Classification of Prostatic Cancer Using Artificial Neural Networks
Torsten Mattfeldt, Maximilian Burger, Hans‐Werner Gottfried, Hans A. Kestler · Birkhäuser Basel eBooks · 2002
A short review of artificial neural networks is provided. Such networks are information processing systems inspired by neurobiological models. The basic principles of two classical networks are informally reported: the multilayer perceptron and learning vector quantization. Artificial neural networks are primarily used for the purposes of prediction, pattern recognition, and process and machine monitoring. Clinico-pathological applications are related to prediction and classification in tumour pathology. For illustration, the authors report on two own studies: prediction of postoperative tumour progression in prostatic cancer from routine and morphometric data, and preoperative biopsy-based staging of prostatic cancer. In these studies better results were obtained by learning vector quantization as compared to multilayer perceptrons. Finally two methods of network validation are shortly described, and some open problems related to neural networks are exposed.