AI helped spot breast cancers earlier, was linked to fewer dangerous cases emerging between screenings

The findings suggest AI could strengthen screening by reducing the number of breast cancers that are missed and diagnosed later, when they are often more aggressive and harder to treat. Photo / Getty Images

Artificial intelligence helped spot breast cancers earlier and was linked to fewer dangerous cases emerging between routine screenings in a Swedish trial of more than 100,000 women.

The findings suggest AI could strengthen screening by reducing the number of breast cancers that are missed and diagnosed later, when they are often more aggressive and harder to treat.

Interval cancers - those diagnosed after a woman has received a clear screening result and before her next routine test - are “a very good measure on how effective the screening method is”, said lead author Kristina Lang, a breast radiologist and clinical researcher at Lund University.

“Since it takes a long time to evaluate breast-cancer mortality, interval cancer rate has been used as a surrogate measure.”

The trial is the first randomised controlled study to examine how AI affects breast-cancer screening outcomes and is the largest test of AI in cancer screening to date.

Earlier results from the same project showed the technology can increase cancer detection while reducing the workload for radiologists.

What had remained unclear was whether those gains meant fewer cancers would appear later, after screening appeared to show no disease.

The results, published in the Lancet, come from a trial embedded in Sweden’s national breast-cancer screening programme.

Almost 106,000 women were assigned to either standard screening, in which mammograms were read independently by two radiologists, or to screening supported by AI.

Early detection

In the AI group, software reviewed images alongside radiologists, flagging higher-risk scans for extra scrutiny while allowing lower-risk images to be read once instead of twice.

Over the two years following screening, fewer women in the AI group were diagnosed with breast cancer after a clear scan.

The rate of such cases was about 12% lower than among women whose mammograms were read without AI.

Women screened with AI support were also less likely to be diagnosed later with invasive tumours or aggressive cancer subtypes, the researchers found.

Importantly, using AI didn’t worsen screening performance or increase false alarms. The rate of false positives was similar in both groups.

“Our study does not support replacing healthcare professionals with AI,” Jessie Gommers, a PhD candidate at Radboud University Medical Centre in the Netherlands and first author of the paper, said in a statement.

“However, our results potentially justify using AI to ease the substantial pressure on radiologists’ workloads.”

Economic impact

Whether AI-supported screening is worth the cost remains an open question.

Previous modelling from Norway suggested the technology could be cost-effective if it helped prevent even a small share of cancers from being missed.

The Swedish trial showed a larger drop than that benchmark, though a full economic analysis of the data is still underway, Lang said.

The researchers said the study has limitations.

It was conducted within Sweden’s screening programme, using a single type of mammography machine and one AI system, and involved mostly experienced radiologists.

The results would need to be confirmed in other countries and health systems, they said.

Further follow-up will also be needed to understand the long-term effects of AI-supported screening, including whether earlier detection reduces the need for aggressive treatment over time.

Sign up to Herald Premium Editor’s Picks, delivered straight to your inbox every Friday. Editor-in-Chief Murray Kirkness picks the week’s best features, interviews and investigations. Sign up for Herald Premium here.