Artificial intelligence assistance may significantly improve the identification of breast cancer during screening using digital breast tomosynthesis (DBT), according to research released February 20 in the Journal of the American College of Radiology.
The study found that when radiologists interpreted DBT scans with the help of AI software, detection of several types of breast cancer increased. These included invasive cancers, lobular cancers, and tumors occurring in women with dense breast tissue. The technology also helped clinicians identify smaller tumors. The investigation was led by Kathy Schilling, MD, of Baptist Health in Boca Raton, Florida.
“Use of AI was able to detect more invasive cancers without an increase in noninvasive cancers or recall rate,” the Schilling team highlighted.
Previous studies have already established that DBT screening improves cancer detection and reduces patient recalls compared with standard mammography. However, DBT produces a much larger number of images, and reviewing these can lengthen reading time for radiologists, potentially contributing to fatigue and burnout.
Although earlier research has suggested potential benefits of artificial intelligence, the authors noted that many prior evaluations were limited and not widely applicable to routine clinical practice. Because of this, they emphasized the importance of assessing AI performance in real-world screening environments.
To explore this, the researchers compared screening outcomes before and after introducing an AI detection system at an outpatient breast imaging center. They examined cancer detection and recall rates and also analyzed differences related to breast density, tumor size, disease stage, and pathology.
The retrospective review used mammography audit data collected from four facilities across two timeframes: 2018–2020 (before AI implementation) and 2020–2022 (after AI adoption). Nine specialized breast radiologists participated in interpreting the studies.
During the pre-AI phase, 54,440 examinations were performed, producing 339 confirmed cancer detections. In the AI-assisted period, 48,742 screenings resulted in 369 confirmed cancers. After AI implementation, the number of cancers detected increased, recall rates remained comparable, more malignancies were discovered in dense breast tissue, and invasive tumors were generally smaller at diagnosis.
The investigators also observed no meaningful rise in diagnoses of ductal carcinoma in situ (DCIS). According to the authors, this indicates that AI can enhance radiologists’ diagnostic accuracy without causing over diagnosis. Detecting cancers at a smaller size may also influence treatment decisions and improve patient outcomes.