Use of artificial neural networks in modeling associations of discriminant factors: towards an intelligent selective breast cancer screening.

In order to improve the costs/benefits ratio of breast cancer (BC) screenings, the author evaluated the performance of a back-propagation artificial neural network (ANN) to predict an outcome (cancer/not cancer) to be used as classificator. Networks were trained on data from familial history of cancer, and sociodemographic, gynecoobstetric and dietary variables. The ANN achieved up to 94.04% of positive predictive value and 97.60% of negative predictive value. Results could operate as guidelines for preselecting women who would be considered as a BC high-risk subpopulation. The procedure--not only based on age factor, but on a multifactorial basis--appears to be a promising method of achieving a more efficient detection of preclinical, asymptomatic BC cases.