Hi again,
> Hi Jesper. Thanks for the reply.
> I understand what you are saying about the SNR in topup input. There is a point of diminishing returns.
>
> One thing I am trying to decide in an acquisition and post-processing scheme is whether I should collect A-->P and then P-->A phase encode for the full set of DTI b1000+256. Based on your answers, it seems I would not need the P-->A for the second set of 256 directions. I could simply use the topup correction from the B0 image volume. Rather, I could simply collect a second set of A-->P and then average those together. Unless...
>
> The whole point of collecting A-->P and P-->A for all the gradient direction volumes is to correct higher order eddy effects in the images that are not in the B0 volume. That was what I was alluding to in my previous post. I understand that this is an acquisition preference, but I am trying to figure out what the advantages would be given that the phase reverse blip is so distal in time to it's "partner (i.e., 16 minutes apart in collection)" and highly vulnerable to motion issues. So, why do people do this? It's got me scratching my head.
>
> Now if I do have the A-->P and P-->A for all the directions and B0, then I need to figure out the best post-processing strategy.
> I am thinking, based on my interpretation of your comments and other things, that blip up/down are used for distortion correction of the B0. Then eddy is used for correction of eddy current effects in the 256 directions with a line input for the topup output. Does eddy use the information in the acquisition parameter to correctly integrate A-->P and P-->A for both 256 direction sets? In other words, does it take advantage of the fact that you have the opposite phase encoded image. Or, does it just eddy current correct everything and average it together and not take advantage of the A-->P and P-->A pairs that were collected.
eddy will use the information in the individual pairs to get better estimates of the EC fields. These are "higher order" in that they go beyond a simple linear model (quadratic or cubic).
Hope that helps. Jesper
>
> Sorry if I am being confusing! I value your feedback.
>
> Brian
>
|