Shape and Boundary Analysis for Classification of Breast Masses
Zhou Weiqiang, Xiangmin Xu, Wei Huang · 2008
Malignant breast tumors appear spiculate or microlobulate in the boundary and irregular in shape. But benign breast masses appear smooth in the boundary and round in shape. We used polygonal modeling to draw Index of spiculation(SI), index of lobule(IF), measure of fractal dimension (FD) and measure of circularity (C) to represent the characteristic of the boundary and the shape of breast masses. The boundary of the mass is divided into three type: 1. spiculate; 2. microlobulate; 3.smooth.The shape of the mass is divided into two types: 1. irregular; 2. sub-circular. Considering the boundary and the shape style the masses can be divided into malignant ones and benign ones. The test is based on a dataset of 93 images from MIAS with 54 benign masses and 39 malignant tumors. The accuracy of the classification reach 0.9265 in terms of the area(Az) under the ROC curve.