Accurate segmentation of the breast region with texture filter in mammograms for CAD applications

Abstract

The first and one of the most important preprocessing steps of the mammogram analysis is segmentation of the breast region. Accurate detection of breast tissue-air line allows focusing on ROI in cancer investigation, by detecting architectural distortion at the border, determination of the retraction location at the cancerous region under skin and, reducing the storage needs in PACS. The structural differences of the breast tissue itself, the variations in breast position, orientation, amount of compression and device characteristics in image acquisitions taken in different times, makes breast segmentation process challenging. In this study, by proposed texture filter method, above 99% value is achieved in completeness, accuracy and quality parameters.