A U.S. Food and Drug Administration (FDA)–cleared artificial intelligence (AI) algorithm was able to detect and correctly localize almost one-third of interval breast cancers in a retrospective evaluation of screening digital breast tomosynthesis (DBT), findings published in Radiology showed. The study authors believe that AI could help further decrease the interval cancer rate.
“Our study shows that an AI algorithm can retrospectively detect and correctly localize nearly one-third of interval breast cancers on screening DBT exams, suggesting its potential to reduce the interval cancer rate and improve screening outcomes,” stated study author Manisha Bahl, MD, MPH, the Director of Breast Imaging Division Quality and Co-Service Chief at Massachusetts General Hospital and Associate Professor at Harvard Medical School. “These findings support integrating AI into DBT screening workflows to enhance cancer detection, but its real-world impact will ultimately depend on radiologist adoption and validation across diverse clinical environments.”
Study Methods and Rationale
Interval breast cancers are cancers that are diagnosed in between regular mammography periods due to symptomatic presentation after a false-negative screening. These cancers are typically more aggressive with rapid growth. 3D mammography with the new DBT approach could potentially reveal some of these fast-growing cancers previously hidden by dense breast tissue, but long-term data with this approach are limited.
“Given the lack of long-term data on breast cancer–related mortality measured over 10 or more years following the initiation of DBT screening, the interval cancer rate was often used as a surrogate marker,” Dr. Bahl explained. “Lowering this rate is assumed to reduce breast cancer-related morbidity and mortality.”
Researchers conducted a retrospective analysis of the screening DBT examinations performed prior to 224 breast cancer diagnoses made between February 2011 and June 2023. The DBT exams were analyzed with the FDA-cleared INSIGHT DBT v1.1.0.0 AI algorithm, supplied by Lunit.
“Several studies have explored the use of AI to detect interval cancers on screening 2D digital mammography exams, but to our knowledge, no previously published literature has focused on the use of AI to detect interval cancers on DBT,” Dr. Bahl noted.
The AI-indicated lesions were scored between 0 and 100, with the highest score being the examination-level score. Two breast imaging radiologists also conducted independent reviews to confirm the accuracy of the AI annotations. “An exam-level analysis gives AI credit for any positive exam, even if its annotation is incorrect or unrelated to the actual cancer site, which may inflate the algorithm’s sensitivity,” Dr. Bahl explained. “Focusing on lesion-level accuracy provides a more accurate reflection of the AI algorithm’s clinical performance.”
Then AI was also tasked with analyzing 1,000 true-positive, true-negative, and false-positive screening DBT examinations.
Key Study Findings
Of the 224 interval cancers analyzed, AI correctly localized 32.6% of the lesions on the DBT examinations. “My team and I were surprised to find that nearly one-third of interval cancers were detected and correctly localized by the AI algorithm on screening mammograms that had been interpreted as negative by radiologists, highlighting AI’s potential as a valuable second reader,” Dr. Bahl said.
The interval cancers that AI correctly detected were larger at surgical pathology (P < .001) and had axillary lymph node positivity (P = .01). “These findings suggest that AI may preferentially detect more aggressive or rapidly growing tumors, or that it identifies missed cancers that were already advanced at the time of screening,” Dr. Bahl said.
In the subsequent analysis of 1,000 DBT examinations, the AI algorithm correctly localized 84.4% of the true-positive cancers, and correctly categorized 85.9% of the true-negative and 73.3% of the false-positive cases as lacking cancer.
Disclosure: For full disclosures of the study authors, visit pubs.rsna.org.
