Hi Mark,
> Say I don't have blip-up-blip-down data to run topup. If I just run eddy on that data, will eddy only model the effects of (i) EC-induced distortions? Or will it also model the effects of (ii) subject movement (and (iii) gradient heating), despite the fact that I won't have a --topup input ?
eddy will explicitly model EC distortions and subject movements, and implicitly also gradient heating since it will look exactly as EC distortions from eddys perspective. How well it will model these things depend on your data. Having blip-up-down data is very helpful, but I have seen very good results also without when the diffusion has been sampled over the whole sphere.
> If so, with high b-value data (where SNR is low), will it help eddy to model the effects of subject movement if we collect b=0 images interspersed throughout the data acquisition (i.e. every 10 or 20 directions we collect a b=0 image) as opposed to collecting b=0 images just at the start of the acquisition ?
I see how you are thinking and it is true that one could estimate b0 movements first and then for example interpolate between the b0's and use that as starting estimates for the dwi estimation. At the moment eddy doesn't do that. The reason for that is partly that it has worked very well without on our data, partly that we often collect data in multiple sessions and an interpolation between b0's migt not work very well without information where the session breaks are.
My suggestion is that you give it a go with your data and sees how it goes. Note also that we now recommend --fwhm=0 when running eddy.
Jesper
|