Novel fully automated Computer Aided-Detection of suspicious regions within mammograms

In this paper we present a novel fully automated scheme for detection of abnormal masses by anatomical segmentation of Breast Region and classification of regions of Interest (ROI). The system consists of three main processing steps, we perform essential pre-processing to remove noise, suppress artifacts and labels, enhance the breast region, extract breast region by the process of segmentation and remove unwanted parts as Pectoral Muscle. After segregating the breast region, we use an Adaptive Segmentation Procedure based on Kmeans Clustering followed by a Merging Regions method. With the obtained Regions of Interest, the extraction of Statistical and Textural Features is done by using gray level co-occurrence matrices (GLCM) and a Decision Tree Classification is performed to isolate normal and abnormal regions in the breast tissue. If any suspicious regions are present, they get accurately highlighted by this algorithm thus helping the radiologists to further investigate these regions. A set of Mini-MIAS mammograms is used to validate the effectiveness of the method. The precision of the method has been verified with the ground truth given in database and has obtained sensitivity as high as 90%. The CAD system proposed is fully autonomous and is able to isolate different types of abnormalities and it shows promising results.