Threshold Based Segmentation Technique for Mass Detection in Mammography

Breast cancer is second leading cause of death among women. Mass of the cancer is initially originates from a single cell but slowly increases in size by rapid multiplication of cells to produces symptoms. Most of the time cancer symptoms are identified at the late stage, when the tumor becomes bigger in size and treatment becomes invasive. Early detection of the cancer before the development of the symptoms may help in less number of modalities for the treatment. Screening is the basic procedure for identification of breast cancer at an earliest and mammography is an efficient screening method, in which abnormalities can be detected. However, it is difficult to identify the tumor in the breast tissue because tumors possessequal intensityin the breast tissue and appears poor in contrast. Hence, the computer aided detection helps for physicians and radiologist to find abnormality at an earliest in the absence of any symptoms. In this study, we used segmentation algorithm to develop an efficient system to find abnormality at the earliest stage. The proposed segmentation algorithm detects clearly defined region of mass using morphological threshold based segmentation technique. The efficiency of the algorithm is measured with 55 images of Mini-MIAS database. These results showed satisfactory segmentation and the accuracy of the algorithm is 94.54% in identification of mass in mammography and false identification rate 5.45%. Thus, the proposed method is compared with traditional Otsu thresholding method,which is more effective comparing to Otsu thresholding segmentation results.