一种改进的交互式医学图像序列分割方法  

A Modified Interactive Segmentation of Medical Image Series

作　　者：罗希平[1] 田捷[1] 

机构地区：[1]中科院自动化研究所人工智能实验室,北京100080

出　　处：《电子学报》2003年第1期29-32,共4页Acta Electronica Sinica

基　　金：国家自然科学基金 (No .60 0 71 0 0 2 ;60 0 72 0 0 7;69931 0 1 0 ;60 1 72 0 57;79990 580 ;30 2 70 4 0 3) ;国家 863计划项目 (No 2 0 0 1AA1 4 4 1 2 0 ) ;国家杰出青年科学基金 (No 60 2 2 50 0 8) ;国家重大基础研究前期研究专项项目 (No 2 0 0 2CCA0

摘　　要：本文介绍了一种结合livewire算法和活动轮廓模型的医学图像序列的分割方法 .我们通过把livewire算法和图像分割中一般的区域增长方法结合来改进livewire算法 ,并用改进后的算法来对医学图像序列中的单张或多张切片进行交互式的准确分割 .然后计算机利用活动轮廓模型来自动分割相邻的未分割切片 .我们通过在活动轮廓模型的边缘点中引入记录已分割物体边缘附近局部区域特征的灰度模型来把已分割切片中的物体与背景的局部区域特征带入相邻的未分割切片中 ,并用由灰度模型定义的区域相似性代替活动轮廓模型中的外能来引导边缘轮廓收敛到物体的实际边缘 .本文还介绍了一种基于livewire算法思想的简单的分割结果交互式修补方法 .实验表明我们的算法仅需少量用户交互就能快速准确的从医学图像序列中分割出感兴趣的物体 .In this paper,we propose an algorithm for the semiautomatic segmentation of medical image series based on the combination of the live wire algorithm and the active contour model.We modify the traditional live wire algorithm by combining it with the region growing method and obtain accurate segmentation of one or more slices in a medical slice series by the live wire algorithm.Then the computer will segment the nearby slice using the active contour model.We introduced a gray scale model to the boundary points of the active contour model to record the local region characters of the desired object in the segmented slice and replace the external energy of the traditional active contour model with the energy decided by the gray scale model.A simple method based on the idea of the live wire algorithm is introduced for the reparation of the automatic segmentation result to guarantee the reliability of the result.Experiment shows that our algorithm can obtain the boundary of the desired object from a series of medical images quickly and reliably with only little user intervention.

关 键 词：医学图像处理 图像分割 活动轮廓模型 LIVE wire算法 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]