Validation of a large medical database

Abstract

Complex clinical problems involving huge experimental evidence require a preliminary validation of observed data. This may avoid biasing due to incorrect sampling and clarify the sample distribution by showing data-inherent regularities. The paper describes the application of unsupervised models of neural networks to the analysis of a very large set of clinical records for the study of osteoporosis. The main result obtained lies in showing the overall uniformity of the data distribution, which indicates a correct unbiased sampling of the considered population.<>