SwedenHealthcare and life sciences2021–2026Lund University (MASAI trial)
AI-supported mammography screening trial
In a randomised trial of over 100,000 Swedish women, AI-supported screening found 29% more cancers and cut reading work by 44%
Problem
European breast screening programmes have each mammogram read by two radiologists, and breast radiologists are in short supply. Screening services need to keep detection high without adding reading work.
Approach
From spring 2021 the MASAI trial, led by Lund University researchers at four Swedish screening sites, randomly assigned women either to standard double reading or to AI-supported reading. The AI system, Transpara, rated each exam's risk: high-risk exams went to two radiologists and the rest to one, with AI marks shown to the readers. Results appeared in The Lancet Oncology (2023, interim safety), The Lancet Digital Health (2025, detection) and The Lancet (2026, final).
Results
Takeaway
AI allocated exams to radiologists, deciding how many readers each needed while retaining radiologists. False positives stayed roughly level (1.5% against 1.4%), and final results recorded 12% fewer interval cancers, the trial's main safety measure.
Categories
- Size
- Large enterprise
- Industry
- Healthcare and life sciences
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Computer vision
- Adoption
- Vendor solution
- Evidence
- Independent study
Sources
Source summaries
The university reports results from almost 106,000 women in the MASAI trial published in The Lancet Digital Health. AI-supported screening found 29% more cancers (338 against 262), including 24% more early-stage invasive cancers, while radiologists' workload fell by 44%. Seven more women were recalled in the AI group, about a 1% increase.
The radiology trade journal summarises the Lancet Digital Health paper and names the AI tool as Transpara. It reports 338 cancers among 53,043 women in the AI arm, a detection rate of 6.4 per 1,000 against 5.0, and a 44% cut in reading workload with no rise in false positives.
The oncology news service reports the final MASAI results in The Lancet. Interval cancers fell 12% (1.55 against 1.76 per 1,000 women) and aggressive-subtype cancers 27%, with false-positive rates of 1.5% and 1.4%. It recalls the 44% workload reduction from the 2023 interim analysis.
Full details are in the original sources.
