在功能磁共振成像中，独立成分分析是一种常用的数据驱动的分析方法。独立成分分析所具有的成分次序不确定性和不同个体间复杂的变异性使得多被试独立成分分析变得困难。 目前存在的多种算法对个体间的相互依赖关系作了不同的假设，要么没有充分保留个体变异，要么没有解决次序不确定性。针对这些问题，本文ᨀ出一种基于脑活动一致性的多被试独立成分分析算法。该方法仅仅对个体间的相互依赖关系作了很少和一般性的假设， 通过选取一致性较高的成分来构造模板，使用模板匹配的策略更好地解决了次序不确定性，同时充分保留了个体变异性。在模拟数据测试中，相比于现有其他方法，新的方法在解决成分次序不确定性和保留个体变异这两方面具有同等或更高的准确性。在真实fMRI数据测试中，比较了阿尔茨海默症病人与健康老年人在默认网络和顶叶记忆网络上功能连接的差异，新方法重复了以往研究的结果，并且与现有的方法得到了较一致的结果。不同于现有方法，新方法做了更少的假设，最大程度保留了独立成分分析数据驱动性这一特点，特别适合于探索性分析。

Independent Component Analysis (ICA) is a common data-driven analytic method  for functional Magnetic Resonance Imaging (fMRI)  data. The  permutation ambiguity of  independent components  in ICA and the complex individual variability among multiple subjects lead to difficulties in multi-subject data analysis. The current algorithms suffer from problems such as ignoring individual variability or mismatching components across  subjects. To address these problems, we propose a multi-subject ICA algorithm based on  spatial consistency of individual components. This algorithm  only makes  very  limited  and general assumptions  on inter-subject interdependence.  Using spatially consistent individual components to  generate group-level templates and template matching strategy to align individual components, this algorithm deals with the permutation ambiguity better, while preserves individual variability as much as possible. Comparisons between the new algorithm and existing methods using simulated data showed that the new algorithm achieves equivalent or higher accuracy  in  resolving permutation ambiguity and preserving individual variability.  In real fMRI data, we examined the functional connectivity difference between healthy elders and patients  with Alzheimer’s disease  on default mode network and parietal memory network. The new algorithm  replicated the results of previous studies,  and  obtained similar results in comparison with other methods. Distinct from the existing methods, the new algorithm makes lesser assumptions and thus extends the data-driven property of ICA to the multi-subject scenario, which is very suitable for exploratory analysis.