08/08/2026
Despite evidence that breast imaging specialists outperform general radiologists, about 70% of screening mammograms are interpreted by generalists due to specialist shortages. AI-supported workflows may help address variability and standardize screening mammography performance.
In a new Radiology (RSNA Journal) article, Matthew McCabe, PhD, DeepHealth in Somerville, MA, and colleagues evaluated whether a multistage AI-driven workflow could improve screening performance of generalists and specialists.
The prospective study included 60 generalists and 35 fellowship-trained specialists who interpreted 577,742 mammograms. The AI workflow incorporated a computer-aided detection and diagnosis tool that routed AI-identified suspicious examinations not initially recalled for additional expert review.
With AI, generalists’ cancer detection rate increased from 3.76 to 4.99 cancers per 1,000 mammograms, and positive predictive value increased from 3.38% to 3.89% despite a higher recall rate (9.06% to 10.40%). Specialists’ detection rates were unchanged, and generalists using AI performed similarly to specialists using AI.
“The benefit of improved positive predictive value of recalls indicates a favorable shift in the screening process despite the observed increase in recall rate,” the authors conclude.
Read the full article: https://bit.ly/4frs1xu