Educational objectives

· Recognise typical EPI artifacts and understand the physical mechanisms behind them.
· Learn which protocol parameters impact the severity or prevalence of the artifacts discussed.
· Discuss prevention and correction strategies, with a focus on solutions applicable in routine scanning.

Introduction

In Echo Planar Imaging (EPI) multiple echoes/k-space lines are acquired after a single excitation. It employs strong frequency-encoding gradients with interspersed low-magnitude phase-encoding gradient blips. This results in a "zigzag" traversal of k-space, where every second line is acquired with reversed readout gradient polarity. This time-efficient traversal of k-space allows to reduce the typical volume acquisition time from minutes to seconds (or even faster with recent advances in hardware and image reconstruction). Due to its unique speed EPI has become the workhorse of many applications which require dynamic imaging such as functional MRI or diffusion imaging. However, the fast acquisition also results in several challenges and characteristic artifacts, which we will discuss in this session.

Why is EPI so susceptible to image artifacts?

EPI is more prone to image artifacts than slower acquisition strategies because of two characteristics:

1. Oscillating readout polarity: Due to the oscillating polarity of readout gradients every second k-space line needs to be time-reversed. However, eddy currents in the scanner hardware or other imperfections remain static with respect to the object which can lead to inconsistencies between odd and even lines in the image reconstruction.
2. Long acquisition window: As multiple k-space lines are acquired per “shot”/excitation the acquisition time of each EPI readout is very long compared to “line-by-line” acquisition techniques. Compared to the oscillating gradients this weakness it not exclusive to EPI, but we will look at how these artifacts manifest specifically in EPI.

EPI artifacts and how to mitigate them

We will discuss the following artifacts which are commonly observed in EPI imaging. The goal is to understand the origin of these imperfections as well as learn about practical approaches on how to avoid and correct for them:

Nyquist (N/2) ghosting artifact
In practice the forward and backward echoes acquired during the oscillating EPI readout are not perfect mirror images of each other, due to imperfections such as eddy currents in response to the rapidly changing gradients. These differences between odd and even lines result in signal intensity displaced in the phase-encoding direction, halfway across the image. If the FOV has N pixels the aliased ghosted shifts N/2 pixels relative to the image, hence the name N/2 ghosting. This characteristic EPI artifact is typically corrected in image reconstruction, using navigator echoes to determine shifts between odd and even lines(1). We will cover how this correction is usually performed, as well as discuss alternative solutions such as (concurrent) field monitoring(2), GIRF-correction(3), and dual-polarity GRAPPA(4). We will also look at how acquisition parameters (e.g., oblique slices) can impact the severity of ghosting artifacts and share our experience on choices regarding the navigator echo acquisition.

Resolution loss
After excitation the signal decays with time constant T2* (or T2 in the case of spin-echo EPI), leading to a modulation of the echo amplitudes as you fill k-space This creates an exponential signal envelope in the phase-encoding direction in k-space, which in image space manifests as a broadening of the point spread function i.e., a resolution loss in the phase-encoding direction. Blurring can be reduced by shortening the EPI acquisition window (e.g., using parallel imaging).

Geometrical distortion
EPI sequences are prone to geometric and intensity distortions caused susceptibility-induced field inhomogeneities due to the low bandwidth in the EPI phase-encoding direction. Distortions can be reduced by improving the shim or increasing the effective bandwidth, by reducing the number of echoes in the phase-encoding direction. To maintain resolution the gaps in k-space are filled either via parallel imaging or interleaving subsequent shots (in-plane segmentation). Alternatively or additionally, distortions can be corrected in post-processing, for example using a B0 field-map(5) or reversed polarity data(6).

Signal dropout
Field inhomogeneity also results in local signal loss or dropout. This can be due to intra-voxel dephasing or because induced field gradients may shift the echoes outside of the restricted k-space sampling window of EPI. Strategies to reduce signal dropout include shortening TE (not always an option, e.g., for fMRI), increasing the spatial resolution, as well as angulation and z-shim(7).

Chemical shift/Fat artifacts
Fat resonates slightly slower than water (shifted by ca. 3.5 ppm), resulting in mis-localization of the fat signal. Due to the low effective bandwidth of EPI in the phase-encoding direction the fat signal can be shifted by centimetres. Fat suppression or water-selective excitation is thus necessary in to avoid that the shifted fat overlaps with the region of interest(8).

Acknowledgements

No acknowledgement found.