preoperative assessment of breast cancer, Am Surg. 75 (2009) 970–975. [33] Schmachtenberg C., Fischer T., Hamm B., Bick U. (2017). Diagnostic Performance of Automated Breast Volume Scanning (ABVS) Compared to Handheld Ultrasonography With Breast MRI as the Gold Standard. Acad Radiol, https://doi.org/10.1016/j.acra.2017.01.021.

IP T

[34] Xu C., Wei S., Xie Y., Guan X., Yang B., Three-Dimensional Assessment of Automated Breast Volume Scanner Compared with Handheld Ultrasound in Pre-Operative Breast Invasive

Ductal Carcinomas: A Pilot Study of 51 Cases, Ultrasound Med Biol. 42 (2016) 2089–2096.

SC R

https://doi.org/10.1016/j.ultrasmedbio.2016.05.007.

[35] Lagendijk M., Vos E.L., Ramlakhan K.P. et al (2018). Breast and tumor volume measurements in breast cancer patients using 3-D automated breast volume scanner images.

U

World J Surg, https://doi.org/10.1007/s00268-017-4432-6.

[36] Förnvik D., Zackrisson S., Ljungberg O. et al, Breast tomosynthesis: Accuracy of tumor

N

measurement compared with digital mammography and ultrasonography, Acta Radiol. 51

A

(2010) 240-247. https://doi.org/10.3109/02841850903524447.

M

[37] Seo N., Kim H.H., Shin H.J. et al, Digital breast tomosynthesis versus full-field digital mammography: comparison of the accuracy of lesion measurement and characterization using

ED

specimens, Acta Radiol. 55 (2014) 661-667. https://doi.org/10.1177/0284185113503636. [38] Marinovich M.L., Macaskill P., Bernardi D., Houssami N. (2018). Systematic review of agreement between tomosynthesis and pathologic tumor sized for newly-diagnosed breast and

comparison

with

PT

cancer

other

imaging

tests.

Expert

Revi

Med

Devices,

https://doi.org/10.1080/17434440.2018.1491306.

CC E

[39] Menezes G.L.G., Van den Bosch A.A.J., Postma E.L. et al, Invasive ductolobular carcinoma of the breast: spectrum of mammographic, ultrasound and magnetic resonance imaging findings correlated with proportion of the lobular component, SpringerPlus. 2 (2013)

A

621. https://doi.org/10.1186/2193-1801-2-621. [40] Makki J., Diversity of Breas Carcinoma: Histological Subtypes and Clinical Relevance, Clin Med Insights Pathol. 8 (2015) 23-31. https://doi.org/10.4137/CPath.S31563.

15

FIGURE CAPTIONS Fig. 1 – True positive case in DBT-ABVS and MRI. Fifty-two years-old woman with index invasive ductal carcinoma of the left breast correctly assessed by DBT-ABVS. On craniocaudal DBT (a), cancer appeared as an irregular, vaguely spiculated mass with radiolucent intralesional component over a high-density background (arrows). Transverse native ABVS image (b) better defined a 13 mm sized hypoechoic lesion with indistinct margins, posterior shadowing and vertical orientation. MRI confirmed the presence of a mass with intense

IP T

contrast enhancement and similar size (13 mm), as shown in the transverse maximum intensity projection reconstruction (c).

SC R

Fig. 2 – False negative case in DBT-ABVS. Multicentric lesion missed by DBT-ABVS in a 49 years-old patient referred to surgery because of invasive ductal carcinoma (grade 3) associated with ductal carcinoma in situ (DCIS) in the left breast, as shown by transverse MRI

image (a). MRI found one additional multicentric DCIS lesion (b), without definite counterpart

U

on DBT (c) and ABVS (d). The index lesion was visible in the upper external quadrant on both

N

DBT (c) and ABVS (not shown).

Fig. 3 – False negative case in MRI. DCIS grade 3 index lesion missed by MRI in a 55 years-

A

old woman. While MRI examination showed no significant findings (a), DBT found a well-

M

defined architectural distortion (arrows), as evident in the upper quadrants of the medio-lateral oblique view (b). The corresponding ABVS finding on the transverse plane (c) was a

ED

hypoechoic, ill-defined nodule with vertical orientation. Fig. 4 – Agreement in measuring lesions’ size between ABVS vs. MRI (a), and DBT vs. MRI

A

CC E

PT

(b) according to Bland-Altman plots (see the text for details).

16

A ED

PT

CC E

IP T

SC R

U

N

A

M

Figr-1

17

A ED

PT

CC E

IP T

SC R

U

N

A

M

Figr-2

18

A ED

PT

CC E

IP T

SC R

U

N

A

M

Figr-3

19

A ED

PT

CC E

IP T

SC R

U

N

A

M

Figr-4

20

Tab. 1 Magnetic resonance imaging protocol used to evaluate breast cancer. TR= time of repetition; TE = time of echo; IR = inversion recovery time; STIR = short tau inversion recovery; EPI = echo-planar imaging; FLASH = Fast low angle shot T2-weighted

DWI

T1-weighted

System 2

System 1

System 2

System 1

System 2

Sequence

STIR

STIR

EPI

EPI

3D FLASH

3D FLASH

Plane

Transverse

Transverse

Transverse

Transverse

Transverse

Transverse

Duration

4.41 min

4.12 min

2.29 min

1.43 min

8.1 min

9.5 min

Field of view (mm)

320x320

340x340

330x165

350x171

340x340

340x340

Matrix size

230x384

336x448

84x168

102x208

512x512

512x512

Number of slices

40

44

24

34

72

80

Slice thickness (mm)

3

3

4

4

2

2

Distance factor (mm)

0.6

0.6

2

1

0

0

TR/TE (ms)

6980/ 73

5610/58

7100/84

5600/60

9/4.76

9/4.76

IRT (ms)

150

165

-

-

-

-

1

2

4

1

1

-

-

0.800

-

-

SC R

U

N

5

0.1000

A

CC E

PT

ED

b-values (sec/mm2)

A

excitations

M

Number of

IP T

System 1

21

Tab. 2 Diagnostic accuracy for breast cancer of combined Digital breast tomosynthesis and Automated breast volume scanner (DBT-ABVS) and Magnetic resonance imaging (MRI). Number in brackets are 95% C.I. TP = true positive findings; FN = false negative findings; FP = false positive findings; TN = true negative findings; PPV = positive predictive value; NPV = negative predictive value TP (n)

FN (n)

FP (n)

TN (n)

Sensitivity (%)

Specificity (%)

PPV (%)

NPV (%)

DBTABVS

99

9

7

45

91.7

86.5

93.4

83.3

90.0

(84.6 – 96.1 )

(74.2 – 94.4 )

(86.9 – 97. 3)

(70.8 – 92. 1)

(84.3 – 94.2)

MRI

10 4

96.3

88.5

(90.8 – 99.0 )

(76.6 – 95.6 )

76.5

86.5

(58.8 – 89.3 ) 91.2

Accuracy (%)

46

94.5

92.0

93.8

(89.1 – 97. 4)

(81.4 – 96. 8)

(88.8 – 97.0)

78.8

84.9

82.6

(74.2 – 94.4 )

(61.0 – 91. 0)

(72.4 – 93. 3)

(72.9 – 89.9)

88.5

83.8

93.9

89.5

(76.6 – 95.6 )

(68.0 – 93. 8)

(83.2 – 98. 7)

(81.1 – 95.1)

31

3

7

6

45

46

A

MRI

8

M

26

N

Non index lesions DBTABVS

A

CC E

PT

ED

(76.3 – 98.1 )

SC R

6

U

4

IP T

All lesions

22

Tab. 3 Overview of lesions missed by Digital breast tomosynthesis and Automated breast volume scanner (DBT-ABVS) vs. both Magnetic resonance imaging (MRI) and histology, and histology alone, and MRI (vs. both DBT-ABVS and histology, and histology alone). IDC = invasive ductal carcinoma; DCIS = ductal carcinoma in situ Lesion type

Number

Index lesion

Non-index

Mean size on

lesion

histology

DBT-ABVS vs. MRI and histology

IP T

(range)

1

-

1 (multifocal)

8 mm

IDC with DCIS

1

1

-

6 mm

DCIS

4

-

4 (1 multicentric,

SC R

IDC

23 mm (6-54)

U

3 multifocal)

1

1

-

8 mm

-

1 (multicentric)

3 mm

-

2 (multifocal)

2.5 mm (2-3)

A

DCIS

DCIS

2

A

CC E

PT

ED

1

M

DBT-ABVS and MRI vs. histology IDC

N

MRI vs. DBT-ABVS and histology

23