Dear Stuart,
>
> I have some diffusion data which has been obtained with 60 directions with a b-value of 3000. However we have found that the gradients are not oriented about the full or half sphere. Is it still ok to run eddy on such data?
it depends on how far away from a whole/half sphere they are. When I first conceived of eddy I thought that it would depend crucially on having data on the whole sphere. Experience, and also testing/validation by Mark Graham, has shown that it isn’t as critical as I first thought. As long as you add --slm=linear to the command it seems that eddy works almost as well on the half sphere as on the whole.
I don’t have much experience of data that is neither on the half or whole sphere. Where do these data come from, and how have the directions been optimised? The half sphere optimisation was originally conceived as a way to obtain optimal data for estimating diffusion parameters, and had nothing to do with motion/EC correction.
If you want to take a look at what your directions look like you can load them in to Matlab and display them with something like
bvecs=load('bvecs’);
figure
plot3(bvecs(1,:),bvecs(2,:),bvecs(3,:),’*')
(with possible reservation for the ‘ that my mail program insist on messing with)
You can then rotate the plot to see what it looks like.
> If not, is there any way we could process the data which would achieve a result comparable to eddy?
I don’t have any experience of using anything else. I think with a b-value of 3000 many other methods might struggle a bit.
Jesper
>
> Kind regards,
>
> Stuart
|