Improving EPI

Echo-planar imaging (EPI) is commonly used for both BOLD imaging (gradient-echo EPI) and diffusion imaging (spin-echo EPI) because it acquires images very quickly. However, the price that we pay for such fast acquisitions are various image artifacts. Most of these are well-know and various methods have been devised to improve EPI images. This page summarizes work at the CNI to improve various EPI imaging problems.

BOLD contrast (in z-scores) overlaid on the mean EPI image from the timeseries. (TR=3sec, TE=25ms, 2mm^3, one 6-minute event-related run)

TIme-series plot for a voxel from the same data.

Signal drop-out

Because of the long read-outs involved, EPI is very sensitive to field inhomogeneity. Sticking a person in the scanner is guaranteed to create some fairly substantial inhomogeneities in the MR field. These field distortions are particularly problematic near air-tissue interfaces, such as the regions around the nasal sinuses and the auditory canals. Unfortunately for those interested in imaging brain regions near these structures, standard EPI images will have significant geometric distortions and signal drop-out in these regions.

Reducing the read-out duration

Shortening the read-out duration with partial k-space methods (such as ASSET acceleration) can help. At the CNI, 2x acceleration works well with our 8-channel coil. The 32-channel coil produces reasonable EPI data with 2.5x or even 3x acceleration. Note, however, that more acceleration means less data acquired per TR, and thus lower SNR.

Reducing the echo time

Shortening the echo time (TE) will also reduce EPI distortions and signal drop-out. However, BOLD contrast increases over several tens of milliseconds and is ideally sampled at a TE of 30ms (add references and signal curves). Even so, a small reduction in the TE can improve EPI image quality while retaining adequate BOLD contrast.