多相变分水平集的脑部磁共振图像分割与偏移场矫正的耦合模型  

A Coupling Model of Multi-phase Variational Level Set for Brain Magnetic Resonance Image Segmentation and Bias Field Correction

作　　者：王昌[1] 刘艳[1] 秦鑫[1] 张文超[1] 于毅[1] 

机构地区：[1]新乡医学院生物医学工程学院,河南新乡453003

出　　处：《航天医学与医学工程》2015年第6期450-455,共6页Space Medicine & Medical Engineering

基　　金：国家自然科学基金(61305147);新乡医学院科研培育基金(2014QN142)

摘　　要：目的为分割有偏移场的脑部磁共振图像,建立了一种多相变分水平集的脑部磁共振图像分割与偏移场矫正的耦合模型。方法依据偏移场特性,定义了基于局部灰度信息的K均值聚类准则,并将聚类准则项整合为变分水平集函数的基本能量项,构造变分水平集的能量函数,求解能量函数的欧拉-拉格朗日方程。将提出的变分水平集算法扩展为多相,最终实现脑组织分割和偏移场的矫正。结果以医生手动分割的脑组织为标准,将本文模型与多相LBF模型的分割结果进行定量分析,结果表明,偏移场强度为20%时,本文模型的分割精度比多相LBF模型提高了10%,且随着偏移场强度进一步增强,本模型的优势更明显。结论本模型抗噪能力强且分割结果受参数影响小,可应用于脑组织的分割和偏移场的矫正。Objective A coupling model of multi-phase variational level set for brain MR image segmentation and bias field correction was established to segment brain Magnetic Resonance image with bias filed. Methods The local K-means clustering criterion was defined based on the features of bias field,and clustering criterion function was integrated into the basic energy term of variational level set function. The energy function of variational level set function was constructed,and Euler-Lagrange equation of the energy function was solved. Extending this variational level set algorithm to multi-phase will eventually realize the brain tissue segmentation and bias field correction. Results Taking the doctor ＇s manual brain tissue segmentation as the standard,quantitative analysis based on the results of segmentation by this model and multiphase local binary fitting model were made. The results were as follows： when the bias field intensity was 20%,the segmentation accuracy using this model was improved by 10% as compared with the multi-phase local binary fitting model; with the increase of the bias field intensity,the advantages of this model became more obvious. Conclusion This model has the advantage of strong anti noise capability and little dependence on parameters,and can be applied to the brain tissue segmentation and bias field correction.

关 键 词：多相 变分水平集算法 偏移场 K均值聚类 分割 脑部磁共振图像 

分 类 号：R319[医药卫生—基础医学] R857