Hello Vish,
Were the data collected on a Siemens 3T Trio scanner using a partial
Fourier acquisition? If so, there is a known artifact of the sort you
described arising from table vibration coupling to the subject's head.
In some subjects the data appears fine, and in others the artifact is
quite severe. Typically (in my experience), the most obvious signal
drop-out is limited to a small number of slices in a small number of
gradient directions. We have dealt with the problem by just excluding
the gradient directions with signal dropout from the tensor estimation
for that subject.
There is a paper by Gallichan et al (HBM, 31:193-202, 2010) describing
the artifact, as well a proposed co-regessor approach for correcting for
the artifact. They also demonstrate that the artifact can be avoided by
acquiring full k-space data. Additionally, I believe that Siemens has
some sort of table modification that can reduce (eliminate?) the
artifact, so you might want to look into that with your Siemens rep (if
you indeed have a Trio scanner).
cheers,
-Mike H.
On Fri, 2010-08-13 at 20:05 +0100, Vishwadeep Ahluwalia wrote:
> Hi,
> I had attached a movie file from a DTI sequence with 30 directions and 3 averages but it didnt go through. There is slice dropout within a brain volume for certain directions. i.e 3-4 slices within the brain volume have almost no signal while other slices are ok. It looks like head motion but i'm not sure. Is someone enlighten me with the cause of this artifact? Also if the artifact takes place in only one of the repetitions can i still use the dataset by discarding the bad repetition ( and accordingly adjusting the bvecs and bvals)?
> What if the artifact is spread out and i want to remove for eg: volumes for 2 directions in the 1st repetition and volumes for 3 directions in the 2nd?
> In such a case can i still get valid results ( although lower SNR for estimation for some directions)?
> -Vish
|