Abstract
The latency variability of hemodynamic responses between activated voxels, within the same functional area, was studied by cerebral blood flow- (CBF-), cerebral blood volume- (CBV-) and blood oxygenation level dependent- (BOLD-) based functional MRI (fMRI) using visual stimulations. This study consisted of two parts. The first part included CBF- (using slice-selective inversion recovery, SSIR) and BOLD-based (BOLD1) experiments performed on six subjects, while the second part included CBV- (using non-slice-selective inversion recovery, NSIR) and BOLD-based (BOLD2) experiments performed on three subjects. Fifty (Part I) or thirty (Part II) repeated single trials, each with 1-s visual stimulus followed by 13-s (Part I) or 19-s (Part II) fixation, were conducted in each experimental run. From each activated voxels, different numbers of trials were randomly selected for averaging, in order to obtain signal time courses with different contrast-to-noise ratios (CNRs). At each specific CNR level, the latency of the averaged signal time course was determined on a voxel-by-voxel basis. And the latency variability was then calculated as the standard deviation of the latency distribution within the visual area (△τ). For all the fMRI techniques, the measured CNR increased and the △τ decreased as more trials were averaged. The CNRs obtained by CBF- and CBV-based fMRI were significantly lower than that by BOLD, when the same numbers of trials were averaged. The relationship between the △τ and the CNR (or number of trials selected for averaging) was empirically determined for all the experiments. Under the same CNR levels, both the SSIR and the NSIR results appeared significantly lower latency variations than those obtained from BOLD. No significant differences were observed between the SSIR and the NSIR results, in both sensitivity and latency variability. This study demonstrated that the CNR could be significantly improved by single-trial averaging, which led to improved temporal variability of fMRI responses. When comparing at the same CNR levels, lower △τ for the SSIR and the NSIR than the BOLD may be due to their improved spatial localization to the brain parenchyma.