 |
 |
On Monday, September 12, 2011 11:56:04 you wrote: > Sorry to revive an old thread -- I am trying to re-threshold some > components at a different --mmthresh, so I thought I would try using the > --ICs and --mix options described here. As a test case I used the same > --mmthresh as I did in the original run of melodic, --mmthresh=0.5, but I > got totally different thresh_zstat files! The intensities in these files > from my original runs ranged from around -100 to 100, but the newly > generated files have intensities ranging from just -0.1 to 0.1. Am I > missing something obvious here? The command I am using for the > re-thresholding is: > > melodic --in=list_of_filtered_input_files --ICs=melodic_IC > --mix=melodic_mix --Ostats --outdir=rethrsh_out --mmthresh=0.5 --verbose > > In addition, if I append the --report option, I get a segmentation fault: > > Melodic Version 3.10 > > Melodic results will be in rethrsh_out > > Reading data file subject1-mcf-brain-FWHM_4mm-in_MNI152 ... done > Reading components melodic_IC ... done > Creating mask ... done > Reading mixing matrix melodic_mix ... done > > Running Mixture Modelling on Z-transformed IC maps ... > IC map 1 ... > calculating mixture-model fit > saving probability map: rethrsh_out/stats/probmap_1 > re-scaling spatial maps ... > thresholding ... > alternative hypothesis test at p > 0.5 > saving thresholded Z-stats image: rethrsh_out/stats/thresh_zstat1 > creating report page ... Segmentation fault As a partial followup to my own question, looking at lines 365-369 in melodic.cc we seem to be getting a different ICmap depending on whether or not we know the inverse of the standard deviation of the noise. During an ordinary run of melodic this information would be in memory (and would be output as Noise_stddev_inv.nii), but when running melodic with --mix and --ICs this information is presumably not in memory. Could this be why I'm getting different zthresh_stat files with --mix and --ICs?