Hi Peter,
nope - because it is FA. This is quite a typical vibration artifact. The signal void in x artificially / spuriously increases FA there as if there were a commissure tract crossing the falx.
Darren - have you tried the --cnf flag option in dtifit?
Cheers,
Andreas
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Kochunov, Peter [[log in to unmask]]
Gesendet: Samstag, 26. März 2011 22:12
An: [log in to unmask]
Betreff: Re: [FSL] Siemen's vibration artifact and Gallichan correction
Hi Darren,
The artifact in your image doesn't look like a typical vibration artifact to me. If I were to hazard a guess, I'd say that it is a gradient spiking artifact. Vibration artifacts mainly produce voids. This looks like you had more than one k=0 lines. If you observe any zebra stripes or dramatic elevation of the background noise in one or several b>0 images that would be a spiking artifact.
Overall, I'd say these data are not recoverable.
pk
________________________________________
From: FSL - FMRIB's Software Library [[log in to unmask]] On Behalf Of Darren G [[log in to unmask]]
Sent: Saturday, March 26, 2011 4:04 PM
To: [log in to unmask]
Subject: Re: [FSL] Siemen's vibration artifact and Gallichan correction
Thanks for the replies.
Peter and Arman- we have had our scanner table fixed by Siemens and that has solved the problem. With the fix implemented we can get data with partial fourier without any dropout. However, I also wanted to correct some of the data we acquired.
Katie- As you have also noticed, unfortunately the data doesn't seem to correct as nicely as in the article. I have attached 2 FA maps showing the artifact as high signal regions. The arrows roughly delineate the ends and middle of the artifacted regions in the original (top) and fixed (bottom) images, and there is less artifact in the fixed image. (The windowing is the same for both images.)
Also, although the article seemed to show cutoffs of < abs(+/- 0.6) we found some areas still showing the artifact with an x-gradient as low as 0.45. We were able to generate plots quite similar to the Gallichan paper and fit a Tukey distribution as suggested, but modeling the function and including the regressor does not correct all areas. It wonder if it would work better with a voxel by voxel correction, but dtifit doesn't allow this. We have also thought of trying to include more than one regressor, by estimating a different regressor for the voxels not fixed the first time through, but have not yet tried this.
If others have had more success than this, I would appreciate feedback and suggestions.
Regards,
Darren
|