Breast cancer is the most common non-preventable cancer among women, with over 200,000 new cases in the United States each year. Conventional two-dimensional mammographic screening is estimated to detect 70-80% of breast cancers, but many cancers are still not detected early enough to permit a cure, and more than 40,000 women die from the disease each year. Conventional 2D mammography is currently the most effective screening tool for the early detection of breast cancer. The current breast cancer screening guidelines seek to manage the tradeoff between detecting cancers earlier on the one hand, and the dangers of the risk of increased radiation dose, the inconvenience and stress of false positive results, and the intrusive removal of small, slow-growing tumors that would not threaten the life of the patient. Statistically valid clinical trials to directly evaluate the effectiveness of different screening guidelines are necessarily expensive and time-consuming to conduct. As a result, there has been significant controversy surrounding these guidelines, and currently no consensus exists. We are designing improved breast cancer screening guidelines by creating a large synthetic cohort of women to which simulated screening policies can be applied. We have developed dynamic programming algorithms that can analytically derive optimal policies that seek to minimize cancer mortality and maximize life expectancy. The screening guidelines can be tailored to each individual woman’s primary risk factors, and the guidelines can evolve in response to non-stationary biological, environmental, detection and tumor growth parameters.