3D matrix pattern based Support Vector Machines for identifying pulmonary cancer in CT scanned images

A novel algorithm of Three Dimension matrix (3D matrix) pattern based Minimum Within-Class Scatter Support Vector Machines (MCSVMs(3Dmatrix)) is presented. Combining Minimum Within-Class Scatter Support Vector Machines (MCSVMs) and higher-order tensor technology, decision functions of MCSVMs(3Dmatrix) are calculated along with three orthogonal directions in the 3D space. And then the final decision is made by Majority Vote Method. In previous reports, each CT image is solely processed and the relation among successive CT scanned images is neglected. The case results in defective judgment at whiles. The proposed method solves the problem effectively and improves the accuracy of classification to a certain extent.