Purpose: Ultrasound has shown value as a supplemental breast cancer screening exam in women with dense breasts, though is often criticized for high false positives. Automated ultrasound (AUS) has the potential to overcome limitations of handheld ultrasound, such as operator dependence and reproducibility, though radiologists may be unfamiliar with interpretation given the unique 3D/multiplanar technique. This study examines mammographically occult AUS detected cancers in an academic center to identify common features, using both traditional BI-RADS terminology and descriptors unique to AUS.
Materials and Methods: An IRB-approved retrospective analysis of AUS-detected cancers at an academic center over a 5-year period was conducted. Mammographically occult breast cancers were evaluated using BI-RADS lexicon terminology as well as AUS features, including data set(s) on which the cancer was identified and visualization of architectural distortion on the coronal images. Patient demographics, cancer size, histology and stage were also included.
Results: From 2015 to 2020, a total of 8891 AUS exams were performed with detection of 20 cancers in 19 women with a negative concurrent screening mammogram and dense breast tissue, with an overall recall rate of 5.4%, PPV1 of 4.3% and incremental cancer detection rate of 2.2/1000. Mean patient age was 61 (SD 12.49), with the majority (18, 94.7%) having heterogenous breast density. Most cancers were stage 1A (17, 85%) and of invasive ductal subtype (15, 78.9%), with an average size of 9mm (range 4-23mm, SD 5.4). BI-RADS features exhibited were most often round (9, 45%) or irregular (9, 45%) in shape, not parallel in orientation (16, 80%), indistinct margins (8, 40%) and hypoechoic echogenicity (20, 100%). Additionally, 13 (65%) exhibited architectural distortion on the AUS unique 3D reconstructed coronal view. 17 cancers (85%) were visualized in at least 2 data sets (AP, medial and/or lateral).
Conclusion: Most cancers detected on AUS were stage 1, IDC and had BI-RADS features typically associated with malignancy. Important imaging features identified unique to AUS included visualization on at least two data sets and coronal view architectural distortion. No cancers were described as an oval, parallel, circumscribed mass, features associated with sonographic probably benign BI-RADS 3 findings.
Clinical Relevance Statement: AUS is often criticized for high false positive rates, though limited literature has been published on the unique features of AUS detected, mammographically occult cancers to aid radiologists in interpretation.
