Region Marking and Grid Based Textural Analysis for Early Identification of Breast Cancer in Digital Mammography

Most preferred method for early diagnosis of breast cancer is Mammography. Subjective diagnostics sometimes lead to wrong diagnosis and in some cases normal tissue can be diagnosed as malicious. We propose a method to reduce the human error by providing an objective analysis so that it could aid subjective analysis and increase the efficiency. First we extract the region of interest marked by radiologist and perform feature extraction, train the features using classifier. In the next stage we take random data from the database which is processed to remove x-ray annotation and pectoral muscle. The processed image is divided into small grids and features are extracted from every grid. The features from every grid is tested for suspicion, either malignant or benign. We have tested our algorithm on 36 data sets. 100% efficiency was observed for removal of x-ray annotation and 94.4% for pectoral muscle. Grid based textural analysis could classify the suspicious region with an efficiency of 91.67%. The efficiency for grid analysis has been tested manually based on the prior knowledge of presence of tumor, provided by the radiologist in MIAS database.