基于静息态fMRI信号复杂度的MCI识别研究  

MCI recognition based on the complexity of resting-state fMRI signal

作　　者：董建鑫[1] 王川[1] DONG Jianxin;WANG Chuan(Yanjing Medical College,Capital Medical University,Beijing101300)

机构地区：[1]首都医科大学燕京医学院,北京101300

出　　处：《北京生物医学工程》2022年第6期564-568,582,共6页Beijing Biomedical Engineering

基　　金：首都医科大学燕京医学院科研基金(18qdky09)资助。

摘　　要：目的基于静息态功能磁共振图像,提取默认网络特征脑区的信号复杂度参数建立轻度认知障碍(mild cognitive impairment,MCI)的分类模型。方法研究数据来源于阿尔茨海默病神经成像数据库,包含48名健康人和53例MCI患者的数据。首先进行独立成份分析,针对分离出的独立成份分别计算对应时间序列的Hurst指数。然后在体素水平上采用双样本t检验选择左侧眶部额下回、左侧额上回和左侧额中回作为特征脑区,计算其Hurst指数作为分类特征。最后用支持向量机对MCI患者进行识别,并评价模型的准确率、灵敏度、特异度以及接收操作特征(receiver operating characteristics,ROC)曲线下面积。结果基于MCI和正常对照两组构建的分类模型,获得了最高88.71%的分类准确率、90.91%的灵敏度和86.21%的特异度,此外,ROC曲线的最大线下面积为0.96。结论Hurst指数可以反映MCI患者异常脑功能活动,基于独立成份分析和支持向量机的方法能有效地识别MCI患者,具有一定的临床辅助诊断意义。Objective Based on resting-state functional magnetic resonance images,signal complexity parameters of the default network characteristic brain regions were extracted to establish the classification model of mild cognitive impairment(MCI).Methods The data were from the Alzheimer’s Disease Neuroimaging Initiative(ADNI)database included 48 normal controls and 53 MCI patients.Firstly,the independent component analysis was carried out,and the Hurst exponents of corresponding time series were calculated for the separated independent components.Then,at the voxel level,two-sample t test was used to select the left orbital part of inferior frontal gyrus,left superior frontal gyrus and left middle frontal gyrus as characteristic brain regions,and their Hurst exponents were calculated as classification features.Finally,support vector machine was used to identify MCI patients,and the accuracy,sensitivity,specificity and area under receiver operating characteristics(ROC)curve of the model were evaluated.Results The classification model based on MCI and normal control group obtained the highest classification accuracy of 88.71%,sensitivity of 90.91%and specificity of 86.21%.In addition,the maximum area under ROC curve was 0.96.Conclusions Hurst exponent can reflect abnormal brain functional activities of MCI patients.Methods based on independent component analysis and support vector machine can effectively identify MCI patients,which has certain clinical diagnostic significance.

关 键 词：轻度认知障碍 静息态功能磁共振成像 HURST指数 独立成份分析 支持向量机 

分 类 号：R318.04[医药卫生—生物医学工程]