摘要
背景:由于脑部MR图像中信息对比度不高,各种脑部组织的形状复杂等特点,分割方法的选择比较困难,单一的算法很难获得满意的分割结果。目的:针对脑部MRI的特点综合利用现有的算法开发和定制有效的分割应用算法。方法:根据邻域连接和Canny水平集分割算法的优缺点,结合图像特征,用邻域连接方法的分割结果作为Canny水平集分割算法的先验分割模型,借以确定出Canny算法的下限阈值,从而完成两种算法的混合分割。结果与结论:采用实验所用混合方法得到的白质和灰质的分割结果,经与专家手工分割结果对比,证明该方法取得了较好的分割效果,从而证明综合利用现有的算法,不仅避免了重复劳动,还能开发和定制出更加有效的分割应用算法,具备很好的应用潜力。
BACKGROUND: In the brain MRI images, for the low information contrast and the complex brain tissue shape, the choice of the segmentation method is more difficult, the single algorithm is difficult to obtain satisfactory segmentation results. OBJECTIVE: To develop and customize effective segmentation application algorithm according to the characteristics of brain MRI by using the existing algorithm synthetically. METHODS: According to the advantages and disadvantages of neighborhood connected and Canny level set algorithms used in the experiment, and combining the image characteristics, we took the results of the neighborhood connected methods as the prior segmentation model of Canny level set algorithm in order to determine the lower threshold of Canny level set algorithm and thus finished hybrid segmentation method. RESULTS AND CONCLUSION: Compared the white matter and gray matter segmentation result of the hybrid segmentation method with the result of the experts’ manual segmentation, it proved that this method had achieved good segmentation effect, and thus proved that making use of the existing algorithm synthetically could not only avoid the repeated labor, but also could develop and customize more effective segmentation application algorithm and had good potential application.
出处
《中国组织工程研究》
CAS
CSCD
2012年第39期7302-7306,共5页
Chinese Journal of Tissue Engineering Research
基金
内蒙古自治区自然科学基金项目(2010Zd26)
内蒙古科技大学校内创新基金项目(2010NC037)(2010NC030)~~