摘要
目的提出一种新颖的基于自组织特征映射和遗传算法的无监督MR脑部图像分割算法。方法本研究算法分为5步:图像预处理去除背景噪声和颅骨部分、提取图像中两类统计特征和几何不变矩、遗传算法降低特征空间维度、训练自组织特征映射完成向量分类和使用梯度熵聚类算法得到分割图像。结果选用国际MR脑图像库和临床实例MR图像进行仿真实验。定性分析表明基于本文算法的分割图像中白质、灰质和脑脊液边界完整清晰;定量评估结果显示本文提出的遗传特征优化算法优于常用的主分量分析法,梯度熵算法所得分割图像优于K-means聚类算法,且本文提出的算法在白质和脑脊液分割方面优于现存最佳的CGMM算法。结论本文提出的分割流程没有涉及任何关于体素分类的先验知识,是一种完全无监督的MR脑部组织自动分割方法,具有很强的稳定性、优越性,且获得高精确性的分割图像。
Objective This study aimed to present a novel unsupervised method for MR brain image segmentation based on selforganizingmaps(SOMs)and genetic algorithms(GAs).Methods In particular,the proposed method was based on five stagesconsisting of image preprocess,extracting first and second order statistical features,feature selection using evolutionary computation,voxel classification using SOM,and entropy-gradient(EG)clustering.Results Both simulated and clinical datasets were evaluatedby different methods.Qualitative analysis showed that the components of the white matter,gray matter and cerebrospinal fluidwere well preserved,and globally all the regions were correctly classified.Quantitative evaluation results showed that the geneticalgorithms can achieve optimized feature set than principal component analysis(PCA).EG had a statistical significance differencewith K-means(P<0.01).Our algorithm outperformed the CGMM method in WM and CSF delineation,and was found to be the mosteffective among other techniques.Conclusion The complete procedure does not use any a priori knowledge regarding voxel classassignment,but reveals a fully unsupervised,automated method for MRI segmentation to directly identify different tissue classes,which can provide better robustness,superiority,and pervasiveness in clinical applications.
作者
丁力
周啸虎
陈宇辰
高伟
DING Li;ZHOU Xiaohu;CHEN Yuchen;GAO Wei(Department of Radiology, Nanjing First Hospital, Nanjing Medical University, Nanjing Jiangsu 210006, China)
出处
《中国医疗设备》
2017年第10期21-26,共6页
China Medical Devices
基金
国家自然科学青年基金(81601477)
关键词
脑疾病
MR脑部图像
图像分割
自组织特征映射
遗传算法
梯度熵聚类
cerebral disease
MR brain image
image segmentation
self-organizing maps
genetic algorithms
entropy-gradient clustering
