Thursday, November 30 | 8:20 a.m.-8:30 a.m. | R1-SSBR10 | Room E450B
Attendees will learn about how AI can classify interval breast cancers as minimal signs and reading-error misses in this session.
In her presentation, Tiffany Yu, MD, from the University of California, Los Angeles will discuss her team's findings, which show that AI could aid in identifying and reducing rates of interval cancer. Although European studies have highlighted how AI can reduce interval breast cancers, the researchers pointed out that, to their knowledge, no comparable U.S.-based study exists. Given different screening practices, the Yu team reviewed and classified interval breast cancers to evaluate the performance of AI within U.S. breast cancer screening guidelines.
It included data from 184,935 screening mammograms, 65% coming from full-field digital mammograms and 35% from digital breast tomosynthesis (DBT) exams. From the total, the team reported 150 interval breast cancers in 150 women, with an age range of 40-87 years.
Of the cancers, nine were classified as interval cancers, of which four were flagged by AI. The researchers also found that AI flagged 71% of findings that were minimal signs-nonactionable, 89% of findings that were minimal signs-actionable, and 94% of missed reading errors.
The team also reported that women aged 40-49 years were disproportionately affected by occult interval cancers (42%) compared with women aged 50 years and older with minimal signs-actionable (27%, p = 0.007).
"Understanding context-specific interval breast cancer classification -- in this case, U.S.-based screening guidelines -- is critical for assessing AI performance on improving screening mammography sensitivity," the team wrote in its abstract.
Find out what else the team discovered by attending this session.