A novel technique for mammogram mass segmentation using fractal adaptive thresholding

Abstract

Digital mammogram is emerged as a most reliable screening technique for the early diagnosis of breast cancer and it paves an opportunity for researchers to develop novel algorithms for computer aided detection. Presence of clusters of microcalcifications as masses in mammograms is an important early indication of breast cancer. Fractal geometry is an efficient mathematical approach that deals with self-similar, irregular geometric objects called fractals. As the breast background tissues have high local self-similarity, which is the basic property of fractals, a new fractal method is proposed in this paper for the detection and segmentation of circumscribed masses from mammograms. The median filtering, label removal and contrast enhancement are done as pre-processing measures which makes the process of segmentation of masses, easier. The proposed technique then segments the circumscribed masses using Fractal adaptive thresholding with the application of morphological operations. This Fractal based mammogram mass segmentation is able to produce encouraging results that substantiate the merit of the proposed technique. © 2014 Springer Science+Business Media Singapore.