脑功能网络的fMRI特征提取及脑部疾病机器识别  

FMRI feature extraction and identification of brain diseases based on the brain functional network

作　　者：黄嘉爽 梅雪[1] 袁晓龙[1] 李振华[1] 

机构地区：[1]南京工业大学自动化与电气工程学院,江苏南京211816

出　　处：《智能系统学报》2015年第2期248-254,共7页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金资助项目(51205185);2012年度江苏省"青蓝工程"中青年学术带头人项目(KYLX_0754)

摘　　要：脑部疾病的机器识别是医学图像领域研究的热点。传统的功能磁共振图像研究方法大多只针对部分脑区。考虑到脑功能网络具有全局性的特征,利用静息态功能磁共振图像数据,在全脑范围内使用极大重叠离散小波变换,分别构建加权和无权脑功能网络,运用复杂网络理论对网络结构进行分析研究,提取网络聚集系数作为分类识别的特征分量。将该文方法用于对精神分裂症患者的识别,由识别率、灵敏度、特异度表明,该方法能够提高识别效果,且具有普遍适应性,能推广到其他脑部疾病的机器识别应用中。The machine recognition of brain diseases is a hotspot issue in the field of medical images. However,traditional fMRI image analysis only treats part of the brain region. Considering the overall characteristics of the brain network,the maximal overlap discrete wavelet transform is used to construct weighted and binary networks based on the rest-fMRI data. The complex networks theory is applied to the network structure analysis. Finally,the clustering coefficient of the network is extracted as the characteristic component of classification identification,which allowed the separation of schizophrenia patients from normal control subjects. This method is applied to the recognition of schizophrenia in this paper. The experimental results of recognition rate,sensitivity and specificity show that this method is able to improve the effect of recognition and has the universal adaptability,which can be extended to the recognition of other brain diseases.

关 键 词：功能磁共振图像 精神分裂症 复杂网络理论 特征提取 脑部疾病 机器识别 

分 类 号：R741[医药卫生—神经病学与精神病学] TP391.41[医药卫生—临床医学]