摘要
针对脑补核磁共振图像分割问题,提出一种基于高斯混合模型融合半监督聚类方法。首先,从原始数据集随机选取MR图像切片;然后,利用k-均值算法聚类;最后,利用统计学和所得聚类的类标签信息完成分类。实验结果表明,该方法加快了收敛速度。相比两种有监督高斯混合模型,该方法取得了更好的分割结果,且无需标记训练数据集。
For the magnetic resonance brain image segmentation problem, we proposed a method which is based on fusing Gaussian mixed model and semi-supervised clustering (SSC). First, we randomly selected MR image slices from original dataset. Then, we used k-mean algorithm for clustering. Finally, we employed statistics and the class label information of the derived clusters to complete the classification. Experimental results showed that the proposed method speeded up the convergence. Compared with two supervised Gaussian mixed models, the proposed method achieved better segmentation results and did not need labelled training dataset.
出处
《计算机应用与软件》
CSCD
2015年第7期70-73,78,共5页
Computer Applications and Software
基金
四川省教育厅科研项目(13ZAO125)
四川省高校重点实验室开放基金项目(2014WZY05)
企业信息化与物联网测控技术四川省高校重点实验室资助项目(2014WZY03)
软件工程专业综合改革项目(B12201002)
关键词
半监督学习
K-均值聚类
高斯混合模型
磁共振脑部图像
统计学参数
Semi-supervised learning k-mean clustering Gaussian mixed model Magnetic resonance brain image Statistics parame-ters