A new two-feature GBAM-neurodynamical classifier for breast cancer diagnosis
Tijana T. Ivancevic, Lorick Jain, Murk J. Bottema · 2003
Like standard discrete artificial neural networks (ANNs), continual neurodynamical systems can be used for the classification and diagnosis of breast cancer. In this paper, a two-feature generalized bidirectional associative memory (GBAM) classifier is formulated in tensorial invariant form. It is implemented in Mathematica 3.0 and tested on two sample features (the radius and perimeter of cell nuclei in fine-needle aspiration slides) from the Wisconsin breast-cancer database. The classification accuracy obtained (86%), together with the invariance of the classification result upon the variation of the dimensions and output form of the neural activation fields, shows the potential classification ability of theoretical classifiers that are directly implemented in computer algebra systems.