Digital Mammography: Development of an Advanced Computer-Aided System for Breast Cancer Detection

Abstract

Abstract : The goal of the project is to develop computer-aided diagnosis (CAD) methods and systems for mammography using advanced computer vision techniques and image information fusion from multiple mammograms to improve lesion detection and characterization. When fully developed, the CAD system can assist radiologists in mammographic interpretation. During this project year, we have performed the following tasks: (1) collected databases of digital mammograms (DMs) and digitized film mammograms (DFMs) for development of the CAD systems, (2) conducted a study to compare the percent dense area manually segmented by experienced radiologists on DMs and DFMs, (3) developed new image enhancement techniques and new false-positive reduction methods for mass detection, and conducted studies to compare the accuracy of mass detection by the CAD systems for DMs and DFMs using FROC analysis, (4) developed automated method for nipple detection on mammograms as a basis of multiple image fusion analysis for CAD systems, and (5) compared the accuracy for classification of malignant and benign breast masses using single-view and fused two-view information on mammograms by computer, and evaluated the effects of CAD on experienced radiologists' characterization of malignant and benign breast masses in two-view temporal pairs of mammograms. In summary, we have investigated a number of areas in CAD of mammographic lesions and evaluated the new techniques for both DMs and DFMs. We have made progress in the six tasks proposed in the project. We have found that our new computer-vision techniques and two-view information fusion approach can improve the performance of the CAD systems. We will continue the development of the CAD systems for%Ms and DFMs in the coming years.