On 06/09/2012 03:52 PM, Mark Jenkinson wrote:
> Dear Stefan,
>
> The fslstats command is used on the partial volume output from FAST.
> Multiplying the mean by the volume is just another way of summing up the partial
> volume contributions from all voxels. Using -V instead just gives you the number
> of non-zero voxels, but we want to take into account the various partial volume
> values (e.g. 0.15 representing 15% of that voxel being a certain tissue-type).
> Doing it this way, by taking into account the partial volume, gives more accurate
> results.
>
> I hope this answers all three of your questions.
> All the best (from China),
> Mark
>
>
> On 9 Jun 2012, at 03:54, Stefan Kreisel wrote:
>
>
>> Hi all.
>> Looking at the sienax script it says somewhere near the bottom - if I've chosen the -2 option (just for simplicity) - ubrain=`echo "2 k $xa $xb * 1 / p" | dc -`; $xa comes from the first output of the fslstats -m -v command just a couple of lines above, $xb. ubrain is what my "raw" brain volume is going to be. There's a couple of things I keep tripping on:
>> 1.) Why would you multiply the mean intensity of all voxels by the mm^3 of your image! (probably some nasty mathematical trick - right); I would have though it be easier to fetch the volume from fslstats -V.
>> 2.) Having said that - why is that multiplication slighly off in comparison to fslstats -V (say 650507.49 in the report.sienax to 650511.125 using fslstats -V)?
>> 3.) @ 2.) I first though it was a rounding inaccuracies - but probably not. If I run sienax without the -2 option and get the full breadth of tissue types (and no longer binary images of each) fslstats -V fails in full futility and one falls back to having to multiply mean by volume of image...
>> So where's the trick I'm missing?
>> Cheers and hope all of you are in China!?
>> Stefan
>>
>>
Perfect - thanks Mark!
|