核磁共振图像脑组织自动提取方法  

Automated MRI brain tissue extraction method

作　　者：张海燕[1] 李海云[1] 

机构地区：[1]首都医科大学生物医学工程学院,北京100069

出　　处：《计算机工程与应用》2014年第16期168-172,共5页Computer Engineering and Applications

基　　金：北京市自然科学基金(No.4122018)

摘　　要：核磁共振图像的脑组织提取是神经图像处理研究中的一个重要步骤。将传统的几何活动轮廓模型与二值水平集函数相结合,提出了一种新型的二值水平集活动轮廓模型,并基于该模型提出了一种能够自动、准确实现MRI脑组织提取的方法。该方法在脑组织内部自动设定最优初始轮廓曲线,将该演化曲线隐含地表示成一个高维函数的零水平集,零水平集在基于区域的图像力驱动下不断演化并达到待分割脑部图像的边缘。将基于该方法的脑组织提取结果与作为金标准的专家手动分割结果和其他流行算法相比较,结果表明提出的脑组织提取方法能够自动、准确和快速地提取MRI脑组织,是一种鲁棒性较好的MRI脑组织提取方法。The segmentation of brain region from non-brain region in magnetic resonance images, also referred to brain tissue extraction, is an important and difficult task due to the convoluted brain surface and weak boundaries between brain and non-brain tissue. This paper proposes an accurate and automatic method of brain tissue extraction based on a novel binary level set active contour model. An initial contour, which is automatically set inside the brain, is driven by a region-based image force until it well fit the brain and non-brain boundaries. Quantitative evaluation of the new method is performed by comparing the result of the new method to those obtained using manual segmentation in 10 sets normal adult MR brain images. Experimental results demonstrate that the proposed method can provide accurate and automatic brain tissue extraction results.

关 键 词：脑组织提取 灰度直方图 活动轮廓模型 水平集 质量评估 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]