A semi-supervised fuzzy GrowCut algorithm to segment and classify regions of interest of mammographic images

摘要

According to the World Health Organization, breast cancer is the most commonform of cancer in women. It is the second leading cause of death among womenround the world, becoming the most fatal form of cancer. Mammographic imagesegmentation is a fundamental task to support image analysis and diagnosis,taking into account shape analysis of mammary lesions and their borders.However, mammogram segmentation is a very hard process, once it is highlydependent on the types of mammary tissues. In this work we present a newsemi-supervised segmentation algorithm based on the modification of the GrowCutalgorithm to perform automatic mammographic image segmentation once a region ofinterest is selected by a specialist. In our proposal, we used fuzzy Gaussianmembership functions to modify the evolution rule of the original GrowCutalgorithm, in order to estimate the uncertainty of a pixel being object orbackground. The main impact of the proposed method is the significant reductionof expert effort in the initialization of seed points of GrowCut to performaccurate segmentation, once it removes the need of selection of backgroundseeds. We also constructed an automatic point selection process based on thesimulated annealing optimization method, avoiding the need of humanintervention. The proposed approach was qualitatively compared with otherstate-of-the-art segmentation techniques, considering the shape of segmentedregions. In order to validate our proposal, we built an image classifier usinga classical multilayer perceptron. We used Zernike moments to extract segmentedimage features. This analysis employed 685 mammograms from IRMA breast cancerdatabase, using fat and fibroid tissues. Results show that the proposedtechnique could achieve a classification rate of 91.28\% for fat tissues,evidencing the feasibility of our approach.