Dear Jack,
> Hi All,
> I have data acquired with multiple b-values and I use topup with default parameters (specifically --scale=0). There appears to be a global scaling applied to the data and the scaling is not the same across b-values. I dealt with this by dividing the unwarped image by the mean of the unwarped image and multiplying by the mean of the original image. This normalization appeared to "fix" the problem, but I'm not sure how accurate it is. Does anyone know how the scaling is computed?
I am not really sure what you mean. There is a scaling applied internally to all images when running topup. This is done in order to make sure that there is a consistent scale for image data so that a given regularisation means the same thing regardless of the arbitrary scaling of images from different sequences/manufacturers.
However, when you apply the estimated warps using applywarp there will be no scaling and the images should have “roughly” the same mean as the original images. The “roughly” is because for a given image parts may have been shifted out of the FOV, which affects the mean.
Note that for example the —iout output is ONLY there to give a visual confirmation of how well the estimation of the field has worked (and that may have some scaling applied to is as it is literally just writing out the internal estimate). For further analysis you should always use the output of applywarp, or eddy which is increasingly replacing applywarp.
Jesper
> thanks!
>
> jack
|