Fractal Analysis of Contours of Mammographic Masses

T.M. Nguyen and R.M. Rangayyan (Canada)

Keywords

Breast cancer, Contour analysis, Fractal analysis, Fractal dimension, Shape analysis, Signatures of contours.

Abstract

Fractal analysis is widely used in image processing, both in characterizing shapes of objects and in assessing tex ture. Breast masses present shape and texture characteris tics that vary between benign masses and malignant tumors in mammograms. Limited studies have been conducted on the application of fractal analysis specifically for classify ing breast masses. The fractal dimension of the contour of a mass may be computed either directly from the two dimensional contour or from one-dimensional signatures derived from the contour. In this paper, we describe three methods to compute the fractal dimension of contours, and apply one of the methods to a database of 57 contours of breast masses. The 57 contours were obtained from 37 re gions of interest (ROIs) related to benign masses and 20 ROIs related to malignant tumors. A one-dimensional sig nature was derived from each contour by using the Eu clidean distance of each contour point to the centroid of the contour. The fractal dimension of the signatures, esti mated by the ruler method, indicated the possibility of dis criminating malignant tumors from benign masses with an accuracy of 95%.

Important Links:

Go Back