改进谱聚类算法在MCI患者检测中的应用研究  

Improved spectral clustering algorithm and its application in MCI detection

作　　者：相洁[1] 赵冬琴[2] 

机构地区：[1]太原理工大学计算机科学与技术学院,山西太原030024 [2]山西财经大学实验教学中心,山西太原030006

出　　处：《通信学报》2015年第4期27-34,共8页Journal on Communications

基　　金：国家自然科学基金资助项目(61170136;61373101;61472270;61402318);山西省科技攻关基金资助项目(20140321002-01)~~

摘　　要：为了利用功能核磁影像(f MRI,functional magnetic resonance imaging)数据进行轻度认知障碍(MCI,mild cognitive impairment)自动检测,对患者的f MRI数据进行聚类分析,得到患者大脑血氧依赖水平(BOLD,blood oxygen level dependence)的变化模式,并将异常模式用于疾病检测中。由于传统谱聚类算法需要计算相似矩阵所有的特征值和特征向量、时间与空间复杂度较高。提出一种改进的谱聚类方法,在相似矩阵的构造以及σ与k值的确定等方面进行了改进,将其用于MCI f MRI数据的聚类与诊断研究中。与传统谱聚类及Nystr?m算法进行的对比实验结果表明,改进的谱聚类方法可以更准确得到患者异常BOLD模式,分类正确率较高,且时间和空间复杂度均小于传统算法。In order to detect mild cognitive impairment （MCI） using functional magnetic resonance imaging （fMRI）, a method based on fMRI clustering was proposed fMRI data were clustered to obtain the blood oxygen level depend- ence（BOLD） change model of MCI patients, then abnormal patterns were used to detect disease. The traditional spectral clustering algorithm needs to calculate all of the eigenvalue and eigenvector, so time and space complexity is higher. An improved spectral clustering method was proposed which modified the similar matrix construction method and the setting method of a and k, and then this method was applied to clustering and detection of MCI patients. To verify the perform- ance of the proposed method, the comparison of the clustering result, classification accuracy using traditional algorithm and Nystr0m is also done. The comparative experimental results show that the proposed method can get BOLD pattern more accurately, the accuracy of MCI detection is higher than the other two algorithms, and the time and space complex- ity are less than the traditional algorithm.

关 键 词：谱聚类 Nystr6m fMRI-BOLD 轻度认知障碍 MCI诊断 

分 类 号：TP181.09[自动化与计算机技术—控制理论与控制工程]