Dynamic nonlinearities in BOLD contrast: neuronal or hemodynamic?

A primary goal of the functional MRI (fMRI) methods development is to characterize the relationship between the blood oxygenation level-dependent (BOLD) signal changes and neuronal activation. Recent studies of blood oxygenation level-dependent (BOLD) signal responses have demonstrated nonlinear behavior with respect to stimulus duration. Specifically, shorter duration stimuli produce larger signal changes than expected from a linear system. The precise reasons for this nonlinearity are not clearly understood. The goal of this study is to further clarify the origin of dynamic BOLD contrast nonlinearities-either neuronal or hemodynamic or both, by a combined approach of task timing modulation, spatial mapping, and modeling/fitting of the BOLD response using the "Balloon model." In this study, we found that (1) in agreement with the literature, the dynamic BOLD "on" response is nonlinear and has significant spatial heterogeneity. Spatial maps of nonlinearity, while highly reproducible, do not correlate with maps of the BOLD response magnitude or latency, but do show some correlation with functional segregation; (2) the dynamic BOLD "off" response is sublinear; and (3) while data fitted with Balloon model hemodynamic parameters, assuming linear neuronal input, generally create nonlinear dynamic BOLD responses, the Balloon model was not able to fit all BOLD contrast response task timing modulations simultaneously. These findings suggest that the dynamic BOLD response may be a linear function of the neuronal input function and that the neuronal input function is not a simple "on/off" boxcar function, but rather a nonlinear function that has an initial overshoot that lasts for approximately 4 s until reaching a steady state. (C) 2002 Elsevier Science B.V All rights reserved.