摘要
为提高医学图像在组织边界不清晰以及灰度不均匀下的分割性能,提出一种基于多类样本间模糊距离的隶属度函数分割方法。通过磁共振序列测量确定反映磁共振图像脑部组织特性的映射图,经预处理后得到样本模糊标签;设计基于多样本类间模糊距离的隶属度函数确定各样本的隶属度,该隶属度的确定综合考虑了同类样本与不同类样本之间的空间距离,降低了同类样本之间的隶属度依赖;训练模糊支持向量机对三种主要脑组织进行分割。在脑图像公开数据集上的分割实验表明,改进算法可有效提高分割精度。
In order to improve the segmentation performance of medical images under unclear tissue boundaries and non-uniformity intensity, a membership function segmentation method based on fuzzy distance between multiple types of samples is proposed. Maps reflecting the characteristics of brain tissues in magnetic resonance images were determined by magnetic resonance sequence measurement, and the sample fuzzy labels were obtained after preprocessing. A membership function based on the fuzzy distance between multiple samples was designed to determine the membership of each sample. The determination of the degree comprehensively considered the spatial distance between the same sample and different samples, and reduced the degree of membership dependence between the same samples. The fuzzy support vector machine was trained to segment the three main brain tissues. Segmentation experiments on public image datasets of brain images show that the improved algorithm can effectively improve the segmentation accuracy.
作者
杨兵
刘晓芳
Yang Bing;Liu Xiaofang(Institute of Computer Application and Technology,China Jiliang University,Hangzhou 310018,Zhejiang,China;Key Laboratory of Electromagnetic Wave Information Technology and Metrology of Zhejiang Province,College of Information Engineering,China Jiliang University,Hangzhou 310018,Zhejiang,China)
出处
《计算机应用与软件》
北大核心
2022年第4期254-262,共9页
Computer Applications and Software
基金
国家自然科学基金项目(61672476)
浙江省大学生科研创新活动计划项目(2019R409055)。
关键词
模糊支持向量机
类间距离
隶属度函数
图像分割
模糊标签
Fuzzy support vector machine
Inter-class distance
Membership function
Image segmentation
Fuzzy labels
