As Head Physician of the Department of Breast Radiology, Dr. Dembrower outlined how her team implemented artificial intelligence (AI) in breast screening following rigorous clinical trials and extensive testing. As a result, the deployment streamlined workflows and reduced radiologists’ workloads, eliminating the need for evening and weekend shifts to clear waiting lists. Her presentation formed part of a European expert session examining how AI can support breast radiology departments.
Why Sweden Turned to AI in Screening
In Sweden, women aged 40–74 receive invitations for breast screening every two years. At Dr. Dembrower’s site alone, around 80,000 women receive invitations annually, with 75–80% attendance. Traditionally, two radiologists double-read all mammograms, detecting about 70% of cancers through screening. However, due to a growing shortage of breast radiologists, the hospital explored whether AI could address these pressures without compromising care.
The ScreenTrustCAD Trial: Evidence in Action
As reported by healthcare-in-europe, launched in April 2021 and completed in June 2022, the ScreenTrustCAD trial evaluated whether AI could replace one reader in a double-reading model. The study included 55,581 women across three arms: two radiologists, one radiologist plus AI, and AI alone. While two radiologists detected 250 cancers, the AI-plus-one-reader approach identified 261. Importantly, detection of advanced cancers remained comparable across all arms. Based on these results, the hospital moved confidently toward implementation.
From Trial to Clinical Practice
After securing regulatory, ethical, and IT safeguards, the AI-supported workflow went live in June 2023. Since then, recall rates have fallen by 26%, false positives by 13%, and overall cancer detection has increased by 12.2%. Consequently, radiologists now focus more on diagnosed patients, with no backlogs or overtime work.
Expert Perspectives and the Road Ahead
Professor Pascal Baltzer of Vienna General Hospital reinforced these findings, citing trials across Europe and Asia. While he cautioned against over-enthusiasm, he concluded that evidence strongly supports AI as an assistant in screening. Ultimately, radiologists remain the decision-makers. As interval cancer data and legal clarity emerge, AI’s role will continue to evolve—integrated closely with clinical expertise.