Particle.news
Download on the App Store

AI Flags Mammogram Signs Years Before Breast Cancer Diagnosis

Authors say tracking serial AI scores could enable more personalized screening pending prospective trials to confirm clinical benefit.

Overview

  • A Radiology paper published in June 2026 reports that three commercial AI computer‑aided detection systems assigned higher cancer risk scores to mammograms years before radiologist diagnosis.
  • The team retrospectively applied the tools to 88,963 mammograms from 31,394 patients in the Swedish VAI‑B database and found elevated AI scores in people later diagnosed with cancer.
  • At a fixed specificity of 90 percent the systems flagged up to 19.7 percent of eventual cancers about six years before diagnosis and higher proportions at four and two years.
  • The study is retrospective and enriched with a high case proportion, so authors warn the findings do not prove earlier cancers would be confirmed in real time and call for prospective validation.
  • If confirmed, serial AI scores could guide closer surveillance for some people but would also require plans to manage false positives, vendor variability, workflow changes, and regulatory review.