Combining fuzzy cognitive maps with support vector machines for bladder tumor grading

Abstract

Fuzzy Cognitive Map (FCM) is an advanced modeling methodology that provides flexibility on the system's design, modeling, simulation and control. This research work combines the Fuzzy Cognitive Map model for tumor grading with Support Vector Machines (SVMs) to achieve better tumor malignancy classification. The classification is based on the histopathological characteristics, which are the concepts of the Fuzzy Cognitive Map model that was trained using an unsupervised learning algorithm, the Nonlinear Hebbian Algorithm. The classification accuracy of the proposed approach is 89.13% for High Grade tumor cases and 85.54%, for tumors of Low Grade. The results of the proposed hybrid approach were also compared with other conventional classifiers and are very promising. © Springer-Verlag Berlin Heidelberg 2006.