Breast screening: pitfalls and potential solutions

Breast cancer prevalence increases with age, lifestyle and genetic predisposition. Yet, the approach to disease detection is simply volume based. About 40 million mammograms are performed each year to detect about 300,000 cancer cases.

This is not efficient and leads to predictable problems. For example, 5% of the exams are deemed suspicious, but only a tenth of these (0.5%) actually have disease. Despite the tendency to over-diagnose, screening programs still miss up to 20% of the resident disease. While this error rate might be tolerable in a low-risk population, it quickly becomes intolerable where risk is elevated.

In this presentation, I will describe our approach to "solving" the errors problem in the legacy system and our plans for tailoring detection strategy to a population's risk characteristics.