Structured neural network of medical diagnosis for hepatobiliary diseases

This study aims to apply a structured neural network system for improving diagnostic capabilities of hepatobiliary diseases. Several neural networks were organized according to a priori knowledge. A priori knowledge consists of five pathological states. We compared the diagnostic accuracy of discriminant function, ordinary and structured neural networks. The diagnostic capabilities among testing data were 96% by structured neural network, 70% by ordinary neural network, and 66% by discriminant function. The structured neural network had the significantly higher diagnostic accuracy than the other two methods.