Comparing fMRI data processing methods with respect to possible inaccuracy

We present the comparison of several methods to analyze functional magnetic resonance imaging data with respect to some inaccuracy in subject's response (e.g. missing response to stimuli). Approaches based on general linear model, independent component analysis, and combined methods were used to analyze both simulated and real data with unknown (real data) and known (simulated data) levels of inaccuracy. Our results suggest combined approaches as the best choice for single subject analyses. No significant differences between the methods were observed for group results.