Realce de contraste para detecção de lesão em mamografia

Abstract

Breast cancer is the second most common type of cancer among women in Brazil and in the world, accounting for only 1% of cases in men. Mammography is the main exam for the early detection of this disease, and when detected early, the treatment can be performed more quickly, increasing the probability of cure. Breast cancer is difficult to detect due to the low contrast that the mammographic images present, difficulting the lesion visualization.Thus, through mammographic image processing algorithms, it is possible that image characteristics such as contrast can be manipulated using contrast enhancement techniques. In this context, this work aims to use contrast enhancement techniques based on histogram of images (HE – Histogram Equalization, CLAHE – Contrast Limited Adaptive Histogram Equalization and RMSHE – Recursive MeanSeparate Histogram Equalization) and to evaluate the performance of the algorithms in mammographic images of different breast densities, in the presence and absence of microcalcifications. We selected 40 images of the four different breast density classes with mammographic findings, being 10 images of each class (craniocaudal and medio lateral oblique incidences). The image processing and visualization were performed through MATLAB and ImageJ softwares, respectively. From the processing we also obtained metrics that quantify contrast and signal to noise ratio in the images, evaluating the techniques results. The analysis of the results showed that the RMSHE method obtained better results.