Biomedical Engineering Reference
In-Depth Information
11
SEGMENTATION OF BRAIN MR IMAGES
USING J-DIVERGENCE-BASED
ACTIVE CONTOUR MODELS
Wanlin Zhu, Tianzi Jiang, and Xiaobo Li
National Laboratory of Pattern Recognition
Institute of Automation, Beijing, China
In this chapter we propose a novel variational formulation for brain MRI segmentation. The
originality of our approach is on the use of J-divergence (symmetrized Kullback-Leibler
divergence) to measure the dissimilarity between local and global regions. In addition, a
three-phase model is proposed to perform the segmentation task. The voxel intensity value
of all regions is assumed to follow Gaussian distribution. It is introduced to ensure the
robustness of the algorithm when an image is corrupted by noise. J-divergence is then used
to measure the “distance” between the local and global region probability density functions.
The proposed method yields promising results on synthetic and real brain MR images.
1.
INTRODUCTION
Image segmentation plays a major role in medical applications since seg-
mentation results often influence further image analysis, diagnoses, and therapy.
Manual delineation of brain tissues performed by an expert is time consuming and
does not allow reproducibility of segmentation results. Thus, interactive semi-
automatic or automated and reliable methods are desirable. However, medical
images acquired from different imaging systems or machines often suffer from
different artifacts, such as noise and the partial volume effect (due to imaging
Search WWH ::




Custom Search