Dear Aditya,

The segmentations we produce are partial volume segmentations,
so it is not enough to know how many non-zero voxels there are.
The values in the voxels represent partial volumes and these should
be added up to give the most accurate estimation.  We definitely see
accuracy advantages by doing things this way.  That is why the values
need to be multiplied.

As for TIV, you can calculate the WM and CSF volumes the same way, but
the bounding CSF is highly influenced by the mask used, which comes from
BET.  So all you get with this method is the volume of the BET mask which
is not tuned to give a very precise measurement of TIV.  Hence we don't
recommend this for very precise measurements.  For normalising by TIV
this can be reasonable to do, but it does depend on the quality of your BET

For documentation, the main source is
and what is on the email list.  I don't think there is any specific documentation
just about the files that are produced.  The ones of interest are detailed in
the relevant sections of the above documentation.

All the best,

On 27 May 2011, at 16:50, Aditya Kumar Kasinadhuni wrote:

> Hi all,
> Can anyone please help me with answers to a couple of questions I have? 
> First, to calculate the total intracranial volume why is it necessary to multiply the outputs from the following command
> fslstats filename -V -M
> I was thinking the the image "filename_struc_GM.nii.gz" consisted of only gray matter and hence knowing the voxel size and number of voxels would be sufficient to calculate the GM volume. If so, can we not extend this to WM and CSF to calculate their volumes and eventually the TIV??
> Second, 
> Is there any place where I can find the description of all the files listed in the /struc directory? I tried to look at the documentation but I think it lists only that from the /stats directory.
> Can someone please throw some light on these doubts that I have? 
> Thank you.