Hi again,

>
> Imagined you would checked it for high-bvalues. Are there any lower
> SNR-levels to be above though?

Well, with a b of 10000 there isn't much signal left, and that seems to work. Eddy uses all the info in all the voxels to estimate a few handfuls of parameters, unlike for example bedpost that try to do it on a voxel-by-voxel basis. Therefore I reckon any data that has a good enough SNR to be useful, eddy will deal with.

>
> Is there anything one should think of when generating a gradient
> scheme to make a good sampling? The half-sphere scheme on your website
> (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/EDDY) looks very uniform.

The more uniformly you can sample, the better. Note that we do NOT recommend the half sphere scheme though. If you are on Mac I can send you a little app for calculating directions?

>
> I will pilot the twice-refocusing to see, but it would be tempting to
> shorten the TE as we have strong gradients and would be able to get in
> the weighting without needed to separate diffusion gradients too much.

That was the thinking in the HCP.

Is that really your address? Will we be seeing cheap flatpack MR scanners from IKEA soon? :-)

> _______________________________
> Finn Lennartsson
> Yxnäs 24
> SE-34393 Älmhult, Sweden
> Cell: + 46 704 838907
> E-mail: [log in to unmask]