A GROUNDBREAKING study in Sweden has found that AI can enhance breast cancer detection in mammography screening while significantly reducing the workload for radiologists.
The trial, conducted within Sweden’s national screening programme, involved 105,934 women across four screening sites (Malmö, Lund, Landskrona, and Trelleborg). Participants were randomly assigned to either AI-supported screening or the conventional method of double reading by radiologists. The AI system, Transpara (version 1.7.0), was used to triage scans for single or double reading and highlight suspicious findings.
Results showed that AI-assisted screening identified 338 cases of cancer among 53,043 women, compared to 262 detected in the conventional group of 52,872 participants. This represents a 29% increase in cancer detection (6.4 per 1,000 screened participants in the AI group versus 5.0 per 1,000 in the control group, p = 0.0021). Notably, AI helped detect more invasive cancers (270 vs 217 cases) and high-grade in situ cancers (68 vs 45 cases).
The recall rate (the proportion of women asked to return for further testing) was slightly higher in the AI group (1,110 recalls vs 1,027), but the increase was not statistically significant. Importantly, the rate of false positives remained virtually unchanged, ensuring that the AI system did not lead to unnecessary anxiety or additional testing.
A key advantage of AI-assisted screening was a substantial reduction in radiologists’ workload. The AI-supported approach resulted in 61,248 screen readings, compared to 109,692 in the standard group – a 44.2% reduction. This suggests AI could help alleviate pressure on healthcare professionals while maintaining or even improving detection rates.
The study’s findings highlight the potential of AI to improve early detection of clinically significant breast cancer while streamlining screening processes. As healthcare systems face increasing demand, AI could play a crucial role in enhancing efficiency without compromising patient outcomes.
Reference
Hernstrom V et al. Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI): a randomised, controlled, parallel-group, non-inferiority, single-blinded, screening accuracy study. Lancet. 2025;DOI:10.1016/S2589-7500(24)00267-X.
