The Evolution of BI-RADS 3

I remember very early in my career sitting next to Mike Linver at a Society of Breast Imaging (SBI) meeting, listening to a presentation on probably benign findings (Breast Imaging Reporting and Data System [BI-RADS] assessment category 3) (1). The presenter had said that an oval circumscribed mass on a baseline mammogram or US could be managed through short-interval follow-up—unless it was palpable. I leaned over to Mike and whispered, “Do you biopsy every probable fibroadenoma if it is palpable?” He replied, “Typically, I do.” We had routinely given these BI-RADS 3 whether palpable or not in our practice. I realized that this was an area of controversy and decided to review our data. Our manuscript was rejected from the first journal because “all palpable breast masses should undergo biopsy.” We subsequently published our work in a different journal demonstrating a less than 2% risk of cancer for palpable oval circumscribed masses (2). Changes in practice should be data driven.

In this issue of the Journal of Breast Imaging, our Scientific Review provides a long-needed data-driven approach to using BI-RADS 3 for MRI (3). I have not been a frequent user of BI-RADS 3 in MRI because of the lack of data for specific findings, so I found this review very helpful. Admittedly, my practice has been to give an oval, circumscribed, T2-hyperintense mass with slow persistent kinetics a BI-RADS 2 rather than a BI-RADS 3 as these are likely fibroadenomas and have all benign features. Perhaps this is a new area of controversy that will spur more research!

Speaking of fibroadenomas, our Radiologic-Pathologic article by Hayes et al (4) is a fantastic review of fibroepithelial lesions with a focus on benign and borderline entities. The pathology images beautifully demonstrate the spectrum of these lesions. Interestingly, they note that while wide surgical excision remains the primary treatment for benign and borderline phyllodes tumors, a focally positive margin may not increase local recurrence. For a review of malignant phyllodes tumor, Lee et al (5) recently provided an excellent radiologic-pathology article on this topic in JBI.

Artificial intelligence (AI) has become a hot topic because of both the promise of improved detection and diagnosis of imaging findings as well as potential improvements in efficiency. This issue of JBI opens with a review of machine learning and AI for breast cancer screening (6). The senior author, Maryellen Giger, is a long-time researcher in this area beginning with computer-aided detection many decades ago. The authors caution that before the implementation of AI as an independent reader, context such as the target population, repeatability, and performance of the algorithm and the radiologist must be considered (6).

An original research article by Retson et al (7) evaluates a Food and Drug Administration–approved AI algorithm for mammography triage, finding that the performance was similar to average practicing radiologists in the Breast Cancer Surveillance Consortium; when the sensitivity of the algorithm was held at 86.9%, the specificity was 88.5%. Interestingly, the algorithm performance was maintained across all breast densities. While I do not anticipate breast radiologists to ever be replaced, I suspect that we would all welcome improvement in efficiency.

Cain et al (8) reviewed contrast-enhanced mammograms and used a quantitative assessment of enhancement to predict malignant from high-risk and benign lesions. Interestingly, human epidermal growth factor receptor-2 (HER2) overexpressing tumors had higher enhancement ratios compared with luminal and basal cancers.

Alvarenga et al (9) describe a low-tech method of improving efficiency in which referring providers place orders for diagnostic evaluation and biopsy if needed for women undergoing screening mammography. The authors describe this process as “reflex testing,” in which the radiologists and breast imaging center staff manage abnormal findings directly. The result was a significant reduction in the time from a BI-RADS 0 to diagnostic testing as well as a reduction in the time from the first diagnostic test to biopsy.

Our original research articles include two surveys of members of the SBI. The first queries the use of abbreviated breast MRI, finding that about half of responding SBI members either currently offer or plan to offer this service (10). While most respondents had similar approaches as to which sequences to use for abbreviated breast MRI, patient eligibility was highly variable. Of concern, only 6% of respondents provide financial assistance, which could widen the health care equity gap. In the second survey article (11), SBI members were asked to provide BI-RADS density assessment for three mammograms based on their routine practice, as well as definitions of density from the fourth (12) and fifth (1) editions of the BI-RADS lexicon. Density assessment was highly variable when comparing usual practice to the fourth edition of the lexicon, but very consistent with the fifth edition. The majority of respondents (79%) thought that density assessment should reflect both potential for masking and overall dense tissue for risk assessment. About half of respondents thought that quantitative methods were useful.

Rai et al (13) provide a beautiful educational review on calcified axillary lesions demonstrating benign and malignant etiologies. Excellent examples of pseudocalcifications are also included. The table of differentials by location is a very useful reference.

The Training and Professional Development article is an in-depth review of why and how breast radiologists should consider coaching by nationally recognized expert Rex Gatto, as well as Wendie Berg and Martha Mainiero (14). In my mind, coaching has two contexts: “good manners” and skill development. For many, coaching may have a negative connotation. Perhaps you said the wrong thing to a technologist when you were tired, or you can’t help taking over procedures from the residents. I consider this “good manners” coaching. A problem is identified, and coaching typically addresses specific alternate methods of managing similar situations. It is unfortunate that this type of coaching, which is often quite successful, may feel punitive as the person is typically referred for coaching by a division head, department chair, or other leader, and outcomes may be reported back to them. On the other hand, a proactive approach to self-identifying areas for improvement and working with a coach may be experienced as empowering. Skill development or leadership coaching is fabulous. This type of coaching provides enlightenment regarding personal leadership style, strengths and weaknesses, and interacting with those around and above you. I have done two sessions with the same coach, which was transformative. Highly recommend!

References