Plurality and resemblance in fMRI data analysis

We apply nine analytic methods employed currently in imaging neuroscience t o simulated and actual BOLD fMRI signals and compare their performances und er each signal type. Starting with baseline time series generated by a rest ing subject during a null hypothesis study, we compare method performance w ith embedded focal activity in these series of three different types whose magnitudes and time courses are simple, convolved with spatially varying he modynamic responses, and highly spatially interactive. We then apply these same nine methods to BOLD fMRI time series from contralateral primary motor cortex and ipsilateral cerebellum collected during a sequential finger opp osition study. Paired comparisons of results across methods include a voxel -specific concordance correlation coefficient for reproducibility and a res emblance measure that accommodates spatial autocorrelation of differences i n activity surfaces. Receiver-operating characteristic curves show consider able model differences in ranges less than 10% significance level (false po sitives) and greater than 80% power (true positives). Concordance and resem blance measures reveal significant differences between activity surfaces in both data sets. These measures can assist researchers by identifying group s of models producing similar and dissimilar results, and thereby help to v alidate, consolidate, and simplify reports of statistical findings. A plura listic strategy for fMRI data analysis can uncover invariant and highly int eractive relationships between local activity foci and serve as a basis for further discovery of organizational principles of the brain. Results also suggest that a pluralistic empirical strategy coupled formally with substan tive prior knowledge can help to uncover new brain-behavior relationships t hat may remain hidden if only a single method is employed. (C) 1999 Academi c Press.