Automatic Breast Pectoral Muscle Segmentation on Digital Mammograms Using Morphological Watersheds

The segmentation of mammograms plays a major role in isolating areas which can be subject of tumor. The identification of these zones is generally done in three steps: pectoral muscle segmentation, hard density zone detection and texture analysis of regions of interest. This paper deals with the segmentation of the pectoral muscle on a mammography image, in order to facilitate the work of experts while analyzing. The developed methodology is based on Morphological Watersheds. This algorithm has been tested on 80 digital mammography images of MIAS database. The performance is evaluated based upon the false positive (FP), false negative (FN) pixel percentage, and mean distance. All the average FN and FP pixel percentages are 3.68% and 2.98%, with the range shown from 0.90 to 0.99 for accuracy and 0.86 to 0.99 for precision rate. The method is also compared with three well-known pectoral muscle detection techniques and in most of the cases; it outperforms the other three approaches