Digital Mammography: Advanced Computer-Aided Breast Cancer Diagnosis

Abstract

Abstract : The goal of (he project is to develop a computer aided diagnosis (CAD) system for full field digital mammography (FFDM) using advanced computer vision techniques and image information fusion from multiple views and bilateral 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: (I) Collected a data set of FFDM cases that contain mammographic lesions. (2) Developed a database management program to store the coded case information to facilitate archiving and retrieval of the cases. (3) Developing a pyramid image enhancement technique for preprocessing the raw image from the FFDM system, which is then as the input to our CAD system. (4) Compared me mass detection accuracy of our CAD algorithm when applied to our processed images and the GE's processed images and found that the detection Sensitivities on the two sets of images were within a few percent over the entire FP range of interest. (5) Compared the performance of the mass detection system on digitized film mammograms and DMs, and found that, with adjustment of the processing parameters in the algorithm, the detection accuracies of the algorithms on both sets of images are comparable. (6) Investigated the effects of the image enhancement filter on the accuracy of mass detection and found that, by replacing the DWCE filter with an adaptive ring filter, the sensitivity of mass detection can be improved up to 15%. (7) Compared the mammographic density segmented from digitized film mammograms and DMs from the same patients, and found that the correlation of the segmented breast density between the two types of images is very high but the estimated percent dense area on DMs it, on average, about 5% lower than that estimated from digitized film mammograms.