Based on data from a series of patients who underwent screening mammography with a "normal" or "benign" first-reading result. The main objective of the study is to evaluate the possibilities of subtracting from the second reading of breast cancer screening a population of examinations for which the risk of discovering cancer is close to zero, using an AI system (here the MammoScreen™ system).

This is a retrospective multicenter study of mammography imaging data contained in the PACS (Picture Archiving and Communication System) of six radiology centers located in the Rhône, Puy de dôme, Cantal and Haute-Loire départements.

In order to be able to collect two-year follow-up data, the files analyzed will concern all women who received a screening mammogram during the years 2015, 2016, 2017, 2018 and 2019. More recent mammograms will be excluded.

Inclusion Criteria:

• Women aged 50 to 74 participating in the organized screening campaign.
• Having had a screening mammogram with a final L1 result of "normal" (ACR 1/ BI-RADS 1) or "benign" (ACR 2/ BI-RADS 2).
• For which a second reading interpretation is available
• Screening mammograms performed during the years 2015 to2019 (In order to have follow-up data (subsequent mammogram or information on the person's future)

Exclusion Criteria:

- Mammography not performed as part of organized screening

Female breast cancer

As this was a retrospective exploratory study, with no prior hypothesis, the number of subjects required was not calculated.

The study population (n=55,000) represents, for the radiology practices concerned, all mammograms examined in second reading by the Auvergne Rhône Alpes regional cancer screening coordination center between 2015 and 2019.