Breast cancer is the most common cancer diagnosis in women. Despite advances in treatment, organised, population-based, and quality-assured mammography screening programmes remain an important component in reducing the disease burden. The "donna" programme of the Cancer League of Eastern Switzerland performs over 50,000 mammographies annually across six cantons.
The mandatory double reading by two radiologists increases the breast cancer detection rate but places a considerable workload on radiological staff. In times of rising healthcare costs, screening programmes are under pressure to reduce costs and improve efficiency. Prospective studies demonstrate that Artificial Intelligence (AI) can improve diagnostic accuracy and be deployed in a cost-effective manner. A prerequisite is appropriate calibration of the AI, which improves both sensitivity and specificity and prevents an increase in cases referred to the consensus conference or to further diagnostic work-up. Possible implementation approaches include replacing one of the two human readings with AI, AI-based triage for pre-selection of mammographies, and the use of AI as a third independent reading. Which approach is optimal remains unresolved. Furthermore, the use of AI opens perspectives for risk-stratified, personalised screening based on automated breast density measurements and AI-generated risk predictions.
As international studies differ from the Swiss setting in terms of workflows, mammography equipment, target age groups, and population composition, their findings cannot be directly transferred. The present research project therefore investigates various AI algorithms using an (Eastern) Swiss screening cohort, with the aim of identifying the solution and its optimal implementation that achieves the highest efficiency while maintaining at least equivalent diagnostic quality.
The findings are intended to contribute to the evidence base regarding the conditions under which a reduction in human readings is feasible, and thereby provide a foundation for an evidence-based review of the existing Swiss requirement for double reading of every screening mammography.
Prof. Dr. Alexander Geissler, Dr. David Ehlig, Marcel Blum, Franziska Ausborn
Krebsliga Ostschweiz
Blumenau-Léonie-Hartmann Stiftung
June 2026 – November 2027