> I'm afraid I still don't understand the (i2-1)... why is it not just
> i1.*i2 to "modulate" the GM with the HDW Jacobian dets?
Assuming that the warping is exact (a big assumption), then the warped grey
matter of one image will be identical to the unwarped grey matter of the
other. The expression is essentially i1.*i2 - i1, which is the modulated
image (theoretically identical to a modulated warped other image) minus the
unmodulated image.
>
> Also, I see your point about smoothing log(det), but there are also
> statistical reasons why people want to take logs; do you think it
> would be okay to smooth first and then take logs?
You hit problems in regions containing only zeros.
>
> I also wonder if the Jacobian fields are naturally smooth enough to
> avoid the need for smoothing -- I think I saw a paper where they'd
> used just a tiny amount of smoothing (e.g. FWHM 1mm, rather than
> something like 8-12) on the Jac dets, and got good results, though I
> guess this depends on the form of regularisation in the warping...
It depends on the accuracy of the inter-subject registration. If it was 100%
exact (extremely unlikely to be feasible), then there would be no need to
restrict the analyses to a particular tissue class and smooth. You could
just analyze the unsmoothed Jacobians. However, as inter-subject
registration is not so accurate, then I figure that it would be a nice idea
to look at smoothed tissue class volumes.
Best regards,
-John
|