Dear Julio,
It is not meaningful to compare the second eigenvector of dtifit with the
second fibre estimate of bedpostx.
Dtifit performs the DTI-single tensor model regression and V2 will be the
second axis of the fitted ellipsoid. Only V1 can be directly interpreted as
a fibre orientation, therefore dtifit can resolve one orientation per voxel.
Bedpostx fits a multi-tensor model (i.e. many ellipsoids) with some
geometrical constraints on each tensor. Each dyad corresponds to a different
fibre orientation and a different compartment of the model. Whereas V1, V2
and V3 from dtifit, all refer to the same model compartment.
Hope this is helpful,
Stam
----- Original Message -----
From: "Julio Duarte" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Wednesday, October 06, 2010 9:40 PM
Subject: [FSL] BedpostX vs dtifit
Dear FSL experts,
I'm comparing the error histograms for the masked dyads2 (threshold 0.05)
obtained with bedpostX and second eigenvector (_V2) on the same voxels using
dtifit after registering 2 images. BedpostX errors (relative to the
reference image) below 60 degrees are consistently lower in frequency than
those computed using dtifit, but after 60 degrees the BedpostX errors are
consistently higher in frequency than those obtained with dtifit. On the
other hand, comparing the error histogram for dyads1 from BedpostX vs first
eigenvector from dtifit are very similar (as I expected).
Have you ever compared these 2 methods along the second eigenvectors? Do
these results have an explanation?
Julio
|