A neuro-fuzzy architecture in a biomedical application

The aim of this paper is to present a neuro-fuzzy architecture for information processing. The main characteristic of this approach is given by its ability to process fuzzy sets and linguistic terms, preserving the simplicity and the potentiality of the connectionist method. The use of a particular notation of the trapezoidal fuzzy sets has permitted a significant enhancement and simplification of the learning algorithm. The proposed architecture has been tested in a complex task of biomedical field, the breast cancer classification. The Wisconsin breast cancer database has been considered. Two strategies have been investigated for capturing the complexity of the information present in the description of input features. The classification accuracy index was used in the validation procedure. The positive results confirm the potentiality and the effectiveness of the proposed architecture.