Okay Ziad, I'll bite...

So, if the centers are in the same orientation as the dataset,
then it should be not toooo painful to apply an expression like:

-expr "a*step((x-$xc)*(x-$xc)+(y-$yc)*(y-$yc)+(z-$zc)*(z-$zc)-$r_squared)"

Here the center is stored in the variables, xc, yc and zc, and
then there is r_squared.

If you want to do multiple spheres per dataset, it might be good
to start with a zero'd dataset, and iteratively add to it. Init
dataset "source+orig" to zeros, and repeat the above only using
'-expr "a+step(...)"' instead of "a*step". Maybe it would be even
better to use "step(a+step(...))" to handle overlapping spheres.

Ahhh, the fun to be had...

- rick