11 Distortion and drop-out
We cover distortion and drop-out at the same time because, while they’re not the same thing, they happen for the same reason: localized perturbations in the magnetic field.
First, we’ll look at distortion — causes, and then solutions. While all MRI techniques need to think about distortion, applications that use EPI read-outs (fMRI and DTI) have very obvious problems with distortion, and fixing distortion is crucial if you want accurate registration (alignment) between your distorted functional data or your DTI data and a reference anatomical image that does not have the same distortion.
Here is a description of 3 different approaches to distortion compensation, with sample commands for the first two.
You will have noticed in some of those images that distortion compensation didn’t produce perfect results. Sometimes, that’s because the algorithm could have done better. But mostly, if you’re using GE EPI acquisitions, it’s because there was also signal drop-out in the same region, and some of the signal simply could not be recovered. So the last two lecture segments cover the causes of dropout, and possible acquisition techniques for avoiding drop-out. Unlike distortion, there’s no data processing solution to drop-out — once you’ve lost the signal, it’s gone.
Here are some acquisition techniques commonly used to address drop-out. Happily, as we move to higher-resolution acquisitions enabled by multiband excitation, faster gradients, and parallel imaging, drop-out becomes less of a concern. So people don’t talk about it as much as they did in the early 2000’s.
Exercises
- To do “traditional” distortion compensation from “first principles” (re-shape the data based on knowledge of how the actual magnetic field at each location was altered), what 4 things do you need?
- What is the name of the tool provided by FSL to do this “traditional” distortion compensation?
- To do “blip-up/blip-down” or “forward/reversed phase encode” distortion compensation, what do you need?
- What are the names of the tools provided by FSL and AFNI to do this kind of distortion compensation?
- Not covered in the ‘lectures’ above, but very useful to figure out for yourself: if you forget to write down key parameters when you’re sitting at the scanner (phase encode direction, echo spacing, difference in echo times in the field mapping images you acquired), how can you recover that information from your data?