Digital Breast Tomosynthesis: Concepts and Clinical Practice

Abstract

Digital breast tomosynthesis (DBT) is emerging as the standard of care for breast imaging based on improvements in both screening and diagnostic imaging outcomes. The additional information obtained from the tomosynthesis acquisition decreases the confounding effect of overlapping tissue, allowing for improved lesion detection, characterization, and localization. In addition, the quasi three-dimensional information obtained from the reconstructed DBT data set allows a more efficient imaging work-up than imaging with two-dimensional full-field digital mammography alone. Herein, the authors review the benefits of DBT imaging in screening and diagnostic breast imaging.

© RSNA, 2019

Summary

This article highlights the current concepts of digital breast tomosynthesis imaging, including synthetic mammography, that are relevant to contemporary breast imaging; the quasi three-dimensional information and the improved lesion conspicuity gained with tomosynthesis are associated with improvements in patient outcomes and efficiency in mammographic imaging.

Essentials

• ■ Multiple studies have shown improved screening outcomes with digital breast tomosynthesis, including lower recall and higher cancer detection rates.

• ■ The additional invasive cancers detected with digital breast tomosynthesis tend to be smaller, lower grade, and have a more favorable prognosis.

• ■ The use of synthetic mammography in place of full-field digital mammography with tomosynthesis provides similar screening outcomes while reducing radiation dose.

• ■ While lesion localization is improved with tomosynthesis, it is important to understand limitations including that the nipple axis of the breast is often not in the center of the breast and therefore is not in the center of the scrollbar and, in some reconstruction algorithms, additional sections are added to the stack to account for bending of the breast compression paddle.

• ■ Due to improvements in lesion detection, characterization, and localization with digital breast tomosynthesis, workflow efficiency may be improved compared with imaging with digital mammography alone.

Introduction

Digital breast tomosynthesis (DBT) involves multiple projections acquired across an arc that are reconstructed into a series of stacked images (1). Depending on the manufacturer, during DBT image acquisition the x-ray tube pivots in an arc that varies between 15° (narrow range) and 60° (wide range) in a plane aligned with the chest wall. In general, the larger angular range of the x-ray tube motion results in more tomographic information and yields better section separation or vertical (z-axis) resolution. An increase in the angular range for the tube motion requires an increase in the number of projections for sufficient sampling (2–4). Regardless of the range of the x-ray tube, the stacked images provide localization information and improved lesion characterization, potentially decreasing or eliminating the need for additional diagnostic work-up (5–11). Although DBT has been performed by using a combination of tomosynthesis imaging plus full-field digital mammography (FFDM), DBT imaging is associated with increased acquisition time and requires a longer interpretation time owing to the larger image set. Concerns over increased radiation dose have prompted the development of synthetic mammography (SM), in which two-dimensional images are reconstructed from the tomosynthesis data set to replace the FFDM portion of the examination (12–17). With DBT becoming the standard of care for both screening and diagnostic breast imaging, herein we review the current literature on DBT, including screening outcomes, tips for lesion localization and characterization, changing workflows for diagnostic imaging, and future developments to improve both efficiency and accuracy in DBT imaging.

Screening Outcomes

Cancer Detection Rate

Compared with FFDM alone, breast cancer screening with DBT plus FFDM is associated with increases in the cancer detection rate (CDR) ranging from 1.2 to 4.6 per 1000 examinations (13,18–28) (Table 1). In prospective European studies, the CDR increased from 27% to 91% with the addition of DBT (18–22,28). In large multisite observational studies from the United States, Friedewald et al (25) demonstrated an increased overall CDR of 29%, and the Population-based Research Optimizing Screening Through Personalized Regimen (PROSPR) consortium (13) showed a 34% increase. There was an increase in cancer detection in both dense and fatty breasts due to improved lesion conspicuity across all breast densities and the “unmasking” of cancers, particularly in dense breasts, with DBT compared with FFDM alone (7,13,18–20,28,34).

Table 1:

Screening Outcomes: Comparison of FFDM Alone and FFDM plus DBT

Note.—CDR = cancer detection rate, DBT = digital breast tomosynthesis, FFDM = full- field digital mammography, PPV₁ = positive predictive value for recall from screening, SM = synthetic mammography. Interval cancer rates were given for two studies and were 1.6 and 1.2 per 1000 screens for digital mammography and DBT, respectively (not significant), for the Screening with Tomosynthesis or Standard Mammography study (12,19) and 0.46 and 0.6 per 1000 screens (not significant) for the study by Conant et al (13).

*Numbers in parentheses are the percentage change.

^†Estimated reduction in false-positive findings.

^‡Study involved one-view DBT and two-view FFDM.

^§SM plus DBT compared with FFDM.

^||Not statistically significant.

**After adjusting for center, age, breast density, and first examination, the odds of biopsy were statistically lower for DBT than FFDM (odds ratio = 0.85; 95% confidence interval = 0.77, 0.93).

Some studies (18,25,26) have shown that the increase in screening-detected cancers achieved with DBT is mostly due to an increase in the detection of invasive cancers rather than ductal carcinoma in situ, which, when lower grade, is sometimes considered as a harm of screening or “overdiagnosis” (35). The detection of additional invasive cancers with DBT has the potential to impact not only the morbidity associated with breast cancer (less aggressive treatment may be possible when cancer is detected at a smaller size and earlier stage) but also long-term breast cancer survival. However, breast cancer survival benefits are difficult to prove without extended follow-up and/or randomized controlled trials.

To further understand the impact of the increased detection of invasive cancers with DBT, data evaluating the cancer subtypes detected with DBT versus FFDM screening is evolving. The Verona study (29), which included 315 cancers, demonstrated that among invasive cancers detected with DBT, there was a higher proportion of cancer with histologic characteristics generally associated with a good prognosis (ie, tubular, papillary, and mucinous subtypes) compared with those detected with FFDM alone. In the Oslo trial (36), there were no significant differences in type, grade, and nodal status; however, the cancers detected only with DBT (n = 52) tended to have low Ki67 scores, which is generally associated with better prognosis and overall survival. In a retrospective study of 261 invasive cancers, Kim et al (37) found a significant association between DBT-only detected cancers and dense breasts (P = .007), small tumor size (≤2 cm; P = .027), and luminal A-like subtype (P = .008), suggesting that DBT screening has benefits in detecting less aggressive subtypes of invasive cancers in women with dense breasts compared with FFDM-only screening. The multisite data from the PROSPR consortium (38) also showed that screening with DBT was associated with the detection of smaller, more often node-negative, human epidermal growth factor 2–negative invasive cancers compared with digital mammography (73 of 99 women [73.7%] vs 276 of 422 women [65.4%], respectively). This shift toward the detection of cancers with a “better prognosis” was largest in the subgroup of women aged 40–49 years, where DBT was associated with seven of 28 cancers (25.0%) categorized as having a poor prognosis, compared with 19 of 47 breast cancers (40.4%) for FFDM.

Although DBT detection of such “better prognosis” cancers at a smaller size may offer women more surgical and treatment choices, it is unclear whether the cancers might not have been detected with FFDM screening a year or two later—without a long-term impact on patient survival (29). Longer-term follow-up of DBT screening is required.

As a complement to observational trials on DBT outcomes (which may be biased by underlying risk profiles of patients and variable utilization of DBT), the Tomosynthesis Mammographic Screening Trial will compare screening outcomes of approximately 165 000 women aged 45–75 years who are randomly assigned to either FFDM or DBT screening during a period of 4.5 years (39,40). The primary aim of the trial, which began accruing patients in July 2017, is to compare the number of advanced cancers in each arm of the trial, with the hypothesis that DBT screening will eventually decrease the number of advanced cancers compared with FFDM screening (40).

Recall Rate

Most prospective and retrospective DBT screening studies have demonstrated lower recall rates and, therefore, improved specificity for DBT plus FFDM compared with FFDM alone (12,13,18–28,30–33) (Table 1). In the retrospective observational U.S. studies, where recall rates are historically much higher than those in Europe, the decreases in recall rate range from 15% to 63% (13,23–27,30–33). In a retrospective analysis of performance metrics from 13 academic and nonacademic breast centers in the United States (n = 454 850), Friedewald et al (25) demonstrated a significant 15% decrease in recall rate with DBT. The observational trial from three U.S. research centers within the PROSPR consortium (n = 142 883) demonstrated a significant 16% decrease in recall rate (13). In most prospective European studies, recall rates have also been shown to decrease; however, the overall reduction is less than that shown in the observational U.S. studies. Of note is one prospective European study where the recall rate with DBT was higher than that in the FFDM arm (20). In that study, one-view DBT was performed—the recall rate was 2.6% with FFDM and 3.8% with DBT; however, both FFDM and DBT rates were very low compared with U.S. standards (20).

When screening recall is evaluated at the lesion level, studies have shown that DBT is associated with a lower recall of asymmetries and focal asymmetries, presumably because the multiple sections in the DBT data set allow the resolution of areas of normal, superimposed tissue that may present as summations or higher density areas at FFDM (19,31,33). Conversely, screening with DBT is associated with a higher recall of masses and distortions, secondary to the “unmasking” effect of both benign and malignant lesions with DBT (33,41,42).

False-Negative and Interval Cancer Rates

False-negative screening studies are those with a negative interpretation and subsequent breast cancer diagnosis before the next routine screening study (43,44). False-negative findings may occur as interval cancers, where a cancer manifests symptomatically (ie, as a breast lump or nipple discharge) or asymptomatically, when a cancer is detected with another imaging modality (eg, US or MRI) (43,44). Cancers diagnosed after negative findings at screening mammography are, in general, associated with less-favorable tumor characteristics and, therefore, greater patient morbidity and mortality compared with screening-detected cancers (45).

A few early, small studies with patient-level data have shown a nonsignificant trend toward decreases in false-negative findings with DBT imaging (12,13,46). The Screening with Tomosynthesis or Standard Mammography trial (n = 7292) (12), which included nine interval cancers (eight invasive cancers and one ductal carcinoma in situ in a patient with Paget disease), estimated an interval cancer rate of 1.23 per 1000 screened with DBT, compared with 1.6 per 1000 screened with FFDM. However, the retrospective study by Bahl et al (47), which included 167 interval cancers (n = 155 281), demonstrated similar interval cancer rates for DBT and FFDM (1.1 vs 1.1 per 1000 examinations; P = .84). The prospective Oslo Tomosynthesis Screening Trial (36) found that the interval cancer rate per 1000 screens for DBT plus FFDM remained similar to previous FFDM-only rounds (2.1 [51 of 24301] vs 2.0 [118 of 59877], respectively; P = .734). The characteristics of the interval cancers (ie, proportion of invasive, size, grade, and nodal status) and the interval cancer rate with DBT screening were similar to those in previous FFDM-only rounds (36). The cancers detected only with DBT (n = 52) were mostly small, low-grade, node-negative invasive cancers with low Ki67 expression, which is known to have a good prognosis (36). Larger studies with extended follow-up are needed to understand the long-term role of DBT in decreasing false-negative studies, rather than just identifying cancers earlier and cancers with a better prognosis (36,48).

Because most of the published studies on DBT were performed in first or prevalent-round screening (incident or subsequent round screening), one might expect improved cancer detection based purely on improvement in lesion conspicuity with DBT (19,20,25,28,29,49,50). To understand whether DBT screening outcomes are sustainable (and therefore may impact longer term outcomes), data on subsequent screening rounds are needed. McDonald et al (14) evaluated 44 468 DBT examinations over 4 consecutive years and assessed screening outcomes at both the population and patient levels in first and subsequent rounds by comparing recall, cancer detection, and false-negative rates with those from previous FFDM-only screening. At the population level, over years 1–3 of DBT, screening recall rates remained lower than that with FFDM (rates of 88, 90, and 92 per 1000 screened with DBT at years 1, 2, and 3, respectively, compared with 104 with FFDM) and CDRs continued to increase compared with FFDM-alone screening (rates of 5.5, 5.8, and 6.1 per 1000 screened with DBT at years 1, 2, and 3, respectively, compared with 4.6 for FFDM). Although the decrease in the recall rate with DBT was statistically significant, the increase in the CDR was not (14). The positive predictive value of positive screening results also increased each year of DBT screening and was statistically significant for DBT years 2 and 3 compared with FFDM alone. To assess the effects of the prevalence or first round of DBT screening, women who underwent only one DBT with previous FFDM screening were compared with those who underwent two and three DBT screenings (incidence DBT screenings) (14). Recall rates in women who underwent two and three DBT screenings continued to drop with each round of DBT screening, as expected. The CDR decreased at the second round of DBT screening (6.2 per 1000 screened) compared with the first DBT round (13 per 1000 screened), but increased again to 7.3 per 1000 screened in women with a third round of DBT screening (14).This fluctuation in CDR may suggest a prevalence round screening effect due to increased cancer detection with the new tomosynthesis modality, with a decrease in CDR at the second, or incidence, round. However, the lower CDR for women who underwent two screenings (6.2 per 1000 screened) is still higher than published data for incidence screening round with FFDM (4.6 per 1000 women) (14,51). There was also a trend toward a lower rate of interval cancers, or symptomatic false-negative findings, with DBT compared with FFDM-alone screening (FFDM rate of 0.7 per 1000 women screened vs 0.5 per 1000 screened in DBT year 1); however, this was not statistically significant (14).

Advantage of DBT in Lesion Evaluation

Lesion Localization

In DBT, the quasi–three-dimensional information gained from the reconstructed image stacks or “sections” allows for improved localization of breast lesions. Superficial lesions such as skin calcifications or sebaceous cysts may be easily localized without the use of tangential views often used in FFDM to show the lesion within the skin layers (52,53). Lesions within the breast may be localized by finding the lesion within the reconstructed stack and using the location of the lesion relative to the DBT plane where the nipple is in focus (54,55). This is particularly helpful in nonpalpable lesions that are better seen or only seen on one view (Fig 1) (56–58). In many cases, the traditional mediolateral view is unnecessary for triangulation, and the patient may proceed directly for US evaluation (Fig 2) (5,6).

Figure 1a:

Images in 57-year-old woman with cancer detected only with digital breast tomosynthesis (DBT). (a–c) Craniocaudal DBT image (a) demonstrates a spiculated mass (arrow) not visible on synthetic craniocaudal (b) or mediolateral oblique (c) mammograms. (d) On the basis of the location in the DBT craniocaudal stack, the mass (arrow) can be localized on DBT image obtained in mediolateral oblique projection. (e) The mass (arrow) is better demonstrated on DBT images in mediolateral projection. (f) US examination shows that finding corresponds to an irregular, hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor negative. F = foot, H = head, L = lateral, M = medial.

Figure 2a:

Images in 78-year-old woman with mass in right breast. (a, b) Images from synthetic mammography (SM) show lesion in inner breast on craniocaudal view (arrow in a) that is not well seen on mediolateral view (b). (c) The position of the lesion (arrow) on the DBT craniocaudal stack helped localize the mass (arrow) to the lower inner part of the breast. (d) With this information, the mass (arrow) is now detected on DBT image in mediolateral oblique view. (e) The patient proceeded directly to US examination, which showed a suspicious irregular hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor 2 positive. F = foot, H = head, L = lateral, M = medial.

Figure 1b:

Images in 57-year-old woman with cancer detected only with digital breast tomosynthesis (DBT). (a–c) Craniocaudal DBT image (a) demonstrates a spiculated mass (arrow) not visible on synthetic craniocaudal (b) or mediolateral oblique (c) mammograms. (d) On the basis of the location in the DBT craniocaudal stack, the mass (arrow) can be localized on DBT image obtained in mediolateral oblique projection. (e) The mass (arrow) is better demonstrated on DBT images in mediolateral projection. (f) US examination shows that finding corresponds to an irregular, hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor negative. F = foot, H = head, L = lateral, M = medial.

Figure 1c:

Images in 57-year-old woman with cancer detected only with digital breast tomosynthesis (DBT). (a–c) Craniocaudal DBT image (a) demonstrates a spiculated mass (arrow) not visible on synthetic craniocaudal (b) or mediolateral oblique (c) mammograms. (d) On the basis of the location in the DBT craniocaudal stack, the mass (arrow) can be localized on DBT image obtained in mediolateral oblique projection. (e) The mass (arrow) is better demonstrated on DBT images in mediolateral projection. (f) US examination shows that finding corresponds to an irregular, hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor negative. F = foot, H = head, L = lateral, M = medial.

Figure 1d:

Images in 57-year-old woman with cancer detected only with digital breast tomosynthesis (DBT). (a–c) Craniocaudal DBT image (a) demonstrates a spiculated mass (arrow) not visible on synthetic craniocaudal (b) or mediolateral oblique (c) mammograms. (d) On the basis of the location in the DBT craniocaudal stack, the mass (arrow) can be localized on DBT image obtained in mediolateral oblique projection. (e) The mass (arrow) is better demonstrated on DBT images in mediolateral projection. (f) US examination shows that finding corresponds to an irregular, hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor negative. F = foot, H = head, L = lateral, M = medial.

Figure 1e:

Images in 57-year-old woman with cancer detected only with digital breast tomosynthesis (DBT). (a–c) Craniocaudal DBT image (a) demonstrates a spiculated mass (arrow) not visible on synthetic craniocaudal (b) or mediolateral oblique (c) mammograms. (d) On the basis of the location in the DBT craniocaudal stack, the mass (arrow) can be localized on DBT image obtained in mediolateral oblique projection. (e) The mass (arrow) is better demonstrated on DBT images in mediolateral projection. (f) US examination shows that finding corresponds to an irregular, hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor negative. F = foot, H = head, L = lateral, M = medial.

Figure 1f:

Images in 57-year-old woman with cancer detected only with digital breast tomosynthesis (DBT). (a–c) Craniocaudal DBT image (a) demonstrates a spiculated mass (arrow) not visible on synthetic craniocaudal (b) or mediolateral oblique (c) mammograms. (d) On the basis of the location in the DBT craniocaudal stack, the mass (arrow) can be localized on DBT image obtained in mediolateral oblique projection. (e) The mass (arrow) is better demonstrated on DBT images in mediolateral projection. (f) US examination shows that finding corresponds to an irregular, hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor negative. F = foot, H = head, L = lateral, M = medial.

Figure 2b:

Images in 78-year-old woman with mass in right breast. (a, b) Images from synthetic mammography (SM) show lesion in inner breast on craniocaudal view (arrow in a) that is not well seen on mediolateral view (b). (c) The position of the lesion (arrow) on the DBT craniocaudal stack helped localize the mass (arrow) to the lower inner part of the breast. (d) With this information, the mass (arrow) is now detected on DBT image in mediolateral oblique view. (e) The patient proceeded directly to US examination, which showed a suspicious irregular hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor 2 positive. F = foot, H = head, L = lateral, M = medial.

Figure 2c:

Images in 78-year-old woman with mass in right breast. (a, b) Images from synthetic mammography (SM) show lesion in inner breast on craniocaudal view (arrow in a) that is not well seen on mediolateral view (b). (c) The position of the lesion (arrow) on the DBT craniocaudal stack helped localize the mass (arrow) to the lower inner part of the breast. (d) With this information, the mass (arrow) is now detected on DBT image in mediolateral oblique view. (e) The patient proceeded directly to US examination, which showed a suspicious irregular hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor 2 positive. F = foot, H = head, L = lateral, M = medial.

Figure 2d:

Images in 78-year-old woman with mass in right breast. (a, b) Images from synthetic mammography (SM) show lesion in inner breast on craniocaudal view (arrow in a) that is not well seen on mediolateral view (b). (c) The position of the lesion (arrow) on the DBT craniocaudal stack helped localize the mass (arrow) to the lower inner part of the breast. (d) With this information, the mass (arrow) is now detected on DBT image in mediolateral oblique view. (e) The patient proceeded directly to US examination, which showed a suspicious irregular hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor 2 positive. F = foot, H = head, L = lateral, M = medial.

Figure 2e:

Images in 78-year-old woman with mass in right breast. (a, b) Images from synthetic mammography (SM) show lesion in inner breast on craniocaudal view (arrow in a) that is not well seen on mediolateral view (b). (c) The position of the lesion (arrow) on the DBT craniocaudal stack helped localize the mass (arrow) to the lower inner part of the breast. (d) With this information, the mass (arrow) is now detected on DBT image in mediolateral oblique view. (e) The patient proceeded directly to US examination, which showed a suspicious irregular hypoechoic mass. Pathologic examination of biopsy specimen revealed invasive ductal carcinoma that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor 2 positive. F = foot, H = head, L = lateral, M = medial.

Although DBT may allow for localization of lesions based on a single DBT view, there are some known pitfalls of using the scroll or localizer bar (55). Identifying the location of a lesion relative to the nipple is an important step in lesion triangulation. In FFDM as well as DBT imaging, the nipple axis is often eccentric and not in the precise center of the breast; therefore, one must appreciate that the nipple will not always localize to the center of the reconstructed DBT image stack on the “scroll bar” (55). In addition, the nipple may be rolled inferiorly or, less often, superiorly on compression in the craniocaudal view or may roll medially or laterally in the mediolateral oblique view. When scrolling through the DBT reconstructed images, the location of the nipple within the DBT image stack should be noted as the first step in lesion triangulation. The clock face localization and depth of a lesion, needed for targeted US or physical examination, will be dependent on the location of the lesion relative to the nipple axis. Therefore, finding the level of the nipple on the scroll bar is crucial. Finally, in larger breasts, very superficial lesions may be more susceptible to even more rolling of the breast and nipple, a concept that is also important in the two-dimensional imaging localization of lesions (55).

In some reconstruction algorithms, additional reconstructed sections are added on the compression paddle side of the DBT stack to offset thickness changes of the breast due to bending of the compression paddle. Usually, there is a slight upward flexing of the compression paddle at the posterior edge of the breasts relative to the detector because the posterior part of the breast and pectoral muscles are often thicker than the anterior part of the breast. For this reason, skin calcifications on the superior (on the craniocaudal view) or medial (on the mediolateral oblique view) skin surfaces, where the compression paddle is, may appear deeper than the expected skin location on the scroll bar. Therefore, to best demonstrate the localization of skin calcifications with DBT, the breast should be positioned with the skin surface of interest closest to the x-ray detector, rather than on the bending paddle side (55).

Lesion Conspicuity

Among the commercially available DBT systems, there is variability in the arc that the x-ray tube pivots across, usually between 15° and 60°. The wider angular range allows a thinner reconstructed section thickness of the in-focus plane (and thus provides superior separation of reconstructed sections) because objects in the different planes are less blurred on images acquired at a smaller angle. Therefore, they may be small variabilities in the conspicuity of the lesion types described in the next sections (3).

Calcifications

When suspicious calcifications are detected with either DBT plus FFDM or DBT plus SM, magnification imaging with FFDM is still necessary to further characterize the calcifications with higher spatial resolution imaging (59,60). In DBT, the resolution of each individual calcification may be compromised depending on the optical density of the calcification, the DBT imaging acquisition, and motion during the longer DBT image acquisition (59,60). However, with SM, the processing algorithm used to reconstruct the two-dimensional–like synthetic image may increase the conspicuity of high-contrast features, including calcifications. In a retrospective reader study of 198 patients, Choi et al (61) evaluated calcification conspicuity and found no significant difference between SM and FFDM alone or with DBT, suggesting that SM with DBT may be sufficient in diagnosing calcifications without FFDM. Lai et al performed a retrospective, multireader observer study of 72 consecutive screening mammograms recalled for microcalcifications (62) and found similar sensitivity and specificity between SM plus DBT and FFDM alone in the detection of microcalcifications. If imaging with DBT plus SM is performed without FFDM for screening, it is important to assess for motion during the DBT acquisition because the SM reconstruction is dependent on the sharpness of features at DBT imaging; in cases where calcifications are seen on one SM view (either the craniocaudal or mediolateral oblique view) and not on the other, motion may have occurred during the acquisition of the DBT images (63). In addition, “pseudocalcifications” may be detected on images from SM as an algorithm-attenuating speckle or quantum noise on an image, as well as structural noise from overlapping structures such as Cooper ligaments that are enhanced at the SM reconstruction (64). Careful scrolling through the DBT sections in the projection of concern, combined with careful searching in the other DBT projection, is recommended. Although pseudocalcifications should not be visible on the DBT stack or source images, true calcifications will come into plane on one or more of the DBT sections.

Architectural Distortion

DBT has been shown to improve the detection of mammographic architectural distortion, the most common missed abnormality in interval cancers (42,65,66) (Fig 3). In a retrospective review of 34 DBT-detected architectural distortions that underwent tomosynthesis-guided vacuum-assisted biopsies, Patel et al (67) found a positive predictive value for biopsies performed (PPV₃) of 26% for architectural distortions that were otherwise occult at FFDM and US. Conversely, in another retrospective study involving 181 architectural distortions (68), 59 (32.6%) were detected only with DBT (not with FFDM); these lesions were associated with a lower rate of US correlation and a lower malignancy outcome. In this study, the final malignancy rate after core needle biopsy and subsequent excisional biopsies was significantly lower for architectural distortions detected only at DBT (five of 59, 10.2%) compared with those detected with FFDM alone (53 of 122, 43.4%) (P < .001), which is due to the increased lesion conspicuity of both benign and malignant lesions detected on DBT images as architectural distortion (68). Therefore, additional imaging is recommended to characterize areas of distortion seen only with DBT imaging. If there is uncertainty regarding a finding, repeating the original tomosynthesis view may be preferred over a spot compression view because spot compression views may occasionally efface the distortion (7). US should be considered as the next step if there is a suspicion of distortion even if visualized on only one view at DBT imaging (34). Concerning or persistent distortion on one or both DBT views with no explanation (prior trauma, biopsy, or surgery) and no US correlate should still be sampled for biopsy with use of tomosynthesis guidance given that the finding is associated with a relatively high probability of malignancy, with a PPV₃ of up to 50% (33).

Figure 3a:

Images in 62-year-old woman with an enlarged axillary lymph node at screening mammography. (a) Full-field digital mammogram in mediolateral oblique view demonstrates an enlarged low left axillary lymph node (arrow). (b, c) DBT images show subtle architectural distortion (arrow) in superior part of breast on mediolateral oblique view (b) and in central part of breast on craniocaudal view (c). (d) US examination demonstrates a suspicious hypoechoic mass at the 12 o’clock location. Pathologic examination of biopsy specimen revealed invasive lobular carcinoma.

Figure 3b:

Images in 62-year-old woman with an enlarged axillary lymph node at screening mammography. (a) Full-field digital mammogram in mediolateral oblique view demonstrates an enlarged low left axillary lymph node (arrow). (b, c) DBT images show subtle architectural distortion (arrow) in superior part of breast on mediolateral oblique view (b) and in central part of breast on craniocaudal view (c). (d) US examination demonstrates a suspicious hypoechoic mass at the 12 o’clock location. Pathologic examination of biopsy specimen revealed invasive lobular carcinoma.

Figure 3c:

Images in 62-year-old woman with an enlarged axillary lymph node at screening mammography. (a) Full-field digital mammogram in mediolateral oblique view demonstrates an enlarged low left axillary lymph node (arrow). (b, c) DBT images show subtle architectural distortion (arrow) in superior part of breast on mediolateral oblique view (b) and in central part of breast on craniocaudal view (c). (d) US examination demonstrates a suspicious hypoechoic mass at the 12 o’clock location. Pathologic examination of biopsy specimen revealed invasive lobular carcinoma.

Figure 3d:

Images in 62-year-old woman with an enlarged axillary lymph node at screening mammography. (a) Full-field digital mammogram in mediolateral oblique view demonstrates an enlarged low left axillary lymph node (arrow). (b, c) DBT images show subtle architectural distortion (arrow) in superior part of breast on mediolateral oblique view (b) and in central part of breast on craniocaudal view (c). (d) US examination demonstrates a suspicious hypoechoic mass at the 12 o’clock location. Pathologic examination of biopsy specimen revealed invasive lobular carcinoma.

Asymmetries

At DBT imaging, the imaging data from multiple additional angles that are reconstructed to create the DBT sections often help resolve asymmetries that are due to summations of normal breast tissue in FFDM or SM (69,70) (Fig 4). In a retrospective analysis of 12 921 DBT and 12 577 digital mammography studies (33), the screening recall rate was lower with DBT than with FFDM for asymmetries (13.3% vs 32.2%, respectively) and focal asymmetries (18.2% vs 32.2%) (P < .0001 for both). This large reduction in what would be false-positive recalls for asymmetries is a significant contributor to the overall improvement in specificity achieved with DBT imaging.

Figure 4a:

Images in 40-year-old woman who presented for screening with an asymmetry seen at full-field digital mammography. (a) Synthetic mammogram (SM) in mediolateral oblique projection demonstrates asymmetry (arrow) in upper part of the breast. (b–d) On sequential DBT reconstructions, each component of the asymmetry is discernible as separate, individual normal structures that are superimposed as a summation artifact on the SM image. (e) SM in the craniocaudal view demonstrates no abnormality. L = lateral, M = medial.

Figure 4b:

Images in 40-year-old woman who presented for screening with an asymmetry seen at full-field digital mammography. (a) Synthetic mammogram (SM) in mediolateral oblique projection demonstrates asymmetry (arrow) in upper part of the breast. (b–d) On sequential DBT reconstructions, each component of the asymmetry is discernible as separate, individual normal structures that are superimposed as a summation artifact on the SM image. (e) SM in the craniocaudal view demonstrates no abnormality. L = lateral, M = medial.

Figure 4c:

Images in 40-year-old woman who presented for screening with an asymmetry seen at full-field digital mammography. (a) Synthetic mammogram (SM) in mediolateral oblique projection demonstrates asymmetry (arrow) in upper part of the breast. (b–d) On sequential DBT reconstructions, each component of the asymmetry is discernible as separate, individual normal structures that are superimposed as a summation artifact on the SM image. (e) SM in the craniocaudal view demonstrates no abnormality. L = lateral, M = medial.

Figure 4d:

Images in 40-year-old woman who presented for screening with an asymmetry seen at full-field digital mammography. (a) Synthetic mammogram (SM) in mediolateral oblique projection demonstrates asymmetry (arrow) in upper part of the breast. (b–d) On sequential DBT reconstructions, each component of the asymmetry is discernible as separate, individual normal structures that are superimposed as a summation artifact on the SM image. (e) SM in the craniocaudal view demonstrates no abnormality. L = lateral, M = medial.

Figure 4e:

Images in 40-year-old woman who presented for screening with an asymmetry seen at full-field digital mammography. (a) Synthetic mammogram (SM) in mediolateral oblique projection demonstrates asymmetry (arrow) in upper part of the breast. (b–d) On sequential DBT reconstructions, each component of the asymmetry is discernible as separate, individual normal structures that are superimposed as a summation artifact on the SM image. (e) SM in the craniocaudal view demonstrates no abnormality. L = lateral, M = medial.

Masses

Because DBT enables evaluation of the breast as in-plane sections, the margins of both malignant and benign lesions are better characterized (7–9). Recent studies have suggested that DBT can replace conventional diagnostic mammography views, such as spot views, and achieve similar sensitivity and specificity (9–11). Benign findings containing fat (eg, galactoceles, hamartomas, lipomas, and evolving fat necrosis) may be better characterized with DBT owing to the in-plane imaging and increased conspicuity of lucent areas of fat, potentially decreasing recall rates, false-positive rates, and additional US evaluation (8) (Fig 5). However, as fat may be seen within malignant lesions on DBT images, it is important to assess the lesion shape, margin, and change over time (71).

Figure 5a:

Images in 48-year-old man who presented with palpable lump in right breast. (a, b) Image from digital breast tomosynthesis (DBT) in mediolateral view (a) demonstrates area of fat necrosis (arrow) not well seen on full-field digital mammogram in mediolateral view (b). Patient reported being in a motor vehicle collision several months earlier, with direct trauma to the breast. No further workup was needed due to the typical appearance of fat necrosis on DBT image.

Figure 5b:

Images in 48-year-old man who presented with palpable lump in right breast. (a, b) Image from digital breast tomosynthesis (DBT) in mediolateral view (a) demonstrates area of fat necrosis (arrow) not well seen on full-field digital mammogram in mediolateral view (b). Patient reported being in a motor vehicle collision several months earlier, with direct trauma to the breast. No further workup was needed due to the typical appearance of fat necrosis on DBT image.

Cancer Conspicuity according to View

Cancers may be seen only on one view due to field of view or have increased conspicuity in one view and not in another. This is more common with invasive lobular cancers, as previously described by Sickles in FFDM (72). Several studies involving DBT have suggested that cancers are more conspicuous on the craniocaudal view, where there is often more direct force from compression on the breast parenchyma (compared with the mediolateral oblique view, where the pectoralis muscle, and even abdomen, may bear the greatest force) and where positioning is more reproducible over time, making subtle changes detectable (7,56,57,73). In a reader study by Korhonen et al (56), who reviewed 197 breast cancers imaged with DBT plus FFDM screening, cancers were significantly more conspicuous on the craniocaudal view than on the mediolateral oblique view with both FFDM and DBT (P < .001). In a study by Rafferty et al (73), which included 34 mixed benign and malignant lesions seen at DBT, 15% of the lesions were seen better on the craniocaudal view. Both craniocaudal and mediolateral oblique views, however, are needed for a complete evaluation because a lesion may be visualized on only one of two views due to the location in the breast.

Practice Considerations

Decreased Mammographic Workup

DBT has been found to improve the diagnostic accuracy for noncalcified lesions compared with supplemental FFDM-only views (74), with fewer additional images obtained at recall with DBT screening compared with FFDM alone (26,33). In a reader study evaluating 548 diagnostic examinations, Park et al (69) found that FFDM spot compression views provided no more diagnostic value over that with DBT imaging, apart from a single case where there was inadequate positioning. In a retrospective study by Greenberg et al (26), 46 of 131 patients who underwent DBT (35.1%) had no additional mammographic views performed at recall (compared with 12 of 190 [6.3%] in the FFDM-only group), suggesting a more efficient recall assessment (P < .001). In the Tomosynthesis Assessment Clinic Trial (75), which involved 144 women, the use of DBT was associated with a significant reduction in the need for additional views (χ² = 17.63, P < .001) and recommendations for US (χ² = 8.56, P = .003). In a recent multicenter retrospective study of 194 437 DBT screenings compared with 131 292 FFDM screenings (76), women screened with DBT were more likely to proceed directly to US without additional mammographic imaging compared with women screened with FFDM alone (11.9% with DBT vs 2.8% with FFDM, P < .001). In addition, the time to biopsy (18 vs 22 days) and final diagnosis (10 vs 13 days) was shorter for DBT. The improved efficiency that is possible with DBT translates to both screening and diagnostic breast imaging and could have an impact on the cost-effectiveness of breast imaging.

SM Examination

Although DBT is associated with improved diagnostic accuracy, some of the limitations of DBT plus FFDM include increased radiation, increased acquisition time, and increased interpretation time when compared with FFDM alone (15–17). The reconstruction of synthetic, two-dimensional–like images from the DBT acquisition to supplant the need for FFDM in combination DBT plus FFDM can significantly reduce the x-ray dose of DBT imaging (17,77,78). In one prospective multi-arm trial comparing FFDM, DBT plus FFDM, and DBT plus SM (79), the estimated average glandular dose for a single mammographic view (±standard deviation) was 1.58 mGy ± 0.61 (range, 0.74–4.51 mGy) for FFDM and 1.95 mGy ± 0.58 (range, 1.05–3.78 mGy) for DBT. By replacing FFDM with SM in the DBT plus FFDM examination, the average glandular dose was decreased by approximately 45%, with the dose with DBT plus SM just 19% higher than that with FFDM alone (49). Svahn et al (17), in a review of 17 publications (and five different types of DBT units), also concluded that while using DBT plus FFDM increases radiation dose by a factor of up to 2.25 compared with FFDM alone, using SM to replace the FFDM portion can bring the radiation dose to a level comparable to that with FFDM alone. Zuckerman et al (77), in a study evaluating the early clinical implementation of DBT plus SM, calculated an actual (versus modeled) average glandular dose of 4.88 mGy per examination for DBT plus SM, compared with 7.97 mGy for DBT plus FFDM, or an average reduction of 39%.

A few early retrospective reader studies demonstrated lower or marginal sensitivity and comparable or higher specificity when DBT plus SM was compared with DBT plus FFDM (50,80). Since then, multiple observational screening studies have shown more promising outcomes (21,22,77,81,82). Two prospective European screening studies have demonstrated comparable recall rates and increased CDRs when comparing DBT plus SM to FFDM only (21,22). In addition, multiple new observational studies have compared screening outcomes with DBT plus SM to those with DBT plus FFDM (77,81,82) (Table 2). The observational study by Zuckerman et al (77) found similar outcomes in recall rates, biopsy rates, and CDRs, and a retrospective study by Ambinder et al (81) found DBT plus SM to have decreased recall rates with no difference in biopsy rate or positive predictive value of positive screening results. Of note, however, is that three studies have shown a decrease in detection of ductal carcinoma in situ, suggesting that SM may not have the same sensitivity for the detection of some early malignant calcific-only lesions (50,77,82). However, in a recent study by Hofvind et al (22) comparing DBT plus SM to DBT plus FFDM, DBT plus SM had a higher CDR for both invasive cancers and ductal carcinoma in situ, suggesting that it is adequate for the detection of ductal carcinoma in situ. The synthetic imaging platform is rapidly evolving, and further evaluation on the detection of both specific imaging finding types and cancer subtypes is needed.

Table 2:

Screening Outcomes: Comparison of DBT plus FFDM to DBT plus SM

Note.—CDR = cancer detection rate, DBT = digital breast tomosynthesis, FFDM = full-field digital mammography, PPV₁ = positive predictive value for recall from screening, SM = synthetic mammography.

*Numbers in parentheses are the percentage change.

^†There was a total of 24 901 patients in the trial. Trial period 1 used an earlier version of image reconstruction software.

^‡Not statistically significant.

Although SM may increase the conspicuity of some lesions with calcifications, distortions, and spiculated margins, pitfalls and artifacts associated with SM include pseudo-calcifications, decreased axillary contrast resolution, bright-band artifact (a bright band of subcutaneous tissue just beneath the skin, also called subcutaneous tissue blurring), and decreased contrast resolution due to artifacts created by high-attenuation foreign bodies such as metal surgical clips (64).

Use of the Breast Imaging Reporting and Data System with DBT

DBT has resulted in a shift in the relative proportions of the Breast Imaging Reporting and Data System (BI-RADS) assessment categories (43) compared with imaging with FFDM alone. In a retrospective study, Raghu et al (84) reviewed all diagnostic mammograms obtained during a 12-month interval of FFDM before DBT (n = 3576) and for 3 consecutive years after implementation of DBT (n = 5288 at year 1, n = 5177 at year 2, n = 4676 at year 3) and compared the FFDM-only group to the DBT year 3 group (where 99% of the patients underwent tomosynthesis). They found that DBT resulted in an increase in the proportion of studies classified as normal or benign (BI-RADS category 1 or 2, 58.7% with FFDM vs 75.8% with DBT; P < .0001), a reduction in the percentage of studies classified as probably benign (BI-RADS category 3, 1181 of 3550 [33.3%] with FFDM vs 764 of 4653 [16.4%] with DBT; P < .0001), and an increase in the PPV₃ (85 of 287 [29.6%] with FFDM vs 182 of 364 [50.0%] with DBT; P < .0001) (84). This change in the distribution of BI-RADS assessments most likely reflects a greater confidence in interpreting DBT compared with FFDM alone. The improved confidence in reader interpretation is most likely due to the ability to resolve areas of tissue superimposition with DBT that may cause false-positive findings at FFDM, coupled with the improvements in lesion conspicuity achieved with DBT (7). Decreased (BI-RADS category 3) short-term follow-up recommendations with similar CDRs can mean less anxiety for patients and decreased costs. At our institution, the proportion of BI-RADS category 3 assignments after recall from screening remained the same after DBT implementation; overall, however, the number of patients assigned to short-term follow-up decreased due to the overall decrease in the number of women recalled with DBT compared with FFDM-only screening (46).

Tomosynthesis-guided Biopsy

DBT-guided vacuum-assisted biopsy allows tissue sampling of lesions detected only with DBT as well as other mammographically visible lesions that were previously biopsied with stereotactic guidance. In the retrospective study by Schrading et al involving 216 mammographic findings seen at FFDM (85), DBT vacuum-assisted biopsy achieved technical success in 100% (51 of 51) of lesions, compared with 93% (154 of 165) of lesions sampled with prone stereotactic vacuum-assisted biopsy. In a retrospective review by Bahl et al (86), technical success was achieved for more lesions with DBT vacuum-assisted biopsy than with prone stereotactic vacuum-assisted biopsy (99.3% [695 of 700] vs 95.1% [410 of 431], respectively; P < .001). Bahl et al also found that more biopsies were performed for noncalcified lesions (eg, architectural distortion, asymmetry, and mass) in the DBT-guided vacuum-assisted biopsy group than in the prone stereotactic vacuum-assisted biopsy group (29.2% [203 of 695] vs 3.4% [14 of 410], respectively; P < .001) (86).The advantages of DBT guidance for biopsy include that z-axis, or depth information, may be obtained without the need for stereotactic imaging pairs and that DBT allows the full detector field for imaging during DBT biopsy (versus the smaller window of conventional FFDM stereotactic systems). DBT-guided procedures have also been shown to have a decreased total biopsy time compared with conventional FFDM stereotactic biopsies (mean, 4 minutes vs 15 minutes in the study by Schrading et al and 12 minutes vs 27 minutes in the study by Bahl et al) (85). Although there is additional glandular dose associated with each individual DBT acquisition during DBT-guided biopsies, because there are fewer exposures obtained with DBT guidance there may be an overall decrease in total radiation dose of DBT guidance compared with stereotactic guidance (85–87).

Biopsy Rates and PPV₃

Some of the early studies reporting outcomes after implementation of DBT screening demonstrated slightly higher biopsy rates but also higher PPV₃ (13,18,24–26,33). In the multicenter retrospective study by Friedewald et al (25), which was mostly prevalent round DBT screening, the model-adjusted rates for biopsy were significantly higher for DBT (19.3 per 1000 screens; 95% confidence interval [CI]= 16.6, 22.1 per 1000 screens) compared with FFDM (18.1 per 1000 screens; 95% CI = 15.4, 20.8 per 1000 screens) (P = .004). However, the mean PPV₃ was 29.2% (95% CI = 26.0%, 32.3%) with DBT plus FFDM versus 24.2% (95% CI = 21.1%, 27.1%) with FFDM (P < .001) (25). In the PROSPR consortium study (13), where patient level and screening round data were available, biopsy rates were significantly higher for DBT compared with FFDM (2.0% vs 1.8%, respectively; P = .0074); however, after adjusting for center, age, breast density, and first examination, the odds of biopsy were significantly lower for DBT than for FFDM (odds ratio = 0.85; 95% CI = 0.77, 0.93). In the observational study of the first 3 years of DBT screening by McDonald et al (14), the rate of biopsies performed in each DBT year did not differ significantly from that of FFDM (FFDM vs DBT year 1: odds ratio = 1.05 [95% CI = 0.87, 1.28], P = .17; FFDM vs DBT year 2: odds ratio = 1.15 [95% CI = 0.94, 1.39], P = .61; and FFDM vs DBT year 3: odds ratio = 1.05 [95% CI = 0.86, 1.29], P = .60). Therefore, it appears that the initial round of DBT screening may be associated with an increased rate of biopsy but that rate returns to a similar level to FFDM at subsequent screening rounds and that biopsies generated from DBT findings are associated with a higher cancer outcome.

Reimbursement and Cost Considerations

In 2015, coverage was approved by Centers for Medicare and Medicaid Services for DBT without copayment for screening indications (although a copayment is required for diagnostic indications). However, DBT is not uniformly covered by commercial insurers, and there is indication that the use of DBT for screening may be associated with insurance coverage, resulting in possible disparities in access for patient subgroups (88). In a study involving 22 primary care centers (88), DBT screening was performed more frequently under private insurance (43.4%) than under Medicaid (36.2%), Medicare (37.8%), other insurance (38.6%), or no insurance (32.9%); there were no sustained differences in the use of DBT for diagnostic testing according to insurance type.

The main drivers of economic value in implementing DBT at the population level are as follows: (a) the reduced screening recall rates, (b) a more direct or expedited mammographic evaluation at diagnostic or problem-solving imaging that may decrease overall cost, and (c) the potential to detect cancers at an earlier stage, reducing costs of later stage treatments (89). A clinical-economic value model analysis of DBT for breast cancer screening among Medicaid beneficiaries demonstrated that the estimated annual cost savings from DBT amounts to $8.14 per patient, which potentially translates into more than $12 000 savings per year for an average-sized Medicaid plan and as much as $207 000 savings per year for a typical state Medicaid program (89). There are also potential indirect cost savings for the patient, including time away from work, transportation, and child care costs. Finally, there is potential reduction in the psychosocial costs related to anxiety associated with false-positive results (90–92).

Dense Breasts and Supplemental Screening

Multiple studies have shown that the increased cancer detection and reduction in screening recall associated with DBT is achieved across all breast densities (13,18,23,28,30). Although the increases in cancer detection with DBT were found when breast density was grouped dichotomously into nondense (BI-RADS categories 1 and 2, almost entirely fatty and scattered fibroglandular, respectively) and dense (BI-RADS categories 3 and 4, heterogeneously dense and extremely dense, respectively) groups, the largest improvements in CDR were achieved in the scattered fibroglandular and heterogeneously dense categories, with the smallest gains in almost entirely fatty and extremely dense breasts (43,93–95). In the Screening with Tomosynthesis or Standard Mammography-2 prospective study (93), the incremental CDR attributed to adding DBT was 5.4 (per 1000 screening examinations) (for DBT plus FFDM) and 6.2 (for DBT plus SM) for dense breasts versus 1.0 (for DBT plus FFDM) and 1.1 (for DBT plus SM) for nondense breasts. A systematic review and meta-analysis of 16 studies by Phi et al (including five diagnostic and 11 screening studies) (96) compared outcome metrics of DBT (without or with FFDM) versus FFDM alone for women with dense breasts. For women categorized as having heterogeneous or extremely dense breasts, in the screening setting, DBT was associated with a higher CDR in both retrospective (n = 115 838 for DBT vs n = 188 419 for FFDM; relative risk = 1.33, 95% CI = 1.20, 1.47) and prospective (n = 6957 for DBT vs n = 6957 for FFDM; relative risk = 1.52, 95% CI = 1.08, 2.11) studies. In addition, Phi et al (96) found that recall rate was significantly lower with DBT plus FFDM than with FFDM alone in retrospective studies (relative risk = 0.72; 95% CI = 0.64, 0.80) but not in the two prospective European studies (relative risk = 1.12; 95% CI = 0.76, 1.63), which probably reflects the already low recall rate with FFDM in the European trials.

The Adjunct Screening with Tomosynthesis or Ultrasound in Women with Mammography-Negative Dense Breasts, or ASTOUND-2, trial (97), a multicenter prospective screening trial involving 5300 women with dense breasts who had negative findings on the FFDM or SM portion of the DBT screening examination, found that the incremental CDR after the addition of DBT was 2.83 per 1000 screens (95% CI = 1.58, 4.67 per 1000 screens) whereas that for hand-held whole-breast US was 4.90 per 1000 screens (95% CI = 3.21, 7.19 per 1000 screens; P = .015); US, however, was associated with more false-positive findings. Although DBT improves cancer detection, US is complementary and will find additional cancers beyond that of DBT, particularly in denser breast. However, the results of the Adjunct Screening with Tomosynthesis or Ultrasound in Women with Mammography-Negative Dense Breasts study underscores the trade-off of increased number of false-positive findings that accompany the additional cancers detected with supplemental screening (97).

Future of DBT Imaging

There is ongoing research to develop next-generation DBT units with varied acquisition arcs to increase spatial resolution, and therefore lesion conspicuity, for both DBT and reconstructed synthetic images (98). Studies evaluating contrast material–enhanced digital mammography and contrast-enhanced DBT suggest that these modalities may be an alternative for supplemental screening for some women in lieu of MRI (99,100).

Although DBT is associated with significantly longer reading times than DM (15,18), interpretation time is expected to decrease with increasing experience with DBT. To address reader efficiency, new presentation modes, including automated “slabbing” of DBT sections to create thicker overlapping sections, have been developed (54). Machine learning–based detection algorithms have also been created to “flag” sections as well as to document the thickness of a detected lesion within the DBT stack to more rapidly guide DBT review (101). In addition, there are now U.S. Food and Drug Administration–approved DBT machine-learning computer-aided detection programs that use lesion detection data to drive the generation of synthetic images. These “computer-aided detection–enhanced” images increase the conspicuity of lesions on SM images, and in early reader studies have been shown to decrease reading time by 29.2% while maintaining performance accuracy (101). The use of machine learning coupled with computer-aided detection is also being investigated to quantitatively classify breast density categories, predict benign versus malignant masses, and improve the diagnostic accuracy and efficiency of radiologists (102–104).

Conclusion

Continued implementation of digital breast tomosynthesis is associated with improvements in screening outcomes, including increased cancer detection rates and improved specificity across all breast densities. The introduction of synthetic reconstruction technology to replace the digital mammography portion of the examination can reduce radiation dose while maintaining the outcomes of screening achieved with digital breast tomosynthesis plus full-field digital mammography. Early studies suggest that digital breast tomosynthesis–detected cancers tend to be smaller and of lower grade, potentially improving prognosis. Early studies have shown similar false-negative rates with digital breast tomosynthesis screening, although larger studies involving longer follow-up may be required to understand specific cancer biologic subtypes and the potential for longer term benefits. In clinical practice, the approach to lesion evaluation and workup is altered due to improved localization and characterization, challenging some of the established mammographic imaging pathways and resulting in improved efficiency and potential for cost savings. New presentation modes and machine learning–based computer-aided detection techniques are being developed for the next generation of digital breast tomosynthesis to improve breast imaging accuracy and efficiency.

In summary, digital breast tomosynthesis is becoming the standard of care in both screening and diagnostic breast imaging due to improvements in patient outcomes and imaging efficiency. Although early data are promising, additional research is needed to better understand the long-term effects of digital breast tomosynthesis imaging, especially in screening outcomes, where the biology of cancers detected, as well as those not detected, is particularly important.