Artificial intelligence (AI) can identify mammographic changes linked to breast cancer years before doctors make a diagnosis, according to a major Swedish study involving Greek researchers.
Published in Radiology, the research analyzed 88,963 mammograms from 31,394 women and assessed three commercially available AI systems. The systems identified warning patterns associated with future breast cancer as early as six years before diagnosis, while elevated risk scores appeared in some cases up to 10 years earlier.
Greek radiologist Pantelis Gialias took part in the research during his doctoral work at Linköping University Hospital in Sweden. Apostolia Tsirikoglou, a Greek researcher at Karolinska Institute, also contributed to the study.
Greek researchers find AI detected cancer warning signs years earlier
The study drew on mammograms collected between January 2008 and April 2019 across four Swedish regions. Among the 31,394 women included, 12,072, or 38.5 percent, received a breast cancer diagnosis during the study period. Their median age at screening was 57.6 years.
At a specificity level of 90 percent, the three AI systems identified between 19 and 19.7 percent of cancers six years before diagnosis. Four years before diagnosis, that figure rose to between 23.3 and 25.2 percent. Two years before diagnosis, the strongest-performing system identified up to 39.3 percent of future cases.
Elevated scores also appeared 10 years before diagnosis in 12.7 percent, 13.8 percent, and 17 percent of cases, depending on the model. The findings suggest that mammograms can contain measurable signs associated with future breast cancer long before a tumor becomes clear enough for a conventional diagnosis.
Changes too subtle for the human eye
Gialias explained that AI can detect small changes in breast tissue that may not appear significant enough for a radiologist to classify as suspicious. These can include subtle alterations in breast architecture or gradual changes in breast density.
On their own, such signs may not provide enough evidence to indicate cancer. AI, however, can identify patterns associated with tumors that later become clinically detectable.
The findings do not mean that AI can definitively diagnose breast cancer a decade in advance. Instead, the systems can flag mammograms linked to a higher likelihood of a future diagnosis. Women who later developed breast cancer generally received higher AI scores, and the systems performed better than breast density alone in predicting future cases.
Greek researchers say AI could reshape cancer screening
The findings could eventually influence how doctors monitor women after routine mammography. Patients whose scans repeatedly produce high or rising AI risk scores could undergo more frequent mammograms or additional imaging instead of following the same screening interval as women at lower risk.
Sweden currently invites women between the ages of 40 and 74 to undergo mammography every two years. Under the traditional system, two radiologists independently review each examination.
Gialias has stressed that AI should support radiologists rather than replace them. Doctors would remain responsible for interpreting results and deciding whether further examinations are necessary.
Previous research involving Gialias found that AI could reduce radiologists’ mammography workload by as much as 34 percent. Another study suggested that AI-assisted screening could lower costs compared with the traditional model in which two radiologists review every mammogram.
More research needed before routine use
Because the latest study was retrospective, researchers worked with earlier mammograms from women whose eventual outcomes were already known. Prospective studies will therefore be needed before AI-generated risk scores become part of routine breast cancer screening.
The research team also plans further work involving women with breast implants, whose mammograms can present additional interpretation challenges.
Overall, the findings strengthen evidence that AI could help identify women at increased risk years earlier and allow doctors to focus closer surveillance on those most likely to benefit.
See all the latest news from Greece and the world at Greekreporter.com. Contact our newsroom to report an update or send your story, photos and videos. Follow GR on Google News and subscribe here to our daily email!
