E5270. Look Again! Reducing False Positive Recalls on Digital Breast Tomosynthesis
Authors
Iris Chen;
University of California, Los Angeles
Melissa Joines;
University of California, Los Angeles
Anne Hoyt;
University of California, Los Angeles
Cheryce Fischer;
University of California, Los Angeles
James Chalfant;
University of California, Los Angeles
William Hsu;
University of California, Los Angeles
Hannah Milch;
University of California, Los Angeles
Background
False positive recalls on screening mammography can significantly increase healthcare cost and patient anxiety. While the literature suggests that digital breast tomosynthesis (DBT) can reduce recall rate, the 10-year false positive recall rate has been estimated to be about 50%, and in clinical practice, recall rate may actually increase compared to that of 2D mammography.
Educational Goals / Teaching Points
This exhibit will include a comprehensive review of imaging features that contribute to false positive recalls on DBT with tips on distinguishing between true and false positive recalls. A quality improvement (QI) module that can be used in academic and private practices to reduce false-positive recalls will be included. The rate of change expected over serial mammograms and effect of length of interval follow-up on recall rates will be discussed with case examples. Lastly, we will include an overview of the current literature and preliminary results from a large-scale single institution study comparing false positive recalls by radiologists compared to artificial intelligence (AI).
Key Anatomic/Physiologic Issues and Imaging Findings/Techniques
Circumscribed masses, benign calcifications, architectural distortion, and intramammary lymph nodes can be more conspicuous on tomosynthesis and may be more easily called back. Additionally, spiculated or stellate masses and architectural distortions have been shown in multiple trials including the STORM, Malmö, ASTOUND, and Oslo trials, to be more conspicuous on tomosynthesis. QI initiatives including the use of an educational module with biopsy-proven findings and a second mammography-reader can help increase specificity. The high NPV of commercially available AI algorithms also has the potential to reduce false positive results if the radiologist can trust the AI system in real-world practice.
Conclusion
It is crucial for breast radiologists to know how to better interpret DBT to decrease false positive recalls and unnecessary healthcare costs.
