KMID : 1147120110170010038
|
|
Journal of the Korean Society of Imaging Informatics in Medicine 2011 Volume.17 No. 1 p.38 ~ p.45
|
|
Demon based Nonrigid Registration in Dynamic Contrast-Enhanced MR Breast Images for Tumor Perfusion Analysis
|
|
Jo Hyun-Hee
Hong Helen Heo Chang-Yong Kim Jong-Hyo
|
|
Abstract
|
|
|
In this paper, we propose a demon based nonrigid registration method in dynamic contrast-enhanced MR breast images for the perfusion analysis of tumor. Our method consists of following three steps. First, breast region is segmented by eliminating breast skin and pectoral muscle, and tissues in the breast region are classified into two classes (fat, and glandular tissue and tumor) using k-means clustering. Second, breast tumor is automatically localized and segmented by using intensity and morphology based method. Third, breasts in pre and post contrast-enhanced MR images are aligned through normalized mutual information based rigid registration, and density difference between corresponding tissues in both MR images is corrected. Nonrigid motion of breast in pre and post contrastenhanced MR images is refined through demon based nonrigid registration with tumor rigidity constraint. To evaluate the performance of our method, we performed the visual inspection and the accuracy evaluation. The experimental results show that the proposed demon based nonrigid registration can accurately deform the breast while maintaining rigid structure of the tumor.
|
|
KEYWORD
|
|
DCE-MRI, Breast Segmentation, Tissue Classification, Tumor Segmentation, Nonrigid Registration
|
|
FullTexts / Linksout information
|
|
|
|
Listed journal information
|
|
|