Morphological Enhancement of Microcalcifications in Digital Mammograms

Abstract

Mammography is a commonly used technique for early detection of breast cancer. In mammograms, microcalcifications show low contrast margin with the background parenchymal tissue (specifically when the background tissue type is fibroglandular) as a result, subjective analysis of these calcifications with respect to their size, shape and morphology presents a daunting challenge even for experienced radiologists. Thus the present work investigates the potential of two morphological techniques i.e., top-hat morphological processing and h-dome morphological processing for enhancement of microcalcifications embedded in variety of background tissue types including fatty, glandular and fibroglandular tissues while restoring their shape and size. The enhancement results are also compared with standard contrast limited adaptive histogram equalization method. For subjective analysis, 25 synthetic images with simulated microcalcifications of various shapes and sizes are used. Objective analysis is carried out on 50 mammographic images taken from benchmark dataset (McGill University mammographic database) by computing quantitative indices like contrast improvement ratio and detail variance/background variance ratios. After rigorous experimentation on both synthetic and benchmark data set it was observed that h-dome morphological processing (with h = 60) is ideally suited for enhancement of microcalcifications while restoring their shape and size.