If you want an average of 10 numbers, you add them up and divide by 10. If
you want an average of 8 numbers you add them up and divide by 8. Here, the
division is by the number of scans that have data at each voxel. The theory
is that if there is a value of exactly zero in an image, then there is no
data, so it is not included in the average. Note that it may go a bit wrong
if data is present in a voxel but its value is exactly zero, rather than the
point being zero because there is no data.
Best regards,
-John
On Tuesday 11 March 2008 14:12, Veronica Garcia Vazquez wrote:
> Dear List,
>
> one of the possible preprocessing steps (as suggested by the list) for
> tensor-based morphometry of gray matter is the calculation of a so
> called "soft-mean" image for each subject. This "soft-mean" image (i3)
> can be calculated by:
> i3 = (i1+i2)./(eps + (i1~=0) + (i2~=0))
>
> where i1 is gray matter segmentation of acquisition 1 and i2 is gray
> matter segmentation of acquisition 2. this "soft-mean" gray matter image
> (i3) can then be used as follows:
>
> i4 = i3 * (Jacobian image -1)
>
> the question: what is the purpose of this "soft-mean" gray matter image?
>
> is it to maintain information on the probability of a voxel belonging
> to gray matter?
>
> thanks,
> joost
> Veronica
|