On Thu, 11 Jun 2009 14:02:34 +0200, roberto toro <[log in to unmask]> wrote:
Dear Roberto,
Looking carefully at your images it seems to me that the bias might not be too high, since
the places with high anisotropy show orientations similar to those in the "good" subject.
That said, the bias is definitively there. It might be caused by the artifact in the B3
acquisition: there is a clear ghost at the top of the image, which will cause image
degradation for that acquisition.
As a previous answer stated, you can try to estimate a global intensity drop in the image
by looking at the ventricles, or at the noise in the background. However, if the changes
in intensity are position-dependent (and if the problem is the ghosty image, this will be
the case), you won't be able to do that.
This is another reason to recommend the acquisition of more directions that those strictly
needed by your model: if one of your measurements gets corrupted, you can't use the
dataset at all.
Best,
-Pablo
>Dear FSLers,
>
>I have DTI data (in-vivo, 6 directions, 7T) from 6 mice, which I
>processed for eddy current correction, and DT fitting. It worked
>nicely for 4 out of the 6 volumes, but the 2 others had a global bias
>in V1 orientation. One volume appeared all red (V1 had an overall
>left/right direction), and the other volume appeared all blue (V1 were
>biased in the dorsal/ventral direction).
>Would there be a way of correcting (subtracting) these global biases?
>Would it be possible that the bias is produced by a global difference
>of intensity in some of the volumes encoding the 6 gradient
>directions?
>
>There's a jpg at http://dl.free.fr/q3VgvTUtr showing a single slice
>from B0-B7. Top half is a mouse that worked OK, bottom half a mouse
>with a left/right bias.
>
>thank you very much in advance!
>
>roberto
|