AI identifies breast cancer up to six years before diagnosis in Swedish study

A Swedish study evaluated three commercial AI‑CAD systems on 88,963 mammograms from 31,394 women collected between 2008 and 2019 across four Swedish regions. The AI flagged subtle signals in prior images, identifying about 20 % of cancers six years before they were clinically diagnosed; detection rates rose to 25 % four years prior and 39 % two years prior.

The same AI‑based screening reduced late‑stage diagnoses, cutting interval cancers by 12 %, invasive cancers by 16 %, large tumours by 19 % and aggressive sub‑types by 27 %. It also lowered radiologists’ reading workload by 44 % without increasing false‑positive rates, with high‑risk cases receiving double reading.

Researchers from Lund University, including Kristina Lång, suggest AI should serve as a decision‑support tool alongside radiologists, potentially improving prognosis for women worldwide as breast‑cancer incidence is projected to rise.