Empirical comparison of sources of variation for FMRI connectivity analysis

In neuroimaging, connectivity refers to the correlations between signals in different brain regions. Although fMRI measures of connectivity have been widely explored, the methods used have varied. This complicates the interpretation of existing literature in cases when different techniques have been used with fMRI data to measure the single concept of "connectivity." Additionally the optimum choice of method for future analyses is often unclear.

In this study, measures of functional and effective connectivity in the motor system were calculated based on three sources of variation: inter-subject variation in task activation level; within-subject variation in task-related responses; and within-subject residual variation after removal of task effects. Two task conditions were compared. The methods yielded different inter-regional correlation coefficients. However, all three approaches produced similar results, qualitatively and sometimes quantitatively, for condition differences in connectivity.

While these results are specific to the motor regions studied, they do suggest that within-subject and across-subject results may be usefully compared. Also, the presence of task-specific correlations in residual time series supports arguments that residuals may not substitute for resting-state data, but rather may reflect the same underlying variations present during steady-state performance.