目的 采用空间独立成分分析法( siCA)和时间独立成分分析法(tICA)对功能磁共振成像(fMRI)信号进行分离,比较信号间的时空特性对2种独立成分分析方法性能的影响.方法 模拟fMRI数据,并将2组独立的信号以及它们的线性混合信号叠加到空间独立的区域,分别利用Infomax、Combi、FBSS和ICA-EMB 4种算法实现sICA和tICA,并对模拟数据中的3组信号进行提取和分离.结果 sICA只能分离空间独立且时间高度独立的信号,无法分离空间相关、时间独立的信号;tICA不仅能够准确分离空间和时间高度独立信号,而且能够准确分离空间高度相关、时间独立的信号,并将时间相关信号整体提取;FBSS和ICA-EMB 2种算法较Infomax和Combi性能稳定.结论 空间或时间独立性假设违背到一定程度时,slCA和tICA对信号分离的结果存在差异.应根据需要选择适合的sICA或者tICA方法对fMRI数据进行处理.%Objective Separate the fMRI signals by using spatial independent component analysis (sICA) and temporal independent component analysis (tICA). Compare the spatiotemporal characteristics among signals which has effect to performance analysis of two kinds of independt signals. Method Simulate fMRI data. Two sets of independent signals and their linear mixture were added in spatially independent regions. sICA and tICA were achieved by using respectively Infomax, Combi, FBSS and ICA-LMB. Then three sets of signals were applied to separate and extract from simulated fMRI. Results sJCA can only separate signals spatially independent and highly temporally independent, can not separate spatially correlated and temporally independent signals. tlCAis not only able to separate highly spatially and temporally independent signals, but also able to successfully separate highly spatially correlated and temporally independent signals, moreover, it can integrally extract temporally correlated signals. FBSS and ICA-EMB algorithms are more stable than Infomax and Combi algorithms. Conclusion The assumption of spatial or temporal independence is violated to a certain degree. There are certain difference of performance between sICA and tICA. It should be based on the characteristics of signals and requirement to select adaptive sICA or tICA method to fulfill blind signal separation in fMRI data.
展开▼