Hi,
We faced unexpected thresholding results from the nonspatial mixture
modelling applied for z-score maps. There are hardly any other values in the
probmap than ones, thousands of them, and some tens of zeros and
additionally very few values between 0-1. Gamma-Gaussian fits look ok (attch
mmfit.png) and mm command usage should be ok (mm --mask --sdf --timingon
--ns -V --logdir).
So in the end there is not much thresholding done for our z-score maps as
mixture modelling for some reason assigns for almost every gamma-fitted
voxel a value of 1 in the probmap. Something seems to go wrong, any help
greatly appreciated!
BR,
Tuomo Starck
|