Dear All,

 

I have used CVS to align around 180 scans to a single, representative subject. I have then used BBR to align the mean b0 image to register the diffusion data to the native space T1 and then apply CVS warps; however, the resulting aligned images for calculated maps from the DWIs are very poor with a great deal of heterogeneity, especially at the boundary but also within white matter. The coefficient of variation of the mean b0 BB registered to the T1 is around 30% globally (excluding the edges of the brain) and 40% or greater for FA. I have reviewed the initial rigid registration of the b0s to the T1s for gross misalignment and found that FLIRT performed pretty well and didn’t fail for any of my subjects and didn’t require manual adjustment; however, the result of after applying the CVS warps is quite poor. The data T1s are 1mm (MPnRAGE) and the b0s are at 2mm. The DWIs were corrected for eddy currents, Gibbs’ ringing, rician noise, and B0 distortions and look pretty great. The DWIs are 70x180x180 at 2mm isotropic and the T1s are 256x256x256 at 1mm isotropic. Are there any suggestions for this issue? I have tried re-aligning the DTI parameter maps to the b0s before applying the CVS warps (although they should be in the same space) and I have tried using both the –t2 and –dti, as well as with/out the –fsl-init flag. Is bbregister shifting the origin for the images at all? Any help would be appreciated.

 

Cheers,

-Austin

--

Austin M. Bazydlo, MS

PhD Candidate

Dept. of Medical Physics

University of Wisconsin-Madison