Abstract
Screening mammography is currently proposed according to patient age, and combined with ultrasound in the case of dense breasts. Imaging developments towards 3D tomosynthesis optimize detection, reduce false positives and limit irradiation. However, outside of cases of high genetic risk, it remains difficult to refine the exploration strategy according to lower-impact risks (family history, alcoholism, etc.). But new, more personalized methods are being studied to determine the optimum imaging strategy for each patient.
At the same time, we need to standardize mammography performance to give all women the same chance. Artificial intelligence techniques such as deep learningcan make this possible by automatically learning to build radiological signatures of cancers. The best algorithms now match the performance of human radiologists. In the future, they will use images to calculate the individual probability of cancer progression, enabling us to optimize patient care.
