A Multi-class Support Vector Data Description Approach for Classification of Medical Image

Abstract

A majority of effective approaches are presented in classification of mammography. In this paper we propose the Hyper sphere Multi-Class Support Vector Data Description (HSMC-SVDD) approach, in order to improve the classification accuracy and training speed when the categories have been increased to more than two classes. The main idea of the HSMC-SVDD is to extend a Hyper sphere One-Class SVDD (HSOC-SVDD) to a HSMC-SVDD as a novel kind of immediate multiple classifiers. Experimental results on the Mammographic Image Analysis Society (MIAS) dataset show that the average training time is 21.369 seconds, compared with the combined classifier proposed by Wei, the training speed has been improved from 10 to 20 seconds and the average testing time is 0.4281 seconds by our approach. And our method can provide 76.6929% classification accuracy.