The impact of susceptibility gradients on cartesian and spiral EPI for BOLD fMRI

High sensitivity to magnetic susceptibility changes and accurate localization of functional activations are key requisites for pulse sequences used for BOLD fMRI. This paper seeks to develop a framework for analysing the performance of various k-space sampling techniques in this respect, with special emphasis on spiral EPI (spiral) and cartesian EPI (EPI) and their performance under influence of induced field gradients (SFGs) and stochastic noise. A numerical method for calculating synthetic MR images is developed and used to simulate BOLD fMRI experiments using EPI and spirals. The data is then examined for activation using a pixel-wise t test. Nine subjects are scanned with both techniques while performing a motor task. SPM99 is used for analysing the experimental data. The simulated spirals provide generally higher t scores at low SFGs but lose more strength than EPI at higher SFGs, where EPI activation is offset from the true position. In the primary motor area spirals provide significantly higher t scores (P < 0.0002). In-plane variation of EPI is higher in phase-encoding direction than in frequency-encoding direction (P < 0.003). In the low SFG areas spirals provide stronger activation than EPI and less spatial variability. Thus, spirals are recommended for fMRI in motor area and language areas.