[Bruce wrote]
sounds reasonable to me, but we should call it something else (normalized FI/CI?)
Yes, dividing out area does look like a first step, justified by the way the per-vertex values are calculated. This gives the result units of mm^-2, which maybe has an interpretation something along the lines of "folds per square millimeter".
But, as noted in off-list emails, I think it would be premature to release an "NewFI" revised only in this way, because it suggests that the other principles of FI are sound, which I'm currently not confident about.
Rudolph and I are in the process of reconciling our understandings of how the FI calculations currently behave, and whether that was what was intended (and probably whether what was intended is even a useful idea).
Anyhow, hopefully the outcome of this will be better calcs implementing something useful, (with documentation of intent and usefulness :-).
Graham
Very interesting debate you keep here, gentlemen.
I went on and divided FI by the area in one reference and one pathological subjects (and age matched references). Now it seems that the "NewFI" is quite correlated with cortical thickness. It even decreases from the frontal, through parietal to occipital lobes and slight thickness perturbation with respect to the normals also affects the "NewFI".
On Tuesday 20 November 2007 21:50:52 Graham Wideman wrote:
[Bruce wrote]
sounds reasonable to me, but we should call it something else (normalized FI/CI?)
Yes, dividing out area does look like a first step, justified by the way the per-vertex values are calculated. This gives the result units of mm^-2, which maybe has an interpretation something along the lines of "folds per square millimeter".
I observe that the "NewFI" almost exclusively increases its value even if the gyri get wider in the pathological subject. In this case I would rather expect the "folds/mm^2" to decrease.
Martin
freesurfer@nmr.mgh.harvard.edu
-
Graham Wideman -
Martin Kavec