Using a field-map to undistort EPIsBR[wiki:FmBackground Background]BR[wiki:FmPower Improvement in power!]BR[wiki:FmFieldmaps Acquiring fieldmaps]BR[wiki:FmDownload Download]BR[wiki:FmRequirements What you'll need to undistort an EPI data series]BR[wiki:FmReference Guide and reference]BR[wiki:FmExample Step-by-step example]BR[wiki:FmBatch Batch mode]
Magnetic field distortions
Ideally, the magnetic field throughout the bore of an MR scanner would be homogenous. Unfortunately, different materials such as bone, air & brain tissue affect magnetic fields in different ways - a property summarised by their magnetic susceptibility. When a structure comprising materials of different susceptibility is placed in a field, this becomes distorted. While some of the large-scale distortions can be removed by shimming, others remain.
Dropout & distortion
The field errors lead to two main problems. EPI sequences are especially affected by small deviations in the magnetic field. First, where the field gradients are large, there will be dephasing within voxels, and signal will be lost. Second, even in regions with there is no dropout, spatial encoding in the phase-encode direction will be disrupted, and images will be distorted. Commonly, the phase encode direction is the anterior-posterior direction, and so distortions will be along this axis.
Jezzard & Balaban (1995) proposed a solution to the problem of distortion. It is possible to directly measure the magnetic field across the head, and then use this information to undistort the EPI images after reconstruction.
The example below shows a mean EPI image before and after undistorting, and a 'structural' image from a non-EPI sequence not susceptible to distortion in the phase-encode direction. The improvement in shape is clear by eye on the coronal and sagital sections.