Ella,
It sounds like your registration procedure might be similar to that achieved by asl_gui/asl_reg depending upon exactly what images you are using where. I might assume you are using your whole head ASL data as the 'low-res structural'?
asl_reg will do: asl data to low res structural (3DOF flirt followed by 6DOF flirt ) and low res structural to structural (6DOF flirt). It currently doesn't do registration between calibration image and raw ASL, nor does it use bbr. So in practice you might find for your data that you can do better than the built in generic registration method and then apply this to the perfusion images. As it stands I don't think the unwarping will be being applied to the perfusion map if you only use the .mat file from all the steps - I am not sure how you would apply the estimated unwarping. However, I would think that your current procedure might be better than trying to use epi_reg on the perfusion image - although the grey/white contrast in the perfusion image might actually help. These are not options I have had the chance to explore and I would be interested to know what difference it makes. Sorry that I cannot give a more definite answer, but overall what you are doing seems like a good idea.
You might also consider a non-linear (e.g. fnirt) reg from structural to standard space.
Michael
On 23 Jan 2014, at 11:32, Ella Hinton <[log in to unmask]> wrote:
> Thanks, that's really helpful - I now have perfusion in the right bits of the brain!
>
> I do think it could be improved though as I have collected fieldmaps, though that I don't *think* I am currently benefiting from them. Prior to finding these great tools, I had a long pathway of registration:
> 1. Mzero(calibration image) to raw ASL (partial FOV) using flirt
> 2. ASL data to wholehead ASL using flirt
> 3. Wholehead to highres structural using bbr with fieldmaps
> 4. highres to standard using flirt
> I then concatenated the mat files and reversed it to create an ASL2Standard.mat file. This is what I used to transform the perfusion_calib map from the asl_gui results. Will this apply the advantages of B0 unwarping from the bbr stage to the perfusion map, if not directly done on the perfusion_calib map? Or is it not that simple?!
>
> I guess I am uncertain exactly what to input into the oxford_asl or gui to ensure its as good a result as I can get.
>
> thanks in advance,
> Ella
>
>
>
>
> Previous message:
>
> Ella,
>
> Those warnings are not a concern they relate to the size of the databased and that some internal variables will not be saved - but you wouldn't need these.
>
> You can still use a low res structural with oxford_asl - you should find that asl_gui will run an oxford_asl command that you can copy and reuse adding the -r option to load in the low res structural data (make sure that this has been BETed). Otherwise you can play with asl_reg. The general recommendation is to register your raw asl data (before tag control subtraction) to the structural and then apply the transformation from that to your (native_space) perfusion image that comes out of the analysis. It is always best to do the actual perfusion quantification with the original data. You might try registration with the perfusion image directly, my experience is that this is less reliable, but you might give it a try.
>
> Michael
>
|