Hi,
Are you sure that your bias field image is in the same space as the EPI?
Your fslmaths command looks right, so I'd need to know more details about how you are getting the bias field.
Also, are you using BBR for the epi to structural registration? That is a lot less sensitive to bias field usually.
All the best,
Mark
On 15 Nov 2013, at 16:32, Xaver Fuchs <[log in to unmask]> wrote:
> Dear experts,
>
> I have EPI datasets which are strongly biased (probably because pre scan normalisation was turned off in the used sequence). It has turned out that the images have very high intensities in posterior and dorsal parts and show a posterior-anterior gradient. A huge problem is, that the EPIs are quite hard to register to the T1 anatomical scan.
> However, the fast's bias field reduction tool apllied to the EPI improves registration significantly. Since the restored image that fast puts out is just one volume and not the whole data set, I was wondering how to apply the bias field correction to the whole 4D set. I tried fslmaths, however, I could not quite figure out how to compute the restored image that fast produces.
> If i devide the original EPIs by the bias field using
>
> fslmaths <epidata> -div <epidata_bias> correctedepis
> the first volume of the resulting epis do not quite look like the epidata_restored that fast produces.
>
> Where is the mistake? Can you help me here?
>
> Best wishes,
> Xaver Fuchs
|