中山大学学报自然科学版 ›› 2017, Vol. 56 ›› Issue (2): 40-47.

• 研究论文 • 上一篇    下一篇

基于张量法的阿尔兹海默症脑图像分类

杨宁1,徐盼盼1,刘佩嘉2,李淑龙1   

  1. 1. 南方医科大学生物医学工程学院//广东省医学图像处理重点实验室,广东 广州 510515;
    2. 华南理工大学数学学院,广东 广州 510641
  • 收稿日期:2016-08-05 出版日期:2017-03-25 发布日期:2017-03-25
  • 通讯作者: 李淑龙(1981年生), 女;研究方向:影像组学,医学图像分析;E-mail:shulong@smu.edu.cn

Prognostic classification of Alzheimers disease brain image-based on tensor method

YANG Ning1, XU Panpan1, LIU Peijia2, LI Shulong1   

  1. 1. School of Biomedical Engineering, Southern Medical University//Guangdong Provincial Key Laboratory of Medical Image Processing,Guangzhou 510515, China;
    2. School of Mathematics, South China University of Technology, Guangzhou 510641, China
  • Received:2016-08-05 Online:2017-03-25 Published:2017-03-25

摘要:

为了识别阿尔兹海默症(Alzheimers Disease,AD)与轻度认知障碍(Mild Cognitive Impairment,MCI)患者,提出了一种基于三阶张量方法的以MRI图像脑灰质灰度为特征的分类方法。采集了70例AD患者,112例MCI患者(包含在随访中转化为AD的,MCI-C:MCI Converters与未转化为AD的,MCINC:MCI Nonconverters各56例),以及70例正常人(NC)的MRI脑图像,提取脑灰质各体素的灰度,获得三阶灰度张量。采用基于张量的独立成分分析,以取得三阶灰度张量的独立成分;为了降低特征维数,利用支持张量机,将张量特征转化为向量特征,再利用递归特征消除法获取有效的主要特征。最后,对四组人群进行分类:AD-NC, MCI-NC, AD-MCI, MCI-C--MCI-NC,此分类模型采用7折交叉验证的方法进行训练测试。此外,还结合样本的基本信息与认知分数进行分类,证明了基本信息、认知分数和脑灰质灰度提供了互补的信息,有助于提升分类效果。结果表明,该方法拥有优良的分类性能,有助于对AD与MCI的诊断治疗。

关键词: 阿尔兹海默症, 轻度认知障碍, 张量, 认知分数, 独立成分分析, 支持张量机, 递归特征消除

Abstract:

A classification method based on the third-order tensors of brain structural magnetic resonance images is proposed to automatically identify Alzheimers disease and mild cognitive impairment. Brain structural magnetic resonance images from 70 AD patients, 112 MCI patients (included patients were converted to AD during follow-up, MCI-C: MCI Converters and patients were not converted to AD during follow-up, MCI-NC: MCI Non-converters) and 70 NCs (normal controls) are collected. The third-order tensors are obtained by extracting image intensity of each voxel of gray matter. In order to obtain the independent components of the third-order tensors, independent component analysis (ICA) is applied. Then, support tensor machine (STM) and recursive feature elimination (RFE) are used to reduce features dimensions and determine dominate features for classification. Finally, the classification of four groups, such as AD-NC, MCI-NC, AD-MCI, MCI-C--MCI-NC, is implemented by using 7-fold cross-validation method. In addition, basic information and cognitive scores are combined with the thirdorder tensor for classification. It is proved that basic information, cognitive scores and image intensity of brain gray matter provide complementary information, which is helpful to improve the classification effect. The experiment results show that this method can achieve excellent classification effect, which contributes to the diagnosis and treatment of Alzheimers disease and mild cognitive impairment.

Key words: AD, MCI, tensor, cognitive scores, independent component analysis, support tensor machine, recursive feature elimination

中图分类号: