Tomosynthesis in Screening Mammography

A successful screening program relies on the ability of radiologists to detect potential malignancies while maintaining an acceptably low recall rate. To do so requires rigorous adherence to quality controls for both personnel and equipment, as warranted by the US Food and Drug Administration (FDA) and the Mammography Quality Standards Act. The controversy that follows screening mammography has largely focused on patient anxiety and increased health care costs due to the number of false alarms and the additional work-up these cases entail.

The hallmark of tomosynthesis is the dual benefit of decreasing false positives while increasing cancer detection in screening mammography. Tomosynthesis allows radiologists to scroll through thin slices of breast tissue, reducing superimposed tissue and the structured noise that limits conventional two-dimensional (2D) mammography. This permits enhanced lesion detection and facilitates lesion margin analysis of both benign and malignant findings. In addition, overlapping tissue can be confidently assessed as normal, and unnecessary recalls can be obviated.

How to Read Screening Mammograms With Tomosynthesis

Batch reading of analog screening was a cost-effective method because labor and time to load and take down films on an alternator made it impractical to read screening cases online. Even for a few years during the transition to digital mammography, in which prior analog images were still needed for comparison, batch reading was still the most practical approach. However, when a practice achieves a state in which the current and recent prior images are all digital, reading online is feasible. Over time, due to tomosynthesis, practices may find that fewer patients are in the recall or follow-up diagnostic pool, and the radiologist has more time to read screening cases online. Thus, although individual tomosynthesis screening cases take more time to view and interpret, overall the caseload and workflow are improved.

Tips for interpreting tomosynthesis images are discussed in Chapter 7 , but a few topics are worth repeating here. Reading any mammogram requires the undivided attention of the radiologist, but reading tomosynthesis cases requires even more concentration. Reading a large number of screening tomosynthesis exams consecutively can potentially be more tiring compared with 2D mammography interpretation. Distractions are common in busy practices, but attempts must be made to minimize them. Taking breaks frequently during the day between cases is also recommended to avoid excessive fatigue.

Much attention has been given to the increased time required to interpret tomosynthesis exams. Given the larger number of images and increased amount of information inherent in tomosynthesis screening, it is inevitable that the exams will take longer to read.

However, radiologists quickly become accustomed to the tomosynthesis hanging protocol, and conversely having only 2D images to interpret seems uncomfortably limited. In addition, it must be recognized that the increase in time required to read tomosynthesis screening exams is balanced by the reduction in time required for diagnostic studies.

A common concern is whether review of previous tomosynthesis images is necessary when assessing subsequent years’ exams. Although individuals may have their own personal protocols, routine review of prior tomosynthesis exams is generally not necessary. Comparing the prior 2D images usually suffices; however, if questionable areas are noted on the current exam, scrolling through the prior tomosynthesis slices is greatly beneficial.

As with reading 2D mammograms, assessing subtle changes over time is essential for detection of early malignancies. Although many cancers will appear more obviously spiculated or distorted on tomosynthesis images compared with 2D images, not all malignancies will have this characteristic appearance. Some cancers will still be identified only as a very subtle focal asymmetry or mass. These findings require careful attention and recall for diagnostic evaluation.

The use of tomosynthesis in both the craniocaudal (CC) and mediolateral oblique (MLO) projections is very important in obtaining the full benefit of tomosynthesis. Although the MLO view may capture more of the breast tissue than the CC view, the CC view produces better compression and separation of tissue that is more uniform from year to year, making the reader’s assessment of changing tissue patterns more confident. There is evidence that suspicious lesions are more frequently detected on the CC than the MLO view, making the CC view essential to mammographic assessment. The CC view also allows for more precise lesion localization information. The benefit of performing screening tomosynthesis in the MLO view only is the reduction of overall radiation exposure, although this may be offset by the loss of potentially vital information provided by the CC tomosynthesis view.

How Is Tomosynthesis Reducing Recall Rates?

The fifth edition of the American College of Radiology Breast Imaging Reporting and Data System (ACR BI-RADS) Atlas states that a recall rate of up to 12% is acceptable for screening mammography. In 2013, results from the Oslo Tomosynthesis Screening Trial were published comparing conventional digital mammography alone with digital mammography plus tomosynthesis. In this single-institution prospective study of 12,631 screening exams, a 15% reduction in false-positive recalls was seen with the use of tomosynthesis. In the United States in 2014, a multicenter retrospective study involving both academic and community-based practices reviewed nearly half a million screening exams and also found a 15% reduction in recall rates, solidly corroborating the Oslo findings. Additional single-institution studies have shown significant reductions in recall rates (15% to 30%) since implementing tomosynthesis.

Although reduced recall rates are noted across all patients, the rates are variable, depending on breast density and patient age. More significant reductions are noted in younger women and in those with dense tissue. In addition, in women undergoing baseline mammography, in which interpretation is more challenging due to lack of comparisons, tomosynthesis significantly reduces recall rates by up to 50%. For practices with limited resources, such findings are important to consider in triaging which patients preferentially receive tomosynthesis.

Technical Issues

In some situations, tomosynthesis imaging can prevent what otherwise would require additional imaging for technical reasons. On 2D mammography, artifacts from deodorant or radiopaque lotions preclude adequate assessment of the tissues beneath. Deodorant, talc, or skin lotion artifact can be readily identified on the skin surface with tomosynthesis. Upon scrolling a few slices beyond the dermal surface, the tissue deep to the skin will be seen with clarity, and the need for additional tangential views or repeat imaging is avoided ( Fig. 5.1 ). Superimposed hair can also produce an artifact that may simulate a mass. Hair artifact may appear superimposed on the 2D image, but it does not appear on the tomosynthesis slices because it is above and not within the compression plates ( Fig. 5.2 ).

Asymmetry

False positives in conventional 2D mammography are often the result of superimposed tissue and frequently recalled as asymmetries. With tomosynthesis, careful scrolling through areas of suspected asymmetry often reveals layers of fibroglandular tissue, overlapping at different planes, without evidence of an underlying mass. By virtue of its ability to separate overlapping tissue, tomosynthesis disperses deceptive 2D asymmetries and can resolve these potential false-positive findings as normal fibroglandular tissue or other benign entities, such as crossing Cooper ligaments or tortuous vessels. Lower recall rates are seen for all mammographic findings—asymmetries, calcifications, masses, and architectural distortion—with the most marked decline in the recall of asymmetries ( Fig. 5.3 ).

Calcifications

The detection of calcifications on tomosynthesis has been controversial. Early prototype studies had data acquisition times of as long as 20 seconds, which introduced the possibility of motion artifacts potentially obscuring calcifications. In addition, calcifications can be spread across multiple tissue planes or exist in loose groups. This may limit the sensitivity of tomosynthesis when viewed in 1-mm intervals because the individual thin slices may not capture the entire group of calcifications if the group occupies more than a few mm of breast tissue. However, tomosynthesis units currently in clinical use can acquire images in as little as 4 seconds. In some cases, overlapping tissue can obscure fine calcifications or small groups of calcifications, and thus tomosynthesis can actually improve their detection and characterization. Importantly, tomosynthesis images are usually read in conjunction with a 2D image, whether conventional or synthesized, and this combined interpretation should permit identification of all calcifications at least equal to if not better than 2D mammography alone.

The localization feature is fundamental to decreasing recalls for many dermal calcifications or deodorant artifacts. A quick glance at the localizer will definitively tell the radiologist where the calcifications are located. If the calcifications are at end slices of the dataset, then they are on the skin surface ( Fig. 5.4 ). Importantly, only those skin surfaces in contact with the receptor or compression paddle will be visualized in those bookend slices. Some dermal calcifications may also be captured tangentially. A correlation between the MLO and CC tomosynthesis image series usually permits the confident characterization of most dermal calcifications. This ability to precisely localize some calcifications to the skin surface permits definitive benign assessments of these findings.

Other examples of typically benign calcifications that can be better characterized by tomosynthesis include vascular calcifications, milk of calcium, and fat necrosis. As overlapping tissue is cleared away, milk of calcium can be better visualized layering dependently within benign microcysts, averting recall for magnification views ( Fig. 5.5 ). 2D views of linear vascular calcifications may also be difficult to discern from suspicious ductal calcifications; however, with tomosynthesis the vessel itself can appear and the reader can confidently see the calcifications associated with the vessel ( Fig. 5.6 ). Fat necrosis may produce rim calcifications around lucent fat-containing masses. This may be better appreciated in the tomosynthesis slice images, in which the calcifications are noted to be associated peripherally with the fatty mass. Postoperative dystrophic calcifications are further discussed in Chapter 12 .