On Monday 19 November 2007 09:07, Graham wrote:
- DVE's description of the FI calc, which it looks like FS implements
(assuming FS's k calcs are faithful -- that code is hard to follow.)
Yep - that the code can be hard to follow. Still, you can assume that FS's k calcs are faithful. But on deeper reflection, the FI calculation in FS has a core area-dependency in the code that probably should be normalized out.
a) Cylindricallity of gyri and sucli do not contribute equally -- one is emphasized and the other suppressed. (KMax,Kmin = (5,0) --> FI of 25, vs KMax,Kmin = (0, -5) --> FI of 0; I think these are the gyral and sulcal versions of the same shape, right?)
Ok - I suspect you need to think of Kmax and Kmin (also known as the principle k1 and k2 curves) in absolute terms when comparing. The sign really only indicates if the surface is curving "inwards" or "outwards", and doesn't convey comparative magnitude. So, in the case where one of the principle curves is -5 and the other is 0, k1=-5 and k2=0 (Kmax=-5 and Kmin=0), resulting in the same values for gyri and sulci.
b) Certain moderate saddle shapes contribute *negatively*, though the *extreme* saddle shapes contribute *zero*.
This I can't immediately comment on, but even if there is contribution from a saddle shape (i.e. k1 should be approximately k2), so the FI should be close to zero.
Switching tracks, you might be interested in the re-vamped 'mris_curvature_stats', which has kind of morphed into the main curvature measure tool in FS.
This tool now computes several principle curvature functions on a given surface (optionally saving each as a curvature file), and also determines area- and vertex-normalized integrals for the different functions. Furthermore, the functions and integrals are also computed for only positive and negative curves (or Gaussian curvatures) separately, so sulci and gyri are easily distinguishable.
One of the measures is a so-called "sharpness function", which is conceptually related to Van Essen's FI:
Sharpness = (k1-k2)^2 Van Essen FI = abs(k1)*(abs(k1)-abs(k2))
The sharpness function gives an indication of how sharply folded a gyrus or sulcus is.
What version of 'mris_curvature_stats' are you running? It has seen some active development over the past summer, so you might not have the newest version. Give me a shout if you'd like the current dev version.
Cheers -=R