Throughout the world, breast cancer is common cancer in women that contribute to high death amongst women. The early diagnosis and corresponding treatment can increase the possibilities of survival. In contrast, the challenging task is to detect the mass early in mammographic images, which is difficult due to noise and contrast. Mammography is the most capable technique used by radiologists frequently, which helps detect abnormal mass at an early stage; it is one of the methodologies to identify breast cancer. Here, a system is used to detect the tumor, with a modification in the grow cut algorithm. A new method is suggested based on segmentation by changing the modified grow cut algorithm by improving the region of interest. A modified grow cut algorithm has changed seed selection in mammogram images from manual into the semiautomatic way and has worked on non-defined borders. The earlier algorithm grows the region of interest only for the neighbor of current pixels and, consequently, a neighbor. Still, a change is made in the methodology for growing the region of interest within the class and between a neighbor cell's neighbors. This will ensure to get a more effective segmented area for abnormal mass than the previous method. The proposed technique is evaluated with the help of the mini M.I.A.S. database by considering circumscribed lesions, speculated lesions. Through result analysis, it is clear that the proposed technique gives better results for speculated, circumscribed lesions based on a comparison of ground truth images and segmented results.