Automatic microcalcifications segmentation procedure based on Otsu's method and morphological filters

Abstract

Microcalcifications are present in a great number of malignant lesions, being considered as a significant sign of malignancy. However, they are detected in less than 50% of mammograms with carcinomas. Given this scenario, systems to highlight infra-clinic lesions, aiding the specialists to make diagnostics, have been studied. These Computer-aided diagnosis systems have been developed based on parameters extracted from microcalcifications. They are basically composed of three parts: segmentation, parameters calculation and lesions classification. This paper presents an automatic microcalcification segmentation method, based on Otsu's method and morphological filters. The proposed method was assessed with 236 regions of interest, captured from 54 images (300 dpi, 8 bits). Based on the joint opinion of two experienced radiologists, the algorithm adequately segmented at least 88.6% of the cases.