Dear Martin,
> finally, I tested topup and eddy on a HCP-adapted diffusion MRI dataset (96 dir, left-right and right-left). Using the "trial and error" approach I managed it all the way through. Eddy took quite a while (and I'm not talking about minutes ...).
that is why there is an OpenMP version.
> My question is: The input for eddy (i.e. the merged raw data DTI_RL.nii.gz + DTI_LR.nii.gz=192 dir) is 92 MB, but the output of eddy 490 MB. The --iout of topup is 30 MB, consisting of 6 b0-values for each encoding direction).
> Since this is the first time I tried to correct for eddy current and susceptibility artifacts, I just wanted to make sure, that this is totally normal and that I didn't miss an error (at least there was no message).
It does sound like a big difference. Does the results look otherwise fine?
eddy, and I think all FSL apps, now writes the results as float by default. If your images are int16 to start with, which is fairly typical, that makes a factor of two. Add to that the compression that, I imagine, will not be able to do an equally good job with float which I think has ~2^25 possible values compared to 2^16. I would not be surprised to see the output 3-4 times larger, and I guess your factor of ~5 isn't to far from that.
It is important though that you check your images by running for example a movie in FSL. If something has gone terribly wrong and the images are just rubbish you would expect the compression to do a very poor job and you would definitely expect large images.
I will consider adding a flag that allows you to set the output format.
Jesper
>
> Thank you very much,
> Martin
>
|