Digital breast tomosynthesis versus full-field digital mammography—Which modality provides more accurate prediction of margin status in specimen radiography?

Digital breast tomosynthesis versus full-field digital mammography—Which modality provides more accurate prediction of margin status in specimen radiography?

European Journal of Radiology 93 (2017) 258–264 Contents lists available at ScienceDirect European Journal of Radiology journal homepage: www.elsevi...

547KB Sizes 1 Downloads 23 Views

Recommend Documents

A prospective study comparing digital breast tomosynthesis with digital mammography in surveillance after breast cancer treatment
•This is the first reported study investigating adjunct digital breast tomosynthesis (DBT) in surveillance after breast

Fully automated nipple detection in digital breast tomosynthesis
•We propose a nipple detection algorithm for DBT.•The algorithm includes a segmentation method of breast for DBT.•The al

Digital radiography
DSA has quickly made the transition from an experimental examination performed in a research institution to a routine pr

Digital tomosynthesis: A new future for breast imaging?
The aim of this article is to review the major limitations in current mammography and to describe how these may be addre

Digital breast tomosynthesis: Dose and image quality assessment
•Comparison among dose and image quality of four DBT systems is described.•Systems performance has been related with the

P6. Digital breast tomosynthesis: Dose and image quality assessment
Digital Breast Tomosynthesis (DBT) offers a gain in sensitivity and in specificity for the detection of breast cancers c

Digital Breast Tomosynthesis: Dose and image quality characterization
Digital Breast Tomosynthesis (DBT) offers a gain in sensitivity and in specificity for detection of breast cancers compa

Pelvimetry by digital radiography
We studied the ability of digital radiographs generated on a computed tomography scanner to replace conventional pelvime

European Journal of Radiology 93 (2017) 258–264

Contents lists available at ScienceDirect

European Journal of Radiology journal homepage:

Research papers

Digital breast tomosynthesis versus full-field digital mammography—Which modality provides more accurate prediction of margin status in specimen radiography?


Heba A. Amera,b, Florian Schmitzbergerb, Barbara Ingold-Heppnerc, Julia Kussmaule, ⁎ Manal F. El Tohamya, Hazim I. Tantawya, B. Hammb, M. Makowskid, Eva M. Fallenbergb, a

Dept of Radiology, Zagazig University Hospitals, Zagazig, Egypt Clinic of Radiology, Charité, Campus Virchow-Klinikum, Augustenburger Platz 1, 13353 Berlin, Germany Institute of Pathology, Charité Campus Mitte, Charitéplatz 1, 10117 Berlin, Germany d Clinic of Gynacolgy and Breast Center, Charité Campus Mitte, Charitéplatz 1, 10117 Berlin, Germany e Clinic of Radiology, Charité Campus Mitte, Charitéplatz 1, 10117 Berlin, Germany b c



Keywords: Specimen radiography Digital breast tomosynthesis Margin evaluation Digital mammography

Objectives: To evaluate the reliability of tumor margin assessment in specimen radiography (SR) using digital breast tomosynthesis (DBT) and full-field digital mammography (FFDM) in comparison to postoperative histopathology margin status as the gold standard. Methods: After ethics committee approval, 102 consecutive patients who underwent breast conservative surgery for nonpalpable proven breast cancer were prospectively included. All patients underwent ultrasound/mammography-guided wire localization of their lesions. After excision, each specimen was marked for orientation and imaged using FFDM and DBT. Two blinded radiologists (R1, R2) independently analyzed images acquired with both modalities. Readers identified in which direction the lesion was closest to the specimen margin and to measure the margin width. Their findings were compared with the final histopathological analysis. True positive margin status was defined as a margin measuring < 1 mm for invasive cancer and 5 mm for ductal carcinoma in situ (DCIS) at imaging and pathology. Results: For FFDM, correct margin direction was identified in 45 cases (44%) by R1 and in 37 cases (36%) by R2. For DBT, 69 cases (68%) were correctly identified by R1 and 70 cases (69%) by R2. Overall accuracy was 40% for FFDM and 69% for DBT; the difference was statistically significant (p < 0.0001). Sensitivity in terms of correct assessment of margin status was significantly better for DBT than FFDM (77% versus 62%). Conclusion: SR using DBT is significantly superior to FFDM regarding identification of the closest margin and sensitivity in assessment of margin status.

1. Introduction With the introduction of mammography screening, the percentage of patients diagnosed with early breast cancer eligible for breast-conserving treatment instead of mastectomy is increasing. The overall goal of any breast-conserving surgery (BCS) is to achieve complete excision of the malignant lesion with adequate tumor-free resection margins since the risk of local recurrence and therefore poorer prognosis is directly related to positive tumor margins [1]. So far, it is still difficult to achieve tumor-free resection margins in the first attempt

at BCS. Several studies report high re-excision rates ranging from 20% − 60% [2–4]. Specimen radiography (SR) is one of the standard techniques performed to examine the integrity of the resection margin while the patient is still in the operating room. The findings guide the surgeon in either completing the operation without additional excision or extending the resection to achieve a tumor-free resection bed in a one-step operation [5–9]. However, the current techniques for performing SR using digital mammography have low sensitivities compared with final histopathological margin analysis [10].

Abbreviations: DBT, digital breast tomosynthesis; FFDM, full-field digital mammography; SR, specimen Radiography; IDC/NST, invasive ductal carcinoma/carcinoma of non-special type; DCIS, ductal carcinoma in situ; BCS, breast conserving surgery; R1/R2, Reader 1/Reader 2 ⁎ Corresponding author. E-mail addresses: [email protected] (F. Schmitzberger), [email protected] (B. Ingold-Heppner), [email protected] (J. Kussmaul), [email protected] (B. Hamm), [email protected] (M. Makowski), [email protected] (E.M. Fallenberg). Received 26 November 2016; Received in revised form 25 May 2017; Accepted 29 May 2017 0720-048X/ © 2017 Published by Elsevier Ireland Ltd.

European Journal of Radiology 93 (2017) 258–264

H.A. Amer et al.

Breast Unit for imaging. The routine surgical strategy to achieve tumor-free margins at our institution is as follows: when invasive cancer without evidence of ductal carcinoma in situ (DCIS) is present, additional excision during the primary operation is performed if a margin width ≤1 mm is measured; an optional extension is indicated for margin widths ≥1 and ≤5 mm (following intraoperative discussion with the surgeon). When there is evidence of DCIS, additional excision is mandatory for margin widths ≤5 mm and optional for margin widths ≤10 mm. On the real time of the operation of cases included in this study, the radiologist recommendations were based only on viewing the FFDM specimen. Whenever the lesion was located centrally in the specimen with abundant visible tumor-free margins in all directions, the surgeon was informed to be able to terminate the procedure; when the lesion was found to be close to a margin, the surgeon was told the direction in which to extend the excision, and further tissue was removed in this direction before the operation was considered to be completed. The surgeons were blinded to the study.

Over the last decade, digital breast tomosynthesis (DBT) has been expressly developed as a supplement to screening and diagnostic digital mammography as it has been shown to lower false positive recall rates and improve the sensitivity and specificity via reducing tissue overlapping and masking [11–13]. We have, so far, found only few studies in the literature that discuss the possibility of using DBT in performing SR [14–16]. However, these studies only evaluate the accuracy of DBT for lesion detection, characterization or lesion size measurement. The aim of our study is to evaluate the reliability and added value of digital breast tomosynthesis in the intraoperative assessment of tumor resection margins in excised specimens, compared to digital mammography, using the final postoperative histopathological analysis of margin status as the gold standard. 2. Materials and methods 2.1. Study design and patient population After ethics committee approval, 102 of 182 patients who underwent specimen radiography during planned BCS for nonpalpable breast lesions between 2010 and 2012 were consecutively included. Inclusion criteria were: patients with biopsy-proven breast cancer who underwent preoperative guided wire localization of the index lesion. Exclusion criteria were: (1) Patients who had previous surgical interventions (2) Patients who had vacuum-assisted biopsy of the lesion because of the frequent presence of artifacts exerted by the biopsy clips or biopsy clips with no remaining visible lesion; (3) benign lesions; and (4) patients with more than one wire-localized breast lesion seen within one breast specimen, as radiologic–pathologic correlation might be difficult.

2.3. Imaging protocol All imaging procedures were performed on the same digital platform, a MAMMOMAT Inspiration mammography system with tomosynthesis option (Siemens Healthcare, Erlangen, Germany). After receiving the specimen with wire in place from the surgeon, routine two view FFDM specimen radiography was performed. The surgical specimen was positioned on the mammography plate and oriented as for the first view, and then rotated 90° laterally to obtain the second view. In the second view, the single cranial clip was imaged enface within the specimen as shown in Fig. 2. The imaging was performed without real compression but the specimen was a little bit flattened with the compression paddle to equalize the tissue in a magnification view of 1:1.7. Images were then acquired for DBT specimen radiography in orientation similar to that obtained for the 1st view FFDM specimen using standard tomosynthesis acquisition techniques, which generated 25 low-dose projection images or frames over a 50° arc. The 25 raw images were reconstructed into a series of 1-mm thick images at 1-mm intervals, spanning the entire specimen tissue thickness and resulting in a set of 20–50 parallel slices depending on specimen thickness. Tomosynthesis was performed without compression but again with slightly flattened breast tissue. After imaging, the surgical specimen was sent to the pathologist for histological assessment.

2.2. Surgical procedure Prior to the planned therapeutic surgical excision, all patients underwent either mammography- or ultrasonography-guided wire localization followed by orthogonal 2-view mammography to show the position of the wire in relation to the mammographic abnormality. Thereafter, surgical wide local excision was performed. To ensure proper specimen orientation, each specimen was clipped in the operating theater, placing 1 clip in the cranial resection margin and 2 clips in the lateral resection margins (Fig. 1) and a non-clipped suture on the anterior surface. After preparation, the specimen was transferred to the

2.4. Image review Two board-certified readers with ten and 5 years of experience in breast imaging (R1, R2) analyzed the anonymized specimen radiographs retrospectively in a randomized manner between FFDM and the DBT specimen. Each modality was reviewed independently with at least 30 days between the two readings in order to exclude potential “shortterm memory” bias. The radiologists were blinded to patient history and tumor information; however, they knew that all women had proven breast cancer, other than that they were completely blinded to any clinical, surgical, histopathological or previous radiological data. Reviewing was performed subsequently to clinical management. Existing radiographic findings either previous preoperative examinations or the previous reports of the specimen by other radiologist were not allowed to be used in the assessment (as there were no preoperative DBT examinations equivalent to the routine preoperative mammography), so readers were instructed to assume that the reading was the initial specimen examination. The reviewers were allowed to mark multiple lesions if present and make a separate rating for each lesion in the same examination as deemed appropriate. The two readers were asked to rate the images from each modality (FFDM and DBT) for: (1) Presence or absence of the lesion within the specimen.

Fig. 1. Diagram of specimen with 3 clips placed for orientation and dotted line indicating the closest margin (i.e., shortest distance between lesion and specimen edge). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)


European Journal of Radiology 93 (2017) 258–264

H.A. Amer et al.

Fig. 2. FFDM specimen views in relation to clip positions.

2.6. Statistical analysis

(2) Lesion conspicuity on a 4-point scale: 1 = no lesion visible 2 = poor 3 = moderate 4 = good. (3) Type of lesion and features:

For Data analysis, the lesions were sub grouped according to their histopathology types to 3 categories: Pure Invasive carcinoma group, DCIS group and Invasive cancers with associated in situ components. A two-sample t-test for equality of proportions was performed. The significance level was defined as p < 0.05. As the two readers were allowed to identify only one margin as the closest margin and record its width, margin width analysis was only possible in cases that had the same direction assignment radiologically and pathologically. For analysis of margin width, diagnostic accuracy, sensitivity, specificity, and negative and positive predictive values were calculated for all tumors and for each tumor subgroup. The Wilcoxon signed-rank test and BlandAltman analysis were used to test correlation between the measures obtained with each imaging modality and histopathology. All analysis were done with the software “R” Project for Statistical Computing (R Development Core Team 2008, Vienna, Austria)

• mass • architectural distortion (AD) • calcifications (Ca + +) • combination. (4) Size of the lesion. Size on tomosynthesis images was measured on the slice with the longest tumor extension. (5) Identification of the direction (medial, lateral, superior, inferior, anterior or posterior) in which the lesion is closest to the specimen edge using the clip positions for orientation. In addition, readers were asked to measure the margin width (=distance of the lesion from the edge) in this direction; see Fig. 1. When more than one lesion was identified in the specimen (e.g. satellite), the closest margin was assigned in relation to the lesion that was closest to the specimen edge either it is the index lesion (that is marked with wire) or the additional lesion

3. Results 3.1. A. Tumor characteristics The final histopathological diagnoses of the 102 specimens were: pure invasive carcinomas in 26 (25%), DCIS in 16 (15%), and a combination of invasive and in situ cancers in 60 (60%) of the specimens. In the group of pure invasive cancers, 19 were invasive carcinoma of nonspecial type (NST, formerly invasive ductal carcinoma), 4 were invasive lobular carcinomas, 2 mucinous carcinomas, and 1 medullary carcinoma. Thirty-three cases were primarily resected with cancer-free margins, classified as clear. In 69 patients, the resection margin was positive for cancer cells (involved margin). Only 6 of the 69 patients (8%) had pure invasive lesions. The margin status results are summarized in Table 1. The high number of involved margins could be explained by the strict protocol we follow in our institution in specimen imaging (> 5 mm for DCIS, > 1 mm for Invasive cancers) which is smaller than those limits applied in other studies which used 1 cm limit for both invasive cancers and DCIS. Therefore the probability of

2.5. Histopathological workup and analysis Each surgical specimen, including any extra tissue that had been secondary excised during the operation, underwent postoperative pathologic workup, which included estimates of size, grade and type of tumor present. Additionally, the minimal distance (in mm) from the tumor to the resection margins in all directions (Anterior, Posterior, Medial, Lateral, Superior, and Inferior) were reported. In case an intraoperative secondary excision was performed on the recommendation of the radiologist, the number and directions of the secondary excision were stated in the operation report. The final histopathology report is made up of two parts; the primary record indicates the smallest margins for the specimen initially resected by the surgeon. The secondary/final record is made after including the secondary resected tissue in the corresponding direction after the recommendation of the radiologist. Only the primary records were used for the purpose of margin comparison between histopathology and Imaging as the secondary resections were not imaged. Whenever histopathology analysis is mentioned later here this is meant to be only for the primary record of the specimen with no inclusion of the findings of the secondary records of the margins after the additional tissue resection.

Table 1 Margin status of the 102 specimens after primary resection based on final histopathological analysis.


Margin Status

Clear margin

Involved margin

DCIS (16) Pure invasive Ca (26) Invasive + DCIS (60)

1 20 12

15 6 48

European Journal of Radiology 93 (2017) 258–264

H.A. Amer et al.

Fig. 3. Pathology results after 1ry and 2ry resection.

involved margins in histopathology is higher. Further pathology details were illustrated by Fig. 3. Radiologically, Table 2 shows the number of non-detected lesions and the rated lesion visibility by the two readers in both modalities.

Table 3 Accuracy of identification of the direction of the closest margin by reader and modality for all breast cancers and cancer subgroups.

3.2. B. Closest margin direction analysis All cancers(102) DCIS (16) Pure invasive Ca (26) Invasive + DCIS (60)

In the 102 cases, R1 identified the smallest margin in the same direction as in pathology in 45 cases in FFDM specimen versus 69 cases in DBT specimen. R2 identified the smallest margin in the same direction as in pathology in 37 cases in FFDM specimen versus 70 cases in DBT specimen. For use of FFDM specimen, accuracy of correct direction detection was 44% for R1 and 36% for R2. For use of DBT specimen, accuracy was 68% and 69% for R1 and R2, respectively, see Table 3. The p-value of the t-test was significant (< 0.001 for R1 and < 0.000001 for R2). Looking at tumor subgroups, accuracy using FFDM specimen was highest (52%) for lesions with a purely invasive component. In DBT specimen, accuracy was highest (72%) for lesions with both invasive and DCIS components. Individual accuracies for all tumors and each tumor subgroup are presented in Table 3.

DBT specimen R1


41 28 32 1

50 14 36 2

43 26 27 6

33 23 41 5



(45) 44% (6) 38% (16) 62% (23) 38%

(37) 36% (6) 38% (11) 42% (20) 33%

(69) (10) (15) (44)

R2 68% 63% 58% 73%

(70) (12) (16) (42)

69% 75% 62% 70%

In specimen with correctly identified direction DBT specimen had significantly higher overall sensitivity and improved accuracy compared with FFDM specimen. Sensitivities/accuracies were 77%/77% for DBT specimen and 62%/73% for FFDM specimen. The difference for sensitivity between the modalities is significant (p = 0.03). Specificity, on the other hand, was better for FFDM specimen with 98% versus 77% for DBT specimen. Sensitivities, specificities, PPV and NPV for all tumors and for individual tumor subgroups are summarized in Table 4.

Table 2 Number of rated lesions and rated lesion visibility in both modalities by the two readers.



3.4. Accuracy of margin status assessment with either modality compared with pathology

Accurate measurement of margin width with each modality is


DBT specimen

crucial for classification as positive or negative for presence of tumor. Assessment of margin status was better for DBT specimen. Fifty-two and 55 margins were classified correctly (true negative or true positive) using tomosynthesis by R1 and R2, respectively. With use of FFDM specimen, 36 and 25 margins were classified correctly by R1 and R2, respectively.

3.3. C. Margin width analysis

FFDM specimen

FFDM specimen

3.5. Wilcoxon signed-rank test Good Intermediate Poor Non visible

The margin width measured from FFDM specimen was found to be significantly different from that measured by pathology in the two subgroups with purely invasive cancer and with cancers having 261

European Journal of Radiology 93 (2017) 258–264

H.A. Amer et al.


Table 4 Sensitivity, specificity, PPV and NPP for FFDM specimen and DBT specimen for all cancers and cancer subgroups regarding margin status assessment. Modality

FFDM specimen

DBT specimen


Cancer type


All cancers DCIS Pure invasive Ca Invasive + DCIS

62% 70% n.a.a

97.4% 100 96.2%

98% 100% n.a.a

58% 42% 76%

73% 75% 61%






77% 95.8% n.a.

77% n.a. 77%

89% 100% n.a

61% n.a 83%

77% 96% 68%






All cancers DCIS Pure invasive Ca Invasive + DCIS




4. Discussion


According to the European guidelines for quality assurance in breast cancer screening and diagnosis [17], the reliable lesion localization e.g. wire localization and specimen radiography is essential for all nonpalpable lesions detected by mammography in women who undergo surgery to confirm adequate resection before skin closure. The same is recommended in the surgical guidelines for the management of breast cancer [18]. Various studies investigating the accuracy and sensitivity of specimen radiography have also looked at different techniques and measures (digital imaging, use of magnification, intraoperative equipment) aimed at improving its accuracy and reliability in determining distance of cancer from the specimen edge compared with the final histological margins [6,7,19,20]. However, reported sensitives are poor, ranging from 30% to 68%, depending on how the safety margin is defined [6,7,20,21]. To the best of our knowledge, only a small number of studies in the literature discuss the possibility of using DBT for performing specimen radiography [14–16]. However, these studies focus on the evaluation of the accuracy of DBT for lesion detection, characterization, or measurement of lesion size but not for intraoperatively assessing margin status of resected specimens. In this study, we therefore evaluated specimen radiography using DBT in more detail with regard to specific breast cancer subgroups, accuracy of identifying the orientation of the closest resection margin, and accuracy of margin status assessment in terms of presence versus absence of cancer.

n.a. due to small number of cases after subgrouping.

Table 5 Wilcoxon signed-rank test p-values and number of cases for margin widths differences between those measured with the two modalities in comparison to pathology. FFDM specimen

DBT specimen





All cancers

< 0.00001

< 0.00001



DCIS No difference 1–3 mm 3–5 mm > 5 mm

0.1696 1 3 1 1

0.1975 2 3 1 –

0.932 3 6 1 –

1 5 5 2 –

Pure invasive Ca No difference 1–3 mm difference 3–5 mm difference > 5 mm difference

0.01 1 8 2 5

0.03 1 6 2 2

0.429 1 10 4 –

0.9242 2 11 3 –

Invasive + DCIS No difference 1–3 mm difference 3–5 mm difference > 5 mm difference

< 0.00001 1 16 4 2

0.0001 – 10 4 6

0.001 9 20 11 4

0.02 6 23 9 4

Total cases





4.1. Closest margin orientation In this study, DBT specimen showed significantly higher accuracies (68% R1, 69% R2) in defining the correct direction of the closest margin compared with FFDM specimen (44% R1, 36% R2). This observation holds for all subgroup analyses except for the results of R1 for the purely invasive cancer subgroup. The significant improvement in closest margin identification by DBT can be attributed to the better delineation of tumor margins and elimination of possible tissue overlapping over parts of the lesions, one example is shown in Fig. 4. We are not aware of other studies that examined the accuracy of DBT specimen for correct orientation of the closest margin, while two studies [6,20] reported 56% and 48% accuracy of FFDM specimen for correct identification of the orientation of the closest margin. We attribute the poor agreement in closest margin direction between radiology and pathology in the present study and earlier studies [6,20] to the fact that we cannot assume with 100% certainty that orientation of the specimen during radiography is always the same as during pathological examination, either because of specimen movement during transfer or because of the nature and shape of specimens, which all are potential errors in specimen handlings [22].

invasive and in-situ components. In contrast, margin width measured from DBT specimen differed significantly from that measured by pathology only for the mixed invasive/in-situ cancer subgroup. P-values obtained with the Wilcoxon signed-rank test are shown in Table 5. 3.6. Bland-altman analysis

4.2. Margin status assessment and margin width analysis For DBT specimen, estimated biases were +0.9 mm for R1 and +0.6 mm for R2, meaning margin widths measured with this modality were close to those determined by histopathology with a slightly overestimation. The values were higher for FFDM specimen (2.8 mm for R1 and 3.7 mm for R2) resulting in a higher overestimation. The results are summarized in. Looking at tumor subgroups, the best margin estimation was achieved for purely invasive cancers using DBT specimen, with a deviation which was almost equal to zero mm. The deviation was greatest for the subgroup of mixed invasive cancers with DCIS components using FFDM specimen (4.4 mm). Generally, the estimated biases were lower for DBT specimen than for FFDM specimen, ranging from 0.6 mm margin underestimation to 1.7 mm margin overestimation for DBT specimen and from 1.5 mm to 4.4 mm margin overestimation for FFDM

The optimal, tumor-free safety margin width is still a matter of debate and ranges from 1 mm to 10 mm, depending on the investigator, institution, and country [23]. For example, Schulz-Wendtland et al. [16], investigating breast specimens using tomosynthesis, defined a minimum margin of 10 mm for all breast cancer types. Mazouni et al. [9] tested different radiological cut-offs (1, 5, 10 mm) for sensitivity and specificity with the 10 mm threshold providing the higher sensitivity (75%). The reported sensitivities of standard FFDM specimen range from 30%–75% [6,7,20,21]. In our institution, we use a margin threshold of ≥ 1 mm for purely invasive cancers and ≥5 mm for DCIS as this are the achievable resection margins regarding our surgical guidelines. These thresholds were also used to calculate the sensitivity in our study. 262

European Journal of Radiology 93 (2017) 258–264

H.A. Amer et al.

Fig. 4. Bland-Altman plots showing the difference between the averages of each modality versus pathology measurements (in mm). The blue line in the middle indicates the estimated biases while the upper and lower blue lines represent the limits of agreement (mean ± 1.96 SD). Right breast FFDM Specimen, first (A) and second (B) views, showing the excised ill defined mass lesion traversed by the localized wire. The closest margin seems to be more on Medial direction. (C) Selected slices T-SR of the same patients demonstrate better delineation of the tumor border with surrounding desmoplastic reaction. In addition small satellite lesion is clearly seen at the caudal edge of the specimen, proved to be another IDC focus by histopathological examination.

pathologists in our study can be explained in the context of two assumptions: first, the underestimation of tumor size due to different growth patterns (mammographically occult, multifocal, diffusely infiltrating or discontinuous with or without evident tumor mass) and, second, changes occurring during postoperative specimen handling. The so-called pancake phenomenon, proposed by Graham et al. [22], describes the observation that breast tissue specimens become flattened and lose almost 50% of their original height. These changes can distort the distance between the lesion and specimen edges compared with the in vivo situation. For those cases with correct orientation identification of the closest resection margin, the Wilcoxon signed-rank test showed that the distance measured using DBT specimen was significantly different (P = 0.01) from that measured in pathology for one reader and nonsignificantly different for the other reader. In contrast, using FFDM specimen, both readers measures were highly significantly different from those measured in pathology (p < 0.000001). In contrast, when looking at tumor types, DBT specimen did not show any significant difference from pathology measures in DCIS and pure invasive groups for either reader. A significant deviation was only found for the mixed invasive + DCIS group. While we were interested in investigating whether the accuracy of specimen radiography is affected by the cancer type, published studies on the accuracy of measuring the extent of breast tumor by different imaging modalities usually relate their results to breast density or the BIRADS score [14,25,27,28]. Our study has several limitations. First, transfer of the specimens from the operative theater to the breast unit as well as specimen

In the our study, there is significantly better sensitivity of DBT specimen over that of FFDM specimen. That is in keeping with the results of Schulz-Wendtland et al. [16] [24], they reported 86.6% sensitivity for DBT versus 78.3% for standard FFDM and 83% for FFDM with 1.0:1.8 magnification. The higher sensitivities for both FFDM and DBT specimen reported by Schulz-Wendthland et al. in comparison to those reported in our study can be explained by the larger margin width (> 1 cm) they used to define safe margins regardless of the cancer subgroup. We also calculated sensitivities for individual tumor subgroups but due to the small sample size and subsequently smaller representation across the groups significant difference across them cannot be found neither by FFDM or DBT specimen… Bland-Altman analysis of margin width measurements in our study revealed the results obtained with DBT specimen to be closest to the measurements in the pathology reports with average overestimations of only 0.8 mm for tomosynthesis versus 3.3 mm for FFDM specimen. On the other hand, there is consistent overestimation of the margin distance by both readers using both modalities, while the overestimation is statistically significantly less apparent for both readers using DBT specimen. Overestimation of margin width reflects the fact that lesion size is mostly underestimated with radiology, as reported in different studies [25,26]. Margin overestimation in our study ranged from 1–13 mm using FFDM specimen and 1–7 mm using DBT specimen. Britton et al. suggest that an 11 mm margin measured in specimen radiography correlates most strongly with achieving a 5 mm histological margin [6]. The discrepancies in margin widths measured by radiologists and 263

European Journal of Radiology 93 (2017) 258–264

H.A. Amer et al.

surgery, Ann. Surg. Oncol. 14 (4) (2007) 1458–1471. [9] C. Mazouni, R. Rouzier, C. Balleyguier, L. Sideris, F. Rochard, S. Delaloge, H. Marsiglia, M.C. Mathieu, M. Spielman, J.R. Garbay, Specimen radiography as predictor of resection margin status in non-palpable breast lesions, Clin. Radiol. 61 (9) (2006) 789–796. [10] R.G. Roth, A.D. Maidment, S.P. Weinstein, S.O. Roth, E.F. Conant, Digital breast tomosynthesis: lessons learned from early clinical implementation, Radiographics 34 (4) (2014) E89–102. [11] J.M.P. Elizabeth, A. Rafferty, E. Liane Philpotts, P. Steven Poplack, H. Jules Sumkin, F. Elkan Halpern, T. Loren Niklason, Assessing radiologist performance using combined digital mammography and Breast tomosynthesis compared with digital mammography alone: results of a multicenter, multireader trial, Radiology 266 (2013). [12] E.A. Rafferty, J.M. Park, L.E. Philpotts, S.P. Poplack, J.H. Sumkin, E.F. Halpern, L.T. Niklason, Diagnostic accuracy and recall rates for digital mammography and digital mammography combined with one-view and two-view tomosynthesis: results of an enriched reader study, AJR. Am. J. Roentgenol. 202 (2) (2014) 273–281. [13] P. Skaane, A.I. Bandos, R. Gullien, E.B. Eben, U. Ekseth, U. Haakenaasen, M. Izadi, I.N. Jebsen, G. Jahr, M. Krager, L.T. Niklason, S. Hofvind, D. Gur, Comparison of digital mammography alone and digital mammography plus tomosynthesis in a population-based screening program, Radiology 267 (1) (2013) 47–56. [14] N. Seo, H.H. Kim, H.J. Shin, J.H. Cha, H. Kim, J.H. Moon, G. Gong, S.H. Ahn, B.H. Son, Digital breast tomosynthesis versus full-field digital mammography: comparison of the accuracy of lesion measurement and characterization using specimens, Acta Radiol. 55 (6) (2014) 661–667. [15] M. Urano, N. Shiraki, T. Kawai, T. Goto, Y. Endo, N. Yoshimoto, T. Toyama, Y. Shibamoto, Digital mammography versus digital breast tomosynthesis for detection of breast cancer in the intraoperative specimen during breast-conserving surgery, Breast Cancer 23 (5) (2016) 706–711. [16] R. Schulz-Wendtland, G. Dilbat, M. Bani, P.A. Fasching, K. Heusinger, M.P. Lux, C.R. Loehberg, B. Brehm, M. Hammon, M. Saake, P. Dankerl, S.M. Jud, C. Rauh, C.M. Bayer, M.W. Beckmann, M. Uder, M. Meier-Meitinger, Full Field Digital Mammography (FFDM) versus CMOS Technology, Specimen Radiography System (SRS) and Tomosynthesis (DBT) – Which System Can Optimise Surgical Therapy? Geburtshilfe und Frauenheilkunde 73 (5) (2013) 422–427. [17] N. Perry, M. Broeders, C. de Wolf, S. Tornberg, R. Holland, L. von Karsa, European guidelines for quality assurance in breast cancer screening and diagnosis. Fourth edition?summary document, Ann. Oncol. 19 (4) (2008) 614–622. [18] Association of Breast Surgery at Baso, Surgical guidelines for the management of breast cancer, Eur. J. Surg. Oncol. 35 (Suppl. 1) (2009) 1–22. [19] S. Goldfeder, D. Davis, J. Cullinan, Breast specimen radiography: can it predict margin status of excised breast carcinoma? Acad. Radiol. 13 (12) (2006) 1453–1459. [20] C. Schmachtenberg, F. Engelken, T. Fischer, U. Bick, A. Poellinger, E.M. Fallenberg, Intraoperative specimen radiography in patients with nonpalpable malignant breast lesions, RoFo: Fortschritte auf dem Gebiete der Rontgenstrahlen und der Nuklearmedizin 184 (7) (2012) 635–642. [21] N.J. Coombs, P.P. Vassallo, A.J. Parker, C. Yiangou, Radiological review of specimen radiographs after breast localisation biopsy is not always necessary, Eur. J. Surg. Oncol. 32 (5) (2006) 516–519. [22] R.A. Graham, M.J. Homer, J. Katz, J. Rothschild, H. Safaii, S. Supran, The pancake phenomenon contributes to the inaccuracy of margin assessment in patients with breast cancer, Am. J. Surg. 184 (2) (2002) 89–93. [23] H. Krontiras, R.B. Lancaster, M.M. Urist, What is a clear margin in breast conserving cancer surgery? Curr. Treat. Options Oncol. 15 (1) (2014) 79–85. [24] R. Schulz-Wendtland, G. Dilbat, M.R. Bani, M.P. Lux, M. Meier-Meitinger, E. Wenkel, S. Schwab, M.W. Beckmann, M. Uder, B. Adamietz, Use of tomosynthesis in intraoperative digital specimen radiography − is a reduction of Breast Re-excision rates possible? Geburtshilfe Frauenheilkd 71 (12) (2011) 1080–1084. [25] D. Fornvik, S. Zackrisson, O. Ljungberg, T. Svahn, P. Timberg, A. Tingberg, I. Andersson, Breast tomosynthesis: accuracy of tumor measurement compared with digital mammography and ultrasonography, Acta Radiol. 51 (3) (2010) 240–247. [26] A.M. Bosch, A.G. Kessels, G.L. Beets, J.D. Rupa, D. Koster, J.M. van Engelshoven, M.F. von Meyenfeldt, Preoperative estimation of the pathological breast tumour size by physical examination, mammography and ultrasound: a prospective study on 105 invasive tumours, Eur. J. Radiol. 48 (3) (2003) 285–292. [27] A. Luparia, G. Mariscotti, M. Durando, S. Ciatto, D. Bosco, P.P. Campanino, I. Castellano, A. Sapino, G. Gandini, Accuracy of tumour size assessment in the preoperative staging of breast cancer: comparison of digital mammography, tomosynthesis, ultrasound and MRI, La Radiologia Medica 118 (7) (2013) 1119–1136. [28] H.S. Mun, H.H. Kim, H.J. Shin, J.H. Cha, P.L. Ruppel, H.Y. Oh, E.Y. Chae, Assessment of extent of breast cancer: comparison between digital breast tomosynthesis and full-field digital mammography, Clin. Radiol. 68 (12) (2013) 1254–1259.

repositioning between FFDM specimen and DBT specimen may have impaired precise and consistent specimen orientation. Second and most important, histopathological examinations were performed by different pathologists without use of standardized procedures for correlation of radiologically and pathologically measured directions. Nevertheless this reflects the common routine standard in a lot of breast centers. Third as this study is based on retrospective reviewing of the images, strict commitment to the marking protocol cannot be assured, even if it is a standard defined procedure, in addition the readers were at disadvantage of being able to use the wire direction as guidance for specimen orientation and also viewing the initial mammography or the mammography taken after wire insertion. Fourth, these results are of single manufacture's DBT equipment which may or may not extend to other manufacturer's equipment. These can vary in detector types and total scan angle. 5. Conclusion DBT is a promising modality in performing specimen analysis. It significantly improves the accuracy of SR regarding identification of the closest margin and sensitivity regarding margin status assessment compared to FFDM. This could reduce re-excision and re-operation rates. Conflict of interest No author has a conflict of interest. Acknowledgements We are grateful to Nikola Bangeman, Ulrich Bick, MD, Christiane Richter- Ehrenstein, MD, Angela Reles, MD and Klaus-Jürgen Winzer for their contribution in the patient recruitment and inclusion. We are thankful to Bettina Herwig for editorial support. References [1] N. Houssami, P. Macaskill, M.L. Marinovich, J.M. Dixon, L. Irwig, M.E. Brennan, L.J. Solin, Meta-analysis of the impact of surgical margins on local recurrence in women with early-stage invasive breast cancer treated with breast-conserving therapy, Eur. J. Cancer 46 (18) (2010) 3219–3232. [2] C. Seretis, Significance of the resection margin and risk factors for close or positive resection margin in patients undergoing breast-conserving surgery (by Drs Gatek and Vrana), J. B.U.ON. 18 (3) (2013) 803–804. [3] J. Atkins, F. Al Mushawah, C.M. Appleton, A.E. Cyr, W.E. Gillanders, R.L. Aft, T.J. Eberlein, F. Gao, J.A. Margenthaler, Positive margin rates following breastconserving surgery for stage I-III breast cancer: palpable versus nonpalpable tumors, J. Surg. Res. 177 (1) (2012) 109–115. [4] J.F. Waljee, E.S. Hu, L.A. Newman, A.K. Alderman, Predictors of re-excision among women undergoing breast-conserving surgery for cancer, Ann. Surg. Oncol. 15 (5) (2008) 1297–1303. [5] F.A. Angarita, A. Nadler, S. Zerhouni, J. Escallon, Perioperative measures to optimize margin clearance in breast conserving surgery, Surg. Oncol. 23 (2) (2014) 81–91. [6] P.D. Britton, L.I. Sonoda, A.K. Yamamoto, B. Koo, E. Soh, A. Goud, Breast surgical specimen radiographs: how reliable are they? Eur. J. Radiol. 79 (2) (2011) 245–249. [7] G. Ciccarelli, M.R. Di Virgilio, S. Menna, L. Garretti, A. Ala, R. Giani, R. Bussone, G. Canavese, E. Berardengo, Radiography of the surgical specimen in early stage breast lesions: diagnostic reliability in the analysis of the resection margins, La Radiologia Medica 112 (3) (2007) 366–376. [8] N. Cabioglu, K.K. Hunt, A.A. Sahin, H.M. Kuerer, G.V. Babiera, S.E. Singletary, G.J. Whitman, M.I. Ross, F.C. Ames, B.W. Feig, T.A. Buchholz, F. Meric-Bernstam, Role for intraoperative margin assessment in patients undergoing breast-conserving