On 13.09.2007, at 17:39, Matthew Webster wrote:
> We've now fixed this bug in fast4 - the corrected version will be
> in the next patch - coming soon!
We've now retried the fast4 T1/T2/prior-probabilities segmentation
with FSL-4.0.2, and while there are no crashes any more, things still
do not work right: The whole brain gets classified as tissue one,
like in a brain mask.
The only hint are a lot of NaN values in the log messages:
> Beginning prior-based initialisation
> Tanaka Iteration 0 bias field 4
> CHANNEL 1 CLASS 1 MEAN 1 STDDEV nan CLASS 2 MEAN 1 STDDEV nan
> CLASS 3 MEAN inf STDDEV nan
> CHANNEL 2 CLASS 1 MEAN 1 STDDEV nan CLASS 2 MEAN 1 STDDEV nan
> CLASS 3 MEAN 1 STDDEV nan
> Tanaka Iteration 1 bias field 4
> CHANNEL 1 CLASS 1 MEAN nan STDDEV nan CLASS 2 MEAN nan STDDEV nan
> CLASS 3 MEAN nan STDDEV nan
The command we used was
> fast4 --verbose -o fast4mmpp/AKYY_1_seg --outbias -a identity.mat --
> segments -S 2 AKYY_1_t1_brain_std AKYY_1_t2_brain_std
Should I upload the data for you to have a look?
Regards,
Timm Wetzel
|