Neural-network-based decision making in diagnostic applications

We discuss the neural network (NN) approach as a new way of looking at the problem of decision making, and we illustrate our discussion with two examples from gynaecology and obstetrics: assessment of intrauterine growth retardation, and assessment of luteinizing hormone surge for predicting ovulation time. We use the noisy test measurements and estimate the rest of the values by the interpolative property of the NN structure. We discuss the following issues: the meaning of dependence and independence of tests in medicine; the NN approach in combination of multiple gynaecology and obstetric tests in order to make medical decisions: a comparison of the NN approach to statistical decision making in medical decisions.