Lesion Localization Using Digital Breast Tomosynthesis: Where Did I Go Wrong? (2024)

Search for other works by this author on:

Article history

Abstract

Introduction

The effectiveness of digital breast tomosynthesis (DBT) in the screening setting has been validated in several studies that show both increased cancer detection rates and decreased recall rates with DBT compared with digital mammography alone (1, 2). Tomosynthesis has also become an important tool in the diagnostic setting to evaluate and guide biopsy of asymmetries and one-view architectural distortions (3–5).

The Breast Imaging Reporting and Data System (BI-RADS) lexicon published by the American College of Radiology describes an asymmetry as a finding only seen on one of the two standard views of the breast (6). According to the lexicon, an apparent mass only seen on one view should still be characterized as an asymmetry. In a retrospective review of one-view findings on screening mammograms, Sickles found that 82.7% of the recalls were due to a summation artifact, and cancers were only found in 1.8% (7). Interestingly, however, in reviews of missed cancers, 9%–38% of missed cancers were retrospectively visible as a one-view finding (8–10). Characterization and localization of one-view findings has improved significantly with the use of DBT, along with the ability to sample these findings using tomosynthesis-guided biopsy.

Lesion localization on digital mammography

Before DBT, the tool kit for mammographic evaluation of one-view findings was limited. The lateral view is the first step in localizing lesions only seen on the mediolateral oblique (MLO) views (Figure 1). For example, if an area of architectural distortion (AD) only seen on the MLO view rises to a higher position on the lateral view, the finding will be located in the medial breast.

Craniocaudal (CC) rolled views can be performed for lesions only seen on the CC view, where the superior breast tissue is rolled slightly (Figure 2). Rolling of the tissues changes the relationship of the superior and inferior breast tissue with respect to each other. When obtaining a CC rolled view, the superior portion of the breast can be rolled medially (CCRM) or laterally (CCRL). If the abnormality moves in the direction of the roll, that is, the lesion moves more laterally on the CCRL and more medially on the CCRM, it will be in the superior breast.

Finally, step oblique views involve changing the obliquity of the X-ray beam in 10–15 degree increments starting from the view that the abnormality is seen. The goal of step oblique views is to obtain a view only slightly different from the one where the asymmetry was seen. As the obliquity of the X-ray beam changes, a real finding will persist, whereas an asymmetry caused by summation will resolve (11). The lateral view, rolled views, and step oblique views all serve to discern a persistent finding from a summation shadow by varying the angle of the beam or the position of the breast tissue.

DBT technology

DBT is a mammographic technique whereby multiple low-dose projection images are acquired in an arc across the breast and are reconstructed into thin “slices.” (12–15) One millimeter single section reconstruction images can be reviewed on a workstation, allowing a 3D estimation of the location of the lesion relative to the breast tissue, thereby reducing the effects of tissue superimposition (16).

The total number of reconstruction slices depends on the thickness of the breast. For example, if the compressed breast thickness is 5 cm, and slices are at 1mm intervals, there will be approximately 50 total slices. It is important to note that 5 additional reconstructed slices are added on to the nondetector (ie, the compression paddle) side of the image, as the edge of the breast can be difficult to define because of its curvature, and it is difficult to measure compression accurately down to the millimeter (17). Therefore, for our example, there will 55 total slices, including the extra 5 slices on the CC view added to the superior portion.

Each slice is assigned a slice number and is accompanied by a scroll bar that marks the position of that slice relative to the entire stack. In our example, a lesion on slice 25 of the CC view is likely in the central portion of the breast on the lateral view. This information from the scroll bar helps localize the position of the lesion on the orthogonal view.

Problem solving using the scroll bar

The ability to localize one-view findings using the scroll bar can help target breast ultrasound, guide the approach for stereotactic or tomosynthesis-guided biopsy, and confirm the dermal location of superficial findings.

When an indeterminate finding is only seen on one view, the scroll bar can help localize where to search for the lesion on the orthogonal view. With this guidance and a focused search, a subtle correlate is often identified on the orthogonal view, where the finding may originally have been felt to be occult (Figure 3). Even when a correlate is not identified on the orthogonal view, the location on the scroll bar can help target breast ultrasound, which can be especially useful in patients with large or complicated breasts. Instead of scanning half of the breast, a quadrant of the breast can be honed in on, which can save time and prevent incidental findings.

If a suspicious finding is identified on one view and no correlate is identified on ultrasound, targeting the lesion for a stereotactic core biopsy can often be challenging. This is especially the case for architectural distortion, which is often more conspicuous on the CC view than on the MLO or lateral views. The scroll bar can be used to guide the approach for the biopsy in this situation. For example, for an area of architectural distortion only seen on the CC view that localizes to the superior breast on the scroll bar, a superior approach should be selected. Using tomosynthesis for biopsy guidance can improve the confidence for targeting one-view architectural distortion (Figure 4).

The scroll bar can also be used to confirm dermal calcifications and reduce recall rates. While skin calcifications often have classic lucent centers, sometimes calcifications may not be identifiable as dermal, based on morphology. Assuming the skin thickness is 3 mm, the skin is imaged on the first and last 3 slices on DBT, where it is either touching the compression paddle or the detector. Calcifications and masses on those slices can therefore be confirmed as dermal (Figure 5). However, calcifications can still be dermal and not be within the first or last 3 slices, localized in the areas where the skin in not touching the equipment. For example, because of the curvature of the breast, the skin near the nipple is located deeper in the stack compared with the posterior portion of the breast, where the breast is thicker. Last, as the extra 5 slices are added to the compression paddle side of the breast, if the measured compression thickness is perfectly accurate, dermal calcifications may appear deeper in the stack on the nondetector side (18). If calcifications are suspected to be dermal but not within the first or last 3 slices, either a reverse view can be obtained with the calcifications against the detector or a skin tangential study can be performed.

Pitfalls with DBT scroll bar localization

While the scroll bar usually accurately localizes lesions on the orthogonal view, there are a few specific situations where the scroll bar can lead the reader astray. Recognizing the factors that lead to the seemingly inaccurate lesion localization can help prevent interpretation errors.

Location Based on the True Lateral View

It is important to remember that the scroll bar predicts the lesion location on the orthogonal view. At the time of screening mammography, nonorthogonal CC and MLO views are performed, and the principles of triangulation have to be taken in to account. When viewing the scroll bar on the CC view, the location of the lesion is predicted on the true lateral view, not the MLO view (Figure 6). As discussed earlier, medial lesions rise higher on a true lateral view compared with the MLO view. Therefore, when a lesion is medial on the CC view, the scroll bar will predict the lesion location to be higher than what is seen on the MLO view. Conversely, a lateral lesion will be located lower on the scroll bar than expected based on the MLO view because the scroll bar reflects that the lesion will be lower in location on a true lateral view compared with the MLO view.

Nipple Positioning

When the breast is appropriately positioned, the nipple should be in the center of the breast and the center of the scroll bar, demarcating the superior and inferior breast on the lateral view and medial and lateral portions on the CC view. However, in some patients, the nipple may not be in the center of the breast or more commonly, it may be malpositioned in one extreme direction.

In these cases, the reader should note the scroll bar location when the nipple is in focus and use that as a reference for localizing the lesion (Figure 7). For example, in a patient with reduction mammoplasty where the nipple is elevated (in focus toward the head on the scroll bar), a lesion toward the middle of the scroll bar on the CC view would actually be in the inferior breast.

Lesions Localizing to the Detector Side

Lesions may erroneously project closer to the detector on the scroll bar in patients with thin breast tissue and when the finding is located in the anterior portion of the breast.

To account for inaccuracies in measurement of the compression thickness, 5 additional slices are added to the stack toward the compression paddle side of the breast to ensure that a portion of the breast is not excluded on the reconstructed images. While the extra 5 slices do not make a significant difference in the majority of cases, in patients with very thin breasts, the extra slices can cause a lesion to localize closer to the detector side (Figure 8). For example, in a patient with a compression thickness of 20 mm, the 5 extra slices account for a significant portion of the image stack, and there are 20% more images toward the compression paddle side of the breast. In these cases, a lesion in the center of the breast will appear to be closer to the detector side on the scroll bar.

Because of the curvature of the breast, the anterior portion of the breast is usually thinner and more compressible than the posterior portion is. The scroll bar, however, reflects the thickness of the entire breast, and thus the anterior portion may be included in only a small portion of the scroll bar. Anterior lesions may project toward the detector side of the scroll bar, as the compression paddle flexes toward the detector in the anterior portion, where the breast is thinner. In such cases, it is helpful to assess the location of the lesion in relation to the nipple, instead of solely relying on the location determined by the scroll bar (Figure 9). The principle of localization on the scroll bar assumes even compression throughout the breast. There may also be differences in compressibility of the breast based on tissue composition that may lead to uneven compression and errors in localization. In such cases, the position of the lesion on the scroll bar relative to the nipple may again be used to aid in localization.

Superficial lesions

In the vast majority of the cases, the first 3 slices in the scroll bar are imaging of skin. Findings in these slices can be considered dermal and, therefore, do not need additional diagnostic evaluation. Rarely, in patients with very thin breast tissue and thin skin, it is possible that findings in these first few slices are already within the breast parenchyma. If there is doubt at the time of the exam, a skin tangential study can be performed to confirm whether a finding is dermal (Figure 10).

Additionally, superficial lesions are more susceptible to rolling with the skin if the breast is not positioned appropriately. This is especially true in large breasts, where the CC view is more susceptible to the rolling of breast tissues. If there is inappropriate positioning, it may seem that the location predicted by the scroll bar on the orthogonal view was inaccurate. In such cases, a repeat view can be performed to confirm that the predicted location on the scroll bar is correct and that the apparent error was positional.

Conclusion

It is important for the breast imaging radiologist to realize that while the scroll bar on DBT is useful for lesion localization, there are specific scenarios when it may not appear accurate. Alternative cues such as the position of the lesion relative to the nipple on the scroll bar can be used to aid localization in such cases. Knowledge of the pitfalls with the scroll bar on DBT can prevent interpretation errors and allow for accurate localization of findings only seen on one view.

Conflict of interest statement

D.G. declares no conflicts. S.M.F. is a recipient of a research grant and is a scientific advisory board member of Hologic, Inc.

To obtain CME for this article, go to: https://sbi-online.peachnewmedia.com

References

© Society of Breast Imaging 2019. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Issue Section:

Educational Review

Email alerts

Citing articles via

More from Oxford Academic