On 25/06/2009 09:05, "Christoph Blau" <[log in to unmask]> wrote:
> Hi everyone,
>
> following some success by a colleague of mine in getting fsl-vbm to
> work on rat brain data, I am following the tutorial at:
> http://www.fmrib.ox.ac.uk/fsl/fslvbm/index.html
>
> As my colleague found out, rodent brain extraction is not similar
> enough to human brain extraction to reliably use fsl's BET. For the
> record, the Brainstrip plugin to MIPAV[1] requires minimal manual
> tweaking of the mask after running the algorithm and is fairly
> feasible, so long as your sample sizes don't run into the hundreds.
Some time ago some of my colleagues had the same problem. I modified the bet
code so that instead of starting with an initial guess of a sphere, it
started with a spheroid with a different radius in each direction, roughly
2:1 in favour of the long axis. We had to supply an initial value for the
centre coordinate and the average radius, but this works quite well since
the rodents were usually well positioned in the same part of the coil. We do
have to manually clean up the anterior and posterior of the brain sometimes,
but other than that the process is easily scriptable.
Dave
>
> My question: I now have the extracted brain datasets (.nii format;
> orientation as required by fslvbm), and would like to see how good a
> job fslvbm_2_template would do on these data. Is there a way of
> running fslvbm_1_bet so it does all the file conversions _without_
> performing any extraction, so that I can put my extracted data in a
> format that fslvbm_2_template can work on? I think tbs_1_preproc used
> to be able to do this, but it is no longer used in the updated fslvbm.
>
> Also, as I supect that a 2mm cube is not really good enough resolution
> to perform accurate registration on a rat brain, I have simply 'blown
> up' my datasets by multiplying the voxel sizes of my rat brains by 10
> in the image header. Does anyone see a problem with doing this?
>
> Many thanks for your help,
> Chris
>
> [1] See: Bazin et al (2007), J Neurosci Meth 165:1, 111-121.
>
> --
> Christoph Blau, MVB
> Trinity College Institute of Neuroscience
> Lloyd Building
> Trinity College
> Dublin 2
> Ireland
>
|