一种基于复合稀疏表示的阿尔茨海默病的诊断方法  

A Diagnosis Method for Alzheimer's Disease based on Hybrid Sparse Representation

作　　者：滕升华[1] 商胜楠 王芳[1] 赵增顺[1,2] 

机构地区：[1]山东科技大学电子通信与物理学院,青岛266590 [2]山东大学控制科学与工程学院,济南250061

出　　处：《生物医学工程研究》2016年第1期7-11,共5页Journal Of Biomedical Engineering Research

基　　金：国家自然科学基金资助项目(61174190;61471225);山东省自然科学基金资助项目(ZR2014FM002);中国博士后科学基金特别资助项目(2015T80717)

摘　　要：利用磁共振影像数据实现对阿尔茨海默病的准确诊断。将常规稀疏表示中的单层字典分解为两层,分别使用各类别的典型样本和类内差异作为两层字典的元素;设计一种两层字典协调工作的复合稀疏表示形式,以期利用训练样本更为精确地表示待识别样本,并构建分类器用于阿尔茨海默病的分类识别。在ADNI数据库的对比实验表明,该方法的识别性能优于支持向量机和同类的稀疏表示分类器。Given magnetic resonance imaging data,the accurate diagnosis method for Alzheimer＇s disease was studied in this paper. By expanding the conventionally single-layer dictionary in sparse representation into two layers,the exemplars in all classes and the intra-class variations were used to construct a two-layer dictionary. Then a hybrid sparse representation was designed collaboratively incorporating the two layers of the dictionary,where the testing samples were expected to be represented more accurately with the training data. Finally a classifier based on this hybrid sparse representation was developed to diagnose Alzheimer＇ s disease. Experiments on Alzheimer＇s Disease Neuroimaging Initiative（ ADNI） dataset show promising results compared with support vector machine and traditional sparse representation based classifier.

关 键 词：阿尔茨海默病 灰质密度图 复合稀疏表示 近邻传播聚类 增广拉格朗日乘子法 

分 类 号：R318[医药卫生—生物医学工程]