One of the most important steps in digital mammography is an adequate segmentation of possible abnormalities. This obviously minimizes errors in further stages such as in classification. However, several factors affect the proper segmentation of mammograms. Mammograms contain low signal to noise ratio (low contrast) and a complicated structured background.In this article we are describing a generic approach for detecting patterns of architectural distortions in mammograms that is both complete and uncommitted to any type of training. Our detection algorithm dynamically updates the pixels intensities by following their neighboring transition zone. Such approach proved to be effective for detecting the edges of all types of breast abnormalities including the Stellate.