Abstract
Purpose
To compare the performance of one-view digital breast tomosynthesis (1v-DBT) to that of three other protocols combining DBT and mammography (DM) for breast cancer detection.
Materials and methods
Six radiologists, three experienced with 1v-DBT in screening, retrospectively reviewed 181 cases (76 malignant, 50 benign, 55 normal) in two sessions. First, they scored sequentially: 1v-DBT (medio-lateral oblique, MLO), 1v-DBT (MLO) + 1v-DM (cranio-caudal, CC) and two-view DM + DBT (2v-DM+2v-DBT). The second session involved only 2v-DM. Lesions were scored using BI-RADS® and level of suspiciousness (1–10). Sensitivity, specificity, receiver operating characteristic (ROC) and jack-knife alternative free-response ROC (JAFROC) were computed.
Results
On average, 1v-DBT was non-inferior to any of the other protocols in terms of JAFROC figure-of-merit, area under ROC curve, sensitivity or specificity (p>0.391). While readers inexperienced with 1v-DBT screening improved their sensitivity when adding more images (69–79 %, p=0.019), experienced readers showed similar sensitivity (76 %) and specificity (70 %) between 1v-DBT and 2v-DM+2v-DBT (p=0.482). Subanalysis by lesion type and breast density showed no difference among modalities.
Conclusion
Detection performance with 1v-DBT is not statistically inferior to 2v-DM or to 2v-DM+2v-DBT; its use as a stand-alone modality might be sufficient for readers experienced with this protocol.
Key points
• One-view breast tomosynthesis is not inferior to two-view digital mammography.
• One-view DBT is not inferior to 2-view DM plus 2-view DBT.
• Training may lead to 1v-DBT being sufficient for screening.
Similar content being viewed by others
Explore related subjects
Find the latest articles, discoveries, and news in related topics.Avoid common mistakes on your manuscript.
Introduction
Digital breast tomosynthesis (DBT) improves breast cancer detection and diagnosis compared to digital mammography (DM) [1,2,3,4,5,6,7,8,9]. Instead of resulting in a two-dimensional (2D) image of the compressed breast as in DM, DBT acquires several low-dose 2D projections over a limited angular range, which are then used to reconstruct a pseudo-3D image of the breast [10]. Therefore, DBT can ameliorate the main limitation of DM: the anatomical noise due to tissue superposition. After years of promising results pointing to its superiority to DM, DBT is now being considered to be used for population-based breast cancer screening [11]. However, there is no agreement regarding how to implement DBT in screening.
Some implementations of DBT involve its use as an adjunct to DM. The main arguments for performing DM in addition to DBT are comparisons with prior mammograms, as well as results of some early studies that suggest that DBT is inferior for calcification detection and characterization [12, 13]. However, the primary negative effects of performing both modalities are an increase in reading time [14, 15] and in radiation dose [16]. The replacement of DM with synthetic 2D mammograms generated from DBT volumes allows for comparison with priors and reduces the radiation dose, while providing similar diagnostic performance to DM [17,18,19]. With regard to calcification detection, new studies indicate that DBT is not inferior to DM even for wide-scan angle DBT systems (those acquiring the projection images over an angular range of 40–50°, more prone to blurring due to a more oblique x-ray incidence) [20]. Thus, the ongoing evolution of the technique suggests the possibility of using DBT as a stand-alone modality.
However, the vast technical differences among commercial systems may need to be considered [10]. Using a wide-scan angle DBT system yields improved depth information [21]. Therefore, these systems better exploit the 3D advantage of DBT over DM, and may make the acquisition of two views of the breast unnecessary. A large screening trial has recently shown that using only the medio-lateral oblique (MLO) view of a wide-scan angle DBT system as a one-view stand-alone technique resulted in a 43 % breast cancer detection increase compared to two-view DM [7], supporting earlier pilot studies [22, 23]. Other preliminary studies also showed that one-view DBT is not significantly different to two-view DM, but they used a subject sample that was enriched, either totally [24, 25] or partially [26], with lesions detected with standard 2v-DM, therefore only allowing for a determination of non-inferiority for DBT.
In our study, we aim to further explore the potential of one-view DBT for breast cancer detection in screening, by performing a retrospective reader study with an enriched case dataset, comparing the clinical performance of one-view DBT (1v-DBT, MLO) with three other protocols: 1v-DBT (MLO) plus 1v-DM (CC), two-view DM (2v-DM) and 2v-DM plus 2v-DBT, something not yet reported in the same study. Furthermore, we also evaluate the strengths of 1v-DBT as a standalone technique stratified by lesion type, breast density and radiologist experience in reading 1v-DBT in a screening setup, which has not yet been studied.
Methods
Study population
This retrospective study was approved by the regional ethics board after summary review, with waiver of a full review and informed consent. The study used DM and DBT images from 181 women (median age 52 years, range 30–88 y) imaged at our hospital between December 2014 and December 2015, who were recalled from screening (33 %) or had a clinical indication for imaging (67 %). All the available cancer cases within the collection date were included, while benign and normal cases were consecutively included to meet the proportions described below. Women with a prior history of breast cancer or for whom the required views for this study were unavailable were excluded (in total, 25). Within the cohort, 76 patients had malignant lesions and 50 had benign lesions, all verified by histopathology. Four patients had multiple lesions, three patients with two different malignant lesions and one patient with one benign and one malignant lesion. In total, 130 lesions were diagnosed with histopathological proof (79 malignant, 51 benign). The remaining 55 patient cases were interpreted as normal (Breast Imaging Reporting and Data System, BI-RADS®, score 1 or 2), and had at least 1 year of negative imaging follow-up (mean follow-up: 378 days). Detailed characteristics of patient cases and identified lesions are presented in Table 1. Density according to BI-RADS® 5th Edition was obtained from the radiological case report from clinical routine.
Patient images
All patients had undergone diagnostic breast imaging using a commercial DBT system (Mammomat Inspiration, Siemens, Erlangen, Germany). For each patient 2v-DM and 2v-DBT (CC and MLO) were obtained. For DBT, the system acquires 25 low-dose projection images over an angular range of approximately 50° (wide-angle breast tomosynthesis) in about 25 s [10]. The projection images are subsequently reconstructed using a filtered back-projection algorithm [27]. Four different unilateral (breast with the most suspicious findings, or randomly selected if normal) image sets resembling four different imaging protocols were created for each patient: one-view breast tomosynthesis (1v-DBT, MLO), 1v-DBT (MLO) plus 1v-DM (CC), 2v-DM and 2v-DM plus 2v-DBT.
Study design
A retrospective reader study consisting of two reading sessions (Fig. 1) was designed to evaluate the detection performance of the four protocols. The first session had a sequential design with three distinct steps per patient, where the readers were shown progressively 1v-DBT (MLO) (step 1), then 1v-DBT (MLO) plus 1v-DM (CC) (step 2) and finally 2v-DM plus 2v-DBT (step 3). The second reading session was performed at least 4 weeks (considered enough for a memory washout period) after the first session and consisted of only the 2v-DM images. A training set consisting of 20 cases was reviewed to begin each session. The readers were blinded to any information about the patient and any prior imaging.
At each step, the reader was asked to annotate and score all the detected suspicious lesions. For each abnormality, the observer gave two scores: a forced BI-RADS® assessment (1 – normal, 2 – benign findings, 3 – low probability of malignancy, 4 - suspicious of malignancy, 5 – highly suspicious of malignancy) as established by the American College of Radiology, and a level of suspiciousness score (10-point scale, from 1 – high probability of benign; to 10 – high probability of malignancy).
The experiment was performed on an in-house developed workstation (CIRRUS Observer, Diagnostic Image Analysis Group, Nijmegen, The Netherlands) using dual 5-MP or one 10-MP mammographic display(s). The workstation also automatically recorded reading times for each modality.
Reference standard
The reference standard, including location and lesion type, was established by one radiologist who did not participate in the study and had 13 years of experience in DM and 2 in DBT, with knowledge of the clinical presentation, additional imaging tests including priors and histopathology reports, when available.
Readers
Six breast radiologists performed the study. They are part of three different institutions across two countries: The Netherlands and Sweden. We recognized two distinct categorical groups of readers; the three readers from The Netherlands, who had no experience in reading 1v-DBT as a stand-alone modality for breast cancer screening, and the three readers from Sweden, who were experienced in this approach due to participation in a large 1v-DBT screening trial [7]. The three experienced readers with 1v-DBT had 3, 12 and 44 years of experience with mammography and 3, 9 and 10 years of experience with DBT. The three inexperienced readers with 1v-DBT had 17, 26 and 35 years of experience with mammography and 2, 2 and 3 years of clinical experience with DBT in combo mode (2v-DM + 2v-DBT).
Statistical analysis
Four different analyses were performed. First, for a precise evaluation, a jack-knife alternative free-response receiver operating characteristic (JAFROC) analysis was performed [28]. For this, the lesion localizations by the readers (considered correct if within 2 cm of the reference standard) and the level of suspiciousness were used. JAFROC provides a figure of merit (FoM) defined as the probability that a correctly marked lesion is rated higher than the highest-rated mark on a normal/benign case [29].
For the second analysis, receiver operating characteristic (ROC) curves and their area under the curve (AUC) were computed. Since ROC analysis requires that diagnostic confidence is expressed in an ordinal scale, level of suspiciousness and not BI-RADS® scores of the most suspicious finding per case were used [30]. ROC analysis was repeated discriminating by lesion type (soft tissue lesions or calcifications; if a lesion was composed of both types, it was counted on each category) and breast density category (low, a and b; or high, c and d).
Significance testing of ROC and JAFROC was performed using the Dorfman–Berbaum–Metz multiple reader, multiple-case mixed-model analysis of variance, which yields a p-value for rejecting the null hypothesis that the four modalities have equal performance. Random-reader and random-case analysis was performed [28, 29, 31].
To study the impact that our results would have in a screening scenario, sensitivity and specificity on a per-case basis were computed using BI-RADS® categories, using BI-RADS® category 3 or higher defined as a positive interpretation. Cases with biopsied benign lesions were considered as false positives if they were rated positive by the readers. Average sensitivity and specificity for all imaging protocols were computed using a generalized linear model (GLM) to account for multiple reader, multiple case repeated measures. Parameter estimates of the GLM were bootstrapped (n=1,000). The all two-way GLM model was built with an unstructured covariance matrix, using modality and reader as factors. To adjust for multiple comparison, the least significant difference correction was used. Model-based Wald 95 % confidence intervals were calculated. Statistical significance among modalities for each reader was estimated using McNemar’s paired test.
Finally, reading times, defined as the time spent on evaluating, scoring and annotating 1v-DBT (first step of the first reading session) and 2v-DM (second reading session), were compared using two-way repeated measures ANOVA. Outliers, defined as times whose values extended beyond 1.5 standard deviations, were removed. Also, mean glandular doses were retrieved from the DICOM (Digital Imaging and Communications in Medicine) headers for comparison. Differences between modalities for each reader were compared using a paired Student’s t-test.
A two-tailed p value lower than 0.05 was considered to indicate significant difference. All analyses were performed using SPSS (version 24, IBM Inc., Armonk, NY, USA) and open-access JAFROC software by Dev Chakraborty (version 4.2.1, DevChakraborty.com).
Results
The JAFROC curves averaged for all readers are shown in Fig. 2, while individual JAFROC FoM per reader and experience are shown in Table 2. There was no statistical difference between 1v-DBT and the other protocols (p=0.522), either on average or per reader. The FoM for 1v-DBT was similar to 2v-DM and slightly lower than for the rest, while it was comparable between experienced and inexperienced readers with 1v-DBT. Only a small difference was found in the JAFROC FoM between experienced and inexperienced observers, the latter performing slightly better with 2v-DM. Relative results did not change if the biopsied benign cases were removed from the analysis, and 1v-DBT yielded similar performance to the other three modalities (p=0.459).
JAFROC analysis averaged for all readers for each reading protocol. The lesion localization fraction is the number of correctly identified lesions divided by total number of lesions (0 ≤ LLF ≤ 1), while the non-lesion localization fraction is the number of marks which are not close to any lesions, divided by total number of images (0 ≤ NLF); note the lack of an upper bound
The average radiologist’s ROC curve is shown in Fig. 3a for each imaging protocol. The AUC of 1v-DBT was not statistically significantly different than that of the other protocols for the average of readers (p=0.391, Table 3 and Fig. 3a). Only in two cases was 1v-DBT significantly different with respect to another protocol: for one reader (experienced with 1v-DBT), the AUC was statistically better for 1v-DBT than for 2v-DM (p=0.011), while for another reader (inexperienced with 1v-DBT) 1v-DBT performed worse than 2v-DM (p=0.035). Experienced readers had only slightly higher AUCs for 1v-DBT than the inexperienced readers (Fig. 3b; experienced = 0.815 (CI: 0.760-0.871), inexperienced = 0.800 (CI: 0.746-0.855), not significant, p=0.775). On the other hand, AUC for 2v-DM was lower (p=0.425) in the experienced (0.793, CI: 716-0.871) compared to the inexperienced group (0.831, CI: 775-0.887).
Using ROC analysis, no significant difference was found between 1v-DBT and the other modalities either by separating the cases by breast density (low, p=0.601; or high, p=0.323), or by separating the lesions by type (soft tissue, p=0.329; or calcifications, p=0.499). It was also seen that 1v-DBT performs better than 2v-DM at low false-positive rates, both in ROC (Fig. 3a) and JAFROC (Fig. 2) analyses.
The sensitivity and specificity per reader and on average for each imaging protocol is shown in Table 4 and Fig. 4. No difference was found between 1v-DBT and the other modalities either for sensitivity (p=0.536) or specificity (p=0.553). There were differences in the results among the six readers (p<0.001). For the group of 1v-DBT experienced radiologists, there was no statistically significant difference either in sensitivity (p=0.776) or specificity (p=0.482) between 1v-DBT and the other protocols. For the inexperienced group, sensitivity increased for all the other protocols with respect to 1v-DBT (only significant for 2v-DM, from 69 % to 79 %, p=0.019), while specificity was slightly higher for 1v-DBT with respect to the other protocols (not significant, p=0.777).
Two examples of cases that were correctly assessed by most readers in 1v-DBT and not in 2v-DM are shown in Figs. 5 and 6, while two cases assessed correctly by most readers in 2v-DM and not in 1v-DBT are displayed in Figs. 7 and 8. Apparently, the effect DBT can have in benign lesions is bidirectional (Figs. 6, and 8), sometimes leading to increased suspiciousness and recall and sometimes reducing suspiciousness and avoiding recall.
The mean glandular dose per study was equal between 1v-DBT (2.41 ± 0.87 mGy) and 2v-DM (2.41 ± 0.83 mGy). The mean dose for 1v-DBT+1v-DM was 3.62 ± 1.25 mGy and for 2v-DBT+2v-DM was 7.23 ± 2.49 mGy. The average reading time was higher for 1v-DBT with respect to 2v-DM (55 s versus 44 s, p<0.001, see Table 5). For three readers there was no statistical difference between 1v-DBT and 2v-DM, and no difference in reading time was found between experienced and inexperienced observers. A total of 41 reading time outliers were identified among all readers (median six outliers per reader, range 2–15).
Discussion
The results of our study suggest that one-view digital breast tomosynthesis is not significantly different than two-view digital mammography and the combination of two-view mammography plus two-view tomosynthesis for breast cancer detection.
The addition of 1v-DM (CC) to 1v-DBT (MLO), one of the protocols recommended by some manufacturers, yielded an increase in sensitivity but also a small decrease in specificity for the inexperienced readers. For the readers experienced with 1v-DBT, no increase in sensitivity was found when adding 1v-DM to 1v-DBT, similar to the results by Lång et al. [7]. In general, for experienced readers, 1v-DBT proved to be enough in terms of sensitivity and specificity, and no added value was found with extra views. The results of the inexperienced 1v-DBT reader group were different. These radiologists operate at a different point along the same ROC curve as the experienced 1v-DBT reader group, either due to local screening practices or due to their not being accustomed to arriving at a decision with a single view. Overall, they had a higher specificity and lower sensitivity for 1v-DBT, that respectively decreased and increased when more images were added. The higher specificity could be explained due to having more experience in reading mammograms. However, their performance in terms of ROC was similar to that of the experienced readers, suggesting that training could lead them to operate at the same point as the more experienced readers with 1v-DBT.
These results are similar when taking lesion localization into account and computing the figure-of-merit of the JAFROC analysis. Nevertheless, as could be expected, the protocol with more images available for the radiologist, 2v-DBT plus 2v-DM, yielded a slightly better, but not significant, performance. We also saw that 1v-DBT performs better than 2v-DM at low false positives, which could be particularly relevant and important for screening.
The experience level with mammography might have also played a role in our results. As suggested by some studies, the least experienced readers with mammography benefit the most from using DBT [26, 32]. In our case, we observed that the less experienced readers with mammography had a lower ROC performance with 2v-DM than the others, but similar performance with 1v-DBT.
When looking at different lesion types, we found that 1v-DBT is not statistically inferior to any other tested protocol for the task of detecting lesions with calcifications, which adds to the results by other authors [33, 26, 20] and suggests that DBT is not inferior for the detection of calcifications even with a wide-angle DBT system.
All these results suggest that the use of 1v-DBT as a stand-alone modality for breast cancer screening may be feasible, since the added value of the other DBT view or any DM views was not found significant in this study. Aside from the Malmö Breast Tomosynthesis Screening Trial, which was performed with 1v-DBT [7], most screening trials with DBT have used a protocol consisting of 2v-DBT with narrow-angle systems [3, 17, 18, 34, 19]. All studies report equivalent increases in breast cancer detection rates, and similar recall rates [11]. Screening is different from clinical practice. A mass screening policy always implies compromises due to constraints of costs, staffing, radiation dose to the population and other factors. In clinical practice, such constraints are less of an issue.
The Malmö Study showed a 43 % increase in cancer detection rate with 1v-DBT compared with 2v-DM. Clearly, adding a CC-view in DBT would increase the cancer detection rate marginally, just as e.g. adding breast ultrasound examination would do, something that is usually considered not feasible except in high risk groups. In the future, screening is in all likelihood going to be individualized based on risk profile. High-risk groups will probably be offered something increasing the sensitivity which may be another DBT view, ultrasound or even MRI, the latter already being the case in many programs for women with the highest risk. Nevertheless, in a screening scenario, we assume that if 2v-DBT is not feasible due to implementation reasons, there is an overall benefit in detection achieved by performing 1v-DBT instead of two-view DM. Finally, it is yet to be seen if 1v-DBT with a narrow-angle DBT system yields at least the same performance as 2v-DM for breast cancer detection, something not assessed in this work due to its single manufacturer limitation.
The increased reading time for DBT in comparison to DM is still one of the pitfalls that can be improved before implementing DBT in a screening setup [15, 35]. Certainly, using one-view instead of two-view DBT, without losing clinical performance, could ameliorate this problem. We observed that 1v-DBT took on average 25 % longer to read than 2v-DM (although for half of the observers reading times were equivalent). Yet, it is also possible that longer loading times of DBT in comparison with DM influenced reading times in our study. Additional training of radiologists on reading DBT, the inclusion of synthetic mammograms and computer aided detection systems might aid speeding the reading of DBT.
The main limitation of our study is the fact that around 50 % of the positive cases in our dataset are recalls from the Dutch DM-based screening program. Therefore, these were lesions already seen in 2v-DM. The lack of a true DBT screening population in our study, or at least enrichment with lesions first detected with either modality, leads to a bias towards 2v-DM, and thus the true benefit of 1v-DBT in screening might be larger than documented in our study. The real sensitivity and specificity of 1v-DBT in screening practice can only be assessed in a screening study, but our study in contrast allows determination of the relative differences in reading mode. We included all the cancer cases available at our institution, but the study could have benefitted from additional cases detected by DBT screening programs. Also, there were two different sets of radiologists involved in the readings, who might have different operating points based on local routine. Another minor limitation may be the stepwise nature of the first reading session, rather than dividing into three different sessions. However, we believe the stepwise scheme could introduce a bias, if at all, against 1v-DBT, which does not affect the conclusion of this work.
Conclusion
Detection performance with 1v-DBT is not statistically inferior to the standard protocols of 2v-DM and 2v-DM+2v-DBT, and its use as a stand-alone modality might be sufficient for readers experienced with this protocol. Based upon the overall equivalent performance in terms of ROC and JAFROC analysis, experience with single-view DBT interpretation might change the operating point of radiologists, making their sensitivity/specificity performance in a screening scenario equivalent to that of two-view DM plus two-view DBT. Therefore, with a wide-angle system and appropriate training, MLO view-only DBT might be feasible for breast cancer screening.
Abbreviations
- 1v-DBT:
-
One-view digital breast tomosynthesis (MLO)
- 1v-DBT+1v-DM:
-
One-view digital breast tomosynthesis (MLO) plus one-view digital mammography (CC)
- 2v-DM:
-
Two-view digital mammography (CC + MLO)
- 2v-DBT+2v-DM:
-
Two-view digital breast tomosynthesis (CC + MLO) plus two-view digital mammography (CC + MLO)
- ANOVA:
-
Analysis of variance
- AUC:
-
Area under the curve
- BI-RADS:
-
Breast imaging reporting and data system
- CC:
-
Cranio caudal
- DBT:
-
Digital breast tomosynthesis
- DICOM:
-
Digital imaging and communications in medicine
- DM:
-
Digital mammography
- FOM:
-
Figure of merit
- JAFROC:
-
Jack-knife alternative free-response receiver operating characteristic
- MLO:
-
Medio-lateral oblique
- ROC:
-
Receiver operating characteristic
References
Andersson I, Ikeda DM, Zackrisson S et al (2008) Breast tomosynthesis and digital mammography: a comparison of breast cancer visibility and BIRADS classification in a population of cancers with subtle mammographic findings. Eur Radiol 18(12):2817–2825
Skaane P, Gullien R, Bjorndal H et al (2012) Digital breast tomosynthesis (DBT): initial experience in a clinical setting. Acta Radiol 53(5):524–529
Ciatto S, Houssami N, Bernardi D et al (2013) Integration of 3D digital mammography with tomosynthesis for population breast-cancer screening (STORM): a prospective comparison study. Lancet Oncol 14(7):583–589
Skaane P, Bandos AI, Gullien R et al (2013) Comparison of digital mammography alone and digital mammography plus tomosynthesis in a population-based screening program. Radiology 267(1):47–56
Friedewald SM, Rafferty EA, Rose SL et al (2014) Breast cancer screening using tomosynthesis in combination with digital mammography. Jama 311(24):2499–2507
Lourenco AP, Barry-Brooks M, Baird GL, Tuttle A, Mainiero MB (2015) Changes in recall type and patient treatment following implementation of screening digital breast tomosynthesis. Radiology 274(2):337–342
Lång K, Andersson I, Rosso A, Tingberg A, Timberg P, Zackrisson S (2016) Performance of one-view breast tomosynthesis as a stand-alone breast cancer screening modality: results from the Malmo Breast Tomosynthesis Screening Trial, a population-based study. Eur Radiol 26(1):184–190
McDonald ES, Oustimov A, Weinstein SP, Synnestvedt MB, Schnall M, Conant EF (2016) Effectiveness of Digital Breast Tomosynthesis Compared With Digital Mammography: Outcomes Analysis From 3 Years of Breast Cancer Screening. JAMA Oncol 2(6):737–743
Rafferty EA, Durand MA, Conant EF et al (2016) Breast Cancer Screening Using Tomosynthesis and Digital Mammography in Dense and Nondense Breasts. Jama 315(16):1784–1786
Sechopoulos I (2013) A review of breast tomosynthesis. Part I. The image acquisition process. Medical physics 40(1):014301
Skaane P (2017) Breast cancer screening with digital breast tomosynthesis. Breast Cancer 24(1):32–41
Spangler ML, Zuley ML, Sumkin JH et al (2011) Detection and classification of calcifications on digital breast tomosynthesis and 2D digital mammography: a comparison. AJR Am J Roentgenol 196(2):320–324
Tagliafico A, Houssami N (2015) Digital breast tomosynthesis might not be the optimal modality for detecting microcalcification. Radiology 275(2):618–619
Zuley ML, Bandos AI, Abrams GS et al (2010) Time to diagnosis and performance levels during repeat interpretations of digital breast tomosynthesis: preliminary observations. Acad Radiol 17(4):450–455
Dang PA, Freer PE, Humphrey KL, Halpern EF, Rafferty EA (2014) Addition of tomosynthesis to conventional digital mammography: effect on image interpretation time of screening examinations. Radiology 270(1):49–56
Svahn TM, Houssami N, Sechopoulos I, Mattsson S (2015) Review of radiation dose estimates in digital breast tomosynthesis relative to those in two-view full-field digital mammography. Breast 24(2):93–99
Skaane P, Bandos AI, Eben EB et al (2014) Two-view digital breast tomosynthesis screening with synthetically reconstructed projection images: comparison with digital breast tomosynthesis with full-field digital mammographic images. Radiology 271(3):655–663
Zuley ML, Guo B, Catullo VJ et al (2014) Comparison of two-dimensional synthesized mammograms versus original digital mammograms alone and in combination with tomosynthesis images. Radiology 271(3):664–671
Zuckerman SP, Conant EF, Keller BM et al (2016) Implementation of Synthesized Two-dimensional Mammography in a Population-based Digital Breast Tomosynthesis Screening Program. Radiology 281(3):730–736
Clauser P, Nagl G, Helbich TH et al (2016) Diagnostic performance of digital breast tomosynthesis with a wide scan angle compared to full-field digital mammography for the detection and characterization of microcalcifications. Eur J Radiol 85(12):2161–2168
Rodriguez-Ruiz A, Castillo M, Garayoa J, Chevalier M (2016) Evaluation of the technical performance of three different commercial digital breast tomosynthesis systems in the clinical environment. Phys Med 32(6):767–777
Svahn T, Andersson I, Chakraborty D et al (2010) The diagnostic accuracy of dual-view digital mammography, single-view breast tomosynthesis and a dual-view combination of breast tomosynthesis and digital mammography in a free-response observer performance study. Radiat Prot. Dosim 139(1-3):113–117
Svahn TM, Chakraborty DP, Ikeda D et al (2012) Breast tomosynthesis and digital mammography: a comparison of diagnostic accuracy. Br J Radiol 85(1019):e1074–e1082
Gennaro G, Toledano A, di Maggio C et al (2010) Digital breast tomosynthesis versus digital mammography: a clinical performance study. Eur Radiol 20(7):1545–1553
Svane G, Azavedo E, Lindman K et al (2011) Clinical experience of photon counting breast tomosynthesis: comparison with traditional mammography. Acta Radiol 52(2):134–142
Wallis MG, Moa E, Zanca F, Leifland K, Danielsson M (2012) Two-view and single-view tomosynthesis versus full-field digital mammography: high-resolution X-ray imaging observer study. Radiology 262(3):788–796
Mertelmeier T, Orman J, Haerer W, Dudam MK (2006) Optimizing filtered backprojection reconstruction for a breast tomosynthesis prototype device(ed)^(eds) Medical Imaging. International Society for Optics and Photonics, pp 61420F-61420F-61412
Chakraborty DP (2005) Recent advances in observer performance methodology: jackknife free-response ROC (JAFROC). Radiat Prot Dosim 114(1-3):26–31
Chakraborty DP (2008) Validation and statistical power comparison of methods for analyzing free-response observer performance studies. Acad Radiol 15(12):1554–1566
Jiang Y, Metz CE (2010) BI-RADS data should not be used to estimate ROC curves. Radiology 256(1):29–31
Chakraborty DP (2011) New developments in observer performance methodology in medical imaging. Semin Nucl Med 41(6):401–418
Thomassin-Naggara I, Balvay D, Rockall A et al (2015) Added Value of Assessing Adnexal Masses with Advanced MRI Techniques. Biomed Res Int 2015:785206
Kopans D, Gavenonis S, Halpern E, Moore R (2011) Calcifications in the breast and digital breast tomosynthesis. Breast J 17(6):638–644
Gilbert FJ, Tucker L, Gillan MG, et al. (2015) The TOMMY trial: a comparison of TOMosynthesis with digital MammographY in the UK NHS Breast Screening Programme--a multicentre retrospective reading study comparing the diagnostic performance of digital breast tomosynthesis and digital mammography with digital mammography alone. Health Technol Assess 19(4):i-xxv, 1-136
Bernardi D, Ciatto S, Pellegrini M et al (2012) Application of breast tomosynthesis in screening: incremental effect on mammography acquisition and reading time. Br J Radiol 85(1020):e1174–e1178
Acknowledgements
The authors would like to thank Suzan Vreemann for valuable discussion about the statistical methodology.
Funding
This study has received funding by Siemens Healthineers (Erlangen, Germany).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Guarantor
The scientific guarantor of this publication is Ioannis Sechopoulos.
Conflict of interest
The authors of this manuscript declare relationships with the following companies: Siemens Healthineers (Erlangen, Germany).
Statistics and biometry
No complex statistical methods were necessary for this paper.
Informed consent
Written informed consent was not required for this study because this requirement is waived by the Dutch review board for retrospective research on patient files.
Ethical approval
Institutional Review Board approval was not required because this requirement is waived by the Dutch review board for retrospective research on patient files.
Methodology
• retrospective
• experimental
• multicenter study
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Rodriguez-Ruiz, A., Gubern-Merida, A., Imhof-Tas, M. et al. One-view digital breast tomosynthesis as a stand-alone modality for breast cancer detection: do we need more?. Eur Radiol 28, 1938–1948 (2018). https://doi.org/10.1007/s00330-017-5167-3
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00330-017-5167-3