COMPUTER AIDED DETECTION OF MAMMOGRAPHIC MASSES ON DIGITAL MAMMOGRAMS
Serhat Özekes, Ali Yılmaz Çamurcu · DergiPark (Istanbul University) · 2005
This paper presents an automated system for detecting masses in mammogram images.The proposed method is based on a two-step procedure: a. regions of interest (ROI) specification, b. rule based classification of regions of interest.In the first step, the intensity values of pixels in mammogram images are used and scanning the pixels in 8 directions is evaluated.By using various thresholds while scanning the pixels, ROIs are specified.In the second step, all ROIs are labeled using Connected Component Labeling (CCL) and two rules are used to categorize ROIs as true masses or not.These rules are based on euclidean distance and regularity values of the ROIs.To test the system's efficiency, we applied it to images from the Mammographic Image Analysis Society database.The accuracy of the system reaches 88.37% with 0.292 false positives per image.