Automatic detection of the retroareolar region in X-ray mammography images

Abstract

Mammographic image analysis is an important tool for the detection and assessment of breast cancer. Previous studies have shown that the performance of image analysis algorithms can be improved by applying them in the retroareolar (RA) region of the breast. However, previous works have relied on subjective, manual segmentation of the RA region. This paper presents a method for the fully-automated detection of the RA region in x-ray mammography images. The method is based on a curvilinear coordinate system that automatically adapts to the breast shape and size. Experiments using logistic regression analysis on images from a publicly available dataset show that the proposed method outperforms the traditional approach in the task of cancer detection.