Meeting Banner
Abstract #3362

An Automatic Template-Based Breast Segmentation Method for Fat-Suppressed MRI

Xiaoyong Wang1, Muqing Lin1, Peter T. Fwu1, Ling-Chuan Chang2, Yi-Ting Wu2, Chin-Yu Chang2, Jeon-Hor Chen3, Min-Ying Su1

1Center for Functional Onco-Imaging, University of California, Irvine, CA, United States; 2Department of Radiology, China Medical University Hospital, Taichung, Taiwan; 3Center for Functional Onco-Imaging, University of California Irvine, Irvine, CA, United States

The purpose is to develop an automatic breast segmentation method for fat-suppressed MRI. The algorithm is based on template matching of chest region using nonrigid registration. The body landmarks defined on the template can be transformed to the subjects space to determine the posterior-lateral boundary of the breasts. The results show that this chest template-based method is robust for different body and breast shapes, with a mean error of 5.4% for breast, and 2.7% for fibroglandular tissue segmentation. This tool can help developing computer-aided-diagnosis method for breast cancer detection, as well as quantitative analysis of breast density for risk management.

Abstract PDF Presentation Video

Keywords

accuracy achieve anal another applied assigned automatic background bias bilateral body boundary bounding breast breasts cancer canny categorized central chest china choosing clustering clusters color commonly contour contribute correction covering cropped curve define deformed dense density dependent detect detection developing development diagnosis differentiating done easier easily edge ensure error evaluating evaluation examples exclude excluded excludes excluding exclusion extracted extreme fatty feasible field final fitting focus generation graphical green healthy highlighted identify impractical include included inclusion inconsistent inhomogeneity intensity kernel landmarks lateral location lung major make management manual manually margin match middle model models moderate muscle muscles near necessary needed neighboring operator original outer phys posterior procedures program quantitative registration reliable remove required requires rigid risk robust robustness segment segmentation segmented selected served several shape shapes skin slice slices smoothed source spine step studies subject superior suppressed suspicious systems table template templates thoracic tissue tissues variations volume volunteers wall wrong wrongly years