Can AI catch breast cancer earlier? New Lancet study shows promising results

A large Swedish trial published in The Lancet suggests AI-supported mammography can detect more breast cancers earlier and reduce aggressive interval cancers, without replacing radiologists

Breast cancer screening could soon get a smart new assistant. A large study from Sweden suggests that artificial intelligence can help detect breast cancer earlier during routine mammograms and reduce the number of cancers diagnosed later.

 

The study, titled Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study and published recently in The Lancet, analysed screening data from nearly 100,000 women undergoing routine mammography between April 2021 and December 2022.

 

Researchers found that AI-supported screening picked up more cancers during routine checks and was linked to fewer aggressive cancers detected later, pointing to a potential shift in how mammograms are read in the future.

 

The researchers randomly assigned women to one of two groups. In the standard group, each mammogram was read independently by two radiologists. In the AI-supported group, an AI system first analysed the images and assessed cancer risk. Scans deemed low-risk were read by one radiologist, while higher-risk scans were reviewed by two radiologists, with AI highlighting suspicious areas.

Did AI detect more breast cancers during screening?

The study found that about 81 per cent of breast cancers in the AI-supported group were detected during screening, compared with 74 per cent in the standard screening group. Earlier detection matters because cancers found at screening are often smaller, less advanced, and easier to treat.

 

Even more striking was the reduction in aggressive cancer subtypes. The AI-supported group had 27 per cent fewer aggressive cancers diagnosed later, suggesting that AI may help catch fast-growing tumours before they progress.

 

The study also showed that interval cancers, which are breast cancers diagnosed between scheduled screening rounds, often because they were missed earlier or developed rapidly, were lower in the AI-supported group compared with the standard group.

Does AI replace radiologists in breast cancer screening?

The researchers stressed that AI acts as a support tool, not a substitute for human expertise. They emphasised that radiologists remain central to decision-making, with AI helping to prioritise cases and flag potential concerns.

 

In effect, AI functioned like a highly attentive assistant, tireless, consistent and fast, but still reliant on trained clinicians to make final judgements.

 

While the results are promising, specialists say more evidence is needed from diverse populations and healthcare systems. Screening programmes differ widely across countries in terms of technology, training and follow-up care. Further studies will help clarify how well AI performs beyond Sweden and whether similar benefits can be replicated elsewhere.