基于知识的三维核医学图像左心室心肌区的提取  

Knowledge-based Automatic Extraction of Left-Ventricular Myocardial Region from 3D Nuclear Medicine Image

作　　者：王成[1] 张剑戈[1] 江旭峰[2] 章鲁[1] 

机构地区：[1]上海交通大学医学院生物医学工程系,上海200025 [2]上海交通大学医学院附属瑞金医院核医学科,上海200025

出　　处：《中国生物医学工程学报》2007年第1期19-23,34,共6页Chinese Journal of Biomedical Engineering

基　　金：上海市教委项目(05BZ15)

摘　　要：从心脏PET或SPECT图像中提取完整的心肌区域是定量分析心功能的前提。心脏的PET和SPECT图像边界模糊,在病理状态下可能有局部显像缺失,致使图像分割困难。本研究提出一种基于医学知识的快速推进法,利用拟合的椭球模型将边界演化推进到局部低显像区,从而分割出一个完整的左心室心肌区域。实验图像测试和实际图像分割表明这种算法对于有显像缺失的三维核医学心脏图像的分割是有效的。Correctly extracted myocardial region from nuclear medicine images is a key step for analyzing heart functions quantitatively. However, it is often difficult to segment myocardial region from the PET or SPECT image due to blurred border and probable non-uniform attenuation of radioisotope in pathological status. This paper presented a modified fast-marching algorithm, in which a fitted ellipsoid model was employed to march the evolution surface into the diminished region to insure the extracting of the left-ventricular wall. The experiment based on the test-images and the real nuclear medicine images verifies that the proposed algorithm is very effective to the 3D myocardial image with local display diminished.

关 键 词：图像分割 椭球拟合 快速推进法 曲面演化 心肌 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]