Orthogonal Polynomial Regression for the Detection of Response Variability in Event-related FMRI

Previous studies using event-related potentials (ERPs) have identified a variety of brain regions that respond to rare nontarget distractor stimuli presented in the visual oddball task. By contrast, event-related functional magnetic resonance imaging (ER-fMRI) studies using similar stimuli have found little or no response in these same regions, suggesting that ER-fMRI may be less sensitive than ERPs for detecting stimulus-evoked activity. It was hypothesized here that variations in ER-fMRI response amplitude evoked by successive stimuli may have reduced detection sensitivity in these previous studies. Multiple regression with orthogonal polynomials (OPR) was used to increase detection sensitivity by employing orthogonal polynomial equations to model such variations in response amplitude. ER-fMRI data was collected from 17 subjects during performance of the visual oddball task, which included frequent nontarget, rare nontarget distractor, and rare target block letter stimuli, to which subjects made a speeded button press response. When compared with multiple regression using main effect regressors alone, OPR identified an increased volume of significant target and distractor evoked main effect response due to reduced error variance. In addition, distractor-evoked activity was found to correlate with orthogonal polynomial regressors but not main effect regressors across a large volume of prefrontal and paralimbic brain regions including dorsal-lateral prefrontal, inferior-lateral prefrontal, and cingulate cortex. These results illustrate that considerable variability is present in ER-fMRI responses evoked by rare distractor stimuli in the oddball task. Such variability can be modeled using OPR, leading to increased detection sensitivity and better anatomical correspondence between the findings of ER-fMRI and ERP studies.