A method to increase temporal resolution in 3D-EPI fMRI using UNFOLD

Introduction: An fMRI study is ideally acquired at a high temporal resolution and with full brain coverage at each scan time point. This prerequisite is especially true in view of the rapid dynamic BOLD signal changes expected when cognitive tasks are studied. A cascade like brain recruitment is typically observed in complex mental exercises where neural recruitment occurs in both parallel and serial fashion. Currently multi-slice, single-shot 2D EPI is used for most applications. We propose a hybrid 3D EPI sequence instead, which is a standard 2D EPI sequence modified such that the third dimension of the excited slab is spatially encoded using 40 phase-encoding steps. So at each TR, the whole volume is excited. With this method, signal changes related to neuronal activity may be detected more reliably and may less likely be lost in gaps between slices as typically seen with multi-slice 2D imaging. Another advantage of Hybrid 3D-EPI is an improvement in SNR due to excitation of the whole volume at each repetition time, although the output image is more T1-weighted compared to the 2D-EPI [2]. Since both whole brain coverage and relatively high spatial resolution is of importance in cognitive fMRI, the temporal resolution in Hybrid 3D-EPI is slowed to an extent that analysis of time-resolved brain recruitment becomes a futile exercise. We applied therefore in this work, in order to increase temporal resolution, UNFOLD [1](in slice encoding direction) and Partial Fourier (in phase enc. direction) techniques to the 3D hybrid EPI sequence. Unlike in previous studies, UNFOLD was applied here along the slice encoding direction. This directly reduced the volume TR by fifty percent. The sequence was tested on healthy volunteers who were exposed to complex cognitive stimuli that lasted up to a time period of eight TRs. The resulting activation patterns were compared between data acquired with full kt-space coverage and the UNFOLD-PF hybrid data. Methods Experiments were performed with an 8-channel head coil on a 3T GE SIGNA machine with gradients constraints Gmax=40mT/m, slew rate S=140T/ms.