Hi Kelly,
On 13 Mar 2007, at 14:00, Kelly wrote:
> I have another question that I would love your opinion on. I am
> analyzing perfusion data from a multi subject project. First, I
> ran my raw images through Soren Christensen's Matlab "pgui" program
> that calculated CBF, CBV, MTT, k1,k2 and delay. My next task was to
> co-register these images to the highest degree of freedom while
> keeping the integrity of the images. I have tried many different
> methods. I came to the conclusion that my best output so far was to:
>
> 1. Create a mean image of my perfusion dataset (2x2x4 dimensions) and
> flirt with 12dof to (2x2x2) MNI_152 brain
>
> 2. Register each image, using nreg, to the mean and flirted template
> image by using CC and -ds 80 then subdivide (so -ds 40) and then
> subdivide again (-ds 20)
I doubt this is optimal. Do you not have a structural image? The
normal FEAT approach would be best, for example a 7DOF affine to the
betted structural and then 12DOF from structural to standard space.
I would not use any nonlinear reg for the within-perfusion data - do
you think you have wildly varying distortions for example? Does
mcflirt not work well for within-session analysis, and flirt (as
above) for between session (fixed effects) analysis?
In general I would think you should be using FEAT for all of this....?
> My ultimate goal is to use randomise for the analysis. But I am not
> sure what the best input is going to be for randomise. I have
> uploaded (using the link you sent me) the 4-D dataset after step 2
> above was completed. The code is 310438. Would you have a look and
> give
> me your thoughts?
>
> They still look quite different to me and when I ran this
> through randomise the first time I didn't really get any significant
> results; however, the project is looking at differences pre and post
> altitude. I know that CBF should go up at altitude; so I know there
> should be a difference in this dataset. (Paired t-test)
>
> randomise -i 4D_CBF_pre_post -o 4D_CBF_20 -v 10 -V -m min_4D.nii.gz -d
> 21pair.mat -t 21pair.con -c 3
>
> I used the minimum image of the dataset to mask the 4-D image so that
> they all have the same dimensions.
>
> Because the mean image is inherently smoothed and has less
> information, I have recently begun running the CBF image registration
> all to all using tbss_2_reg hoping that this might give me a more
> accurate result...but this takes a while=)
>
> I have read the paper on how tbss works for the FA data. I think this
> methodology is brilliant and I am planning to use it for my diffusion
> dataset that corresponds with the perfusion data that I discussed
> above. I understand why it works so well with the white matter
> tracts.
>
> I am a beginning masters student and have only recently begun to
> educate myself on MR imaging, so my knowledge is very limited in this
> subject. This may be a stupid question but Have you ever explored
> trying to make a similar program that fine tunes the coregistration of
> perfusion (like CBF or MTT) data? I understand that the real values
> that you find in FA data are not present in the perfusion data...but
> maybe normalizing to a certain value set depending on max and min or
> time would provide this?...
I would think generally that perfusion analysis would follow the same
general ideas as FMRI analysis - along with various optional ways of
solving correspondence such as using cortical modelling (e.g.
FreeSurfer) etc. - do you think that perfusion should be treated
fundamentally differently from FMRI?
Cheers, Steve.
>
> Thanks so much for your help. I really am enjoying learning FSL and
> how robust a package it is! I look forward to your reply.
>
> Kelly Brown
> University of Colorado HSC
> Altitude Research Center
>
>
> ______________________________________________________________________
> __
> Interested in getting caught up on today's news?
> Click here to checkout USA TODAY Headlines.
> http://track.juno.com/s/lc?s=198954&u=http://www.usatoday.com/news/
> front.htm?csp=24
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|