Multicenter Mammography Study Shows Greater Than 10 Percent Increase in Breast Cancer Detection with Adjunctive AI

Can adjunctive AI have an impact in single-read settings for screening mammography programs?

In a recent prospective multicenter study, recently published in Nature Communications, researchers compared cancer detection rates (CDRs) and recall rates (RRs) for unassisted digital mammography (DM) interpretation and the use of adjunctive artificial intelligence (AI) software (Lunit Insight MMG, version 1.1.7.1, Lunit) in 24.543 women (median age of 61). The study authors noted that 67.5 percent of the cohort had dense breasts.

Researchers found that the combination of breast radiologist evaluation with adjunctive AI led to 140 screening-detected breast cancer cases (5.7 percent CDR) in comparison to 123 cases detected by unassisted breast radiologists (5.01 percent CDR). Noting a 13.8 percent increase in CDR with adjunctive AI, the study authors said there was also no statistically significant difference in RRs between the use of adjunctive AI and unassisted assessment by breast radiologists (1,113 recall cases, 4.53 RR percent vs. 1,100 recall cases, 4.48 RR).

“The assistance of AI-CAD that led to improved CDRs did not affect RRs, providing reassurance to radiologists when using AI-CAD in their routine practice in a single-reading setting,” wrote lead study author Yun-Woo Chang, M.D., who is affiliated with the Department of Radiology at Soonchunhyang University Seoul Hospital in Seoul, Korea, and colleagues.

In an exploratory analysis, the study authors also found that the adjunctive AI software enhanced breast cancer detection for general radiologists (120 cases, 4.89 percent CDR) compared to unassisted mammography interpretation (95 cases, 3.87 percent CDR).

However, with general radiologists, the researchers noted the use of adjunctive AI led to 1690 recall cases (6.89 percent RR) vs. 1,548 recall cases (6.31 percent RR) with unassisted reading.

“Although comparison with prior mammograms was possible in this simulation analysis similar to a real clinical setting of BR reading, these results are consistent with (general radiologists) relying more on AI-CAD results, given their relatively lower self-confidence in interpreting mammography compared to (breast radiologists), which seems to have induced increased false-positive RRs,” explained Chang and colleagues.

(Editor’s note: For related content, see “New Mammography Studies Assess Image-Based AI Risk Models and Breast Arterial Calcification Detection,” “Study: Mammography AI Leads to 29 Percent Increase in Breast Cancer Detection” and “Can Mammography-Based AI Enhance Breast MRI Use in Patients with Intermediate Risk for Breast Cancer?”)

In regard to study limitations, the authors noted the observational nature of the trial and a relatively short follow-up period. The researchers pointed out that digital mammography was used to assess the impact of the AI software but conceded the increasing use of digital breast tomosynthesis (DBT) in breast cancer screening.