AI cuts interval breast cancers in Swedish trial

An AI tool cut interval breast cancers by 12 per cent in a Swedish screening trial of more than 105,000 women.

The study also found 27 per cent fewer aggressive breast cancers detected at screening when AI was used.

Interval cancers are cancers found between routine screening appointments because they were missed at the original scan. They are often more dangerous and linked to higher death rates than cancers found at screening.

The MASAI trial is described as the first large randomised study to test whether AI can improve mammography screening, which uses low-dose X-rays to examine breast tissue for signs of cancer.

The AI tool, called Transpara Detection and developed by ScreenPoint Medical, supported radiologists in analysing mammography images.

Earlier results from the same trial showed that Transpara Detection increased cancers found by 29 per cent and reduced radiologist workload by 44 per cent compared with standard double-reading, where two radiologists independently review each scan.

The latest findings indicate higher accuracy with AI support. Sensitivity, the ability to detect cancer, was 6.7 percentage points higher in the AI group while specificity, the ability to rule out healthy cases, was maintained. Results were similar across age groups and breast density levels.

Women screened with AI had 16 per cent fewer invasive interval cancers and 21 per cent fewer large interval cancers than those in the standard screening group.

The system also helps doctors assess risk more precisely by subdividing suspicious findings into BI-RADS 4 categories A, B and C. BI-RADS (Breast Imaging Reporting and Data System) is a standardised scale that guides whether a patient needs closer monitoring, further tests or treatment.