الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
Breast cancer is a leading cause of death among all cancers in women, and its incidence is rising. Prevention has not been possible since the cause of this disease is not yet fully understood. Studies have shown that early detection of breast cancer by periodic mammographic screening of asymptomatic women can reduce breast cancer mortality [1]. In this thesis, we propose an adaptive region-based measure of image edge profile acutance [2] which characterizes the transition in density of a region of interest (ROI) along normals to the ROI at every boundary pixel, and thereby quantifies-i-the diffusion of the density of a mass into the surrounding tissue. We investigate the potential of acutance in characterizing the radiographic definition of tumor edges, and propose its application to discriminate between benign and malignant mammographic tumors. In addition, we use shape factors such as moments, Fourier descriptors, chord length statistics, and compactness based upon contours of masses, as well as moments based upon density distributions to aid in classification of circumscribed and spiculated masses. We further investigate four-group classification of tumors as circumscribed benign (CB), circumscribed malignant (CM), spiculated benign (SB), and spiculated malignant (SM). The BMDP (Biomedical Programs) package [3] was used for classification of the lesions using acutance and the shape factors as features in various combina¬tions. Thirty-nine images from the Mammographic Image Analysis Society (MIAS) database and an additional set of fifteen local cases were selected for this study. The cases included sixteen CB, seven CM, twelve SB, and ninteen SM lesions.