Automated detection methods for architectural distortions around skinline and within mammary gland on mammograms

TLDR: Two detection approaches for architectural distortions existing around skinline and within mammary glandular tissues are developed and it is concluded that these methods were effective to detect architectural distortions.

Abstract: The architectural distortion is a very important finding in interpreting breast cancers as well as microcalcification and mass on mammograms. However, it is more difficult for physicians to detect architectural distortion than microcalcification and mass. The purpose of this study is to develop two detection approaches for architectural distortions existing around skinline and within mammary glandular tissues. The detection methods for depressed areas around skinline consisted of three steps. The binarization technique was performed to extract the mammary gland region. In order to determine suspect areas, the top-hat processing was performed. The false positives were eliminated by the features of their sizes and positions. The distorted areas within mammary gland region were detected by the following steps. The structure of mammary gland was extracted by using curvature. The candidates were determined by concentration index. The false positives were eliminated by their isotropy indexes, sizes, pixel values and contrast. Our image database consisted of 17 cases with focal retraction areas around skinlines (case A) and 38 cases with architectural distortions within mammary glands (case B). As a result, the sensitivities were 94% and 84% in case A and case B, respectively. It was concluded that our methods were effective to detect architectural distortions.