混频对多尺度特征提取方法分析fMRI数据的影响

摘要

Objective To discuss the impact of aliasing frequency on the performance of multiscale feature extraction (MFE) for fMRI data analysis. Methods Under the conditions of removing and not removing aliasing frequencies. MFE was employed to analyze the simulated and the auditory fMRI data. In addition, the results revealed by MFE were compared with those of the general linear model (GLM) implemented with SPM8 software. Results Whether removing aliasing frequencies or not, MFE showed the same specificity as that of GLM. However, in terms of the sensitivity, the performance of MFE without removing aliasing frequencies was better than that of MFE when removing aliasing frequencies, and the latter was better than that of GLM. Conclusion In case of correlation analysis employed, aliasing frequencies do not influence the specificity of MFE, while removing these frequencies will decrease its sensitivity.%目的 探讨混频对多尺度特征提取(MFE)方法分析fMRI数据的影响.方法 分别在去除和不去除混频条件下用MFE分析模拟数据及听觉fMRI试验数据,并与由SPM8软件运行的广义线性模型(GLM)方法的结果进行比较.结果 MFE在去除和不去除混频两种条件下的特异度均与GLM相同,但MFE不去除混频时的灵敏度优于去除混频时的灵敏度,后者又优于GLM.结论 在使用相关分析检测激活的条件下,混频不影响MFE的特异度,但去除混频降低其灵敏度.