基于非刚性配准的交互式医学图像序列分割  

Interactive medical image sequences segmentation based on non-rigid registration

作　　者：胡永祥[1] 蒋鸿[2] 陈洪波[1] 

机构地区：[1]中南大学信息物理工程学院生物医学工程研究所,长沙410083 [2]湖南工业大学计算机科学与技术系,湖南株洲412008

出　　处：《计算机工程与应用》2007年第19期231-233,236,共4页Computer Engineering and Applications

摘　　要：提出了一种新的交互式医学图像序列分割算法,该算法将非刚性配准技术和解剖先验知识相结合把图像分割问题转化为图像配准问题。首先采用Demons算法进行图像配准,用光流法计算瞬时位移,设计了一个新的停止准则使其能自适应地确定迭代次数,并将它在金字塔型的多尺度框架下实现。然后用配准得到的形变域对已精确分割的图像进行形变就能自动地获得未分割目标图像的分割结果。扩展上述过程就可实现整个图像序列分割。试验结果表明该算法用户干预少、分割速度快、分割结果准确。This paper presents a new algorithm for medical image sequences segmentation,which converts image segmentation into image registration by combining non-rigid registration technology with prior knowledge about anatomy.Firstly,we select Demons algorithm to register image,adopt optical flow method to compute instantaneous displacement,design a new stop criterion for Demons algorithm and implement it in a pyramid multi-scale framework.Then,the deformation fields obtained from image registration are used to wrap the prior image that has been segmented.So the segmented result is obtained.Repeating this process,the segmented results of the whole image sequence can be achieved.Experimental results demonstrate the new algorithm can rapidly get accurate results with few interventions.

关 键 词：非刚性配准 医学图像序列 交互式分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]