Nonlinear regression of functional MRI data: an item recognition task study

A classical item-recognition task was used to examine human verbal working memory in an event-related functional MRI (fMRI) study. A highly flexible experimental design incorporated a broad variation of memory load and delay time. This design allows for only three or four repetitions per stimulus condition. In a first step, linear regression analysis was applied to locate functional activation in the fMRI data. As a second step, the time course of the hemodynamic response (HR) was analyzed using nonlinear regression, which served to quantify the dependency between HR shape properties and stimulation conditions in several regions-of-interest. On the basis of this study, a closer description of the frontoparietal network involved in verbal working memory was possible.