Print

---

Dear all,

We are re-running some analyses with MELODIC in FSL 5 that had run previously on FSL 4.1.9 (using BBR, which has become available in 5)
The analyses, which used to finish normally, now exit after making a HTML report of component #1.

In the main (gica) directory we see the following things

1. at the very start, before report_firstlevel.html is created, 
    2 files ./logs/feat4_gica.o31777 and ./logs/feat4_gica.e31777 appear (both size 0)
    the latest log created before that, ./logs/feat9 only contains HTML code

2. the last 25 lines in ./logs/gica are
    $ tail -25 ./logs/gica

     Convergence after 35 steps 

   Sorting IC maps

   Writing results to : 
     groupmelodic.ica/melodic_IC
     groupmelodic.ica/melodic_Tmodes
     groupmelodic.ica/melodic_mix
     groupmelodic.ica/melodic_FTmix
     groupmelodic.ica/melodic_ICstats
     groupmelodic.ica/mask
   ...done
   Creating report index page ...done


   Running Mixture Modelling on Z-transformed IC maps ...
     IC map 1 ... 
      calculating mixture-model fit 
      saving probability map:    groupmelodic.ica/stats/probmap_1
      saving mixture model fit:  groupmelodic.ica/stats/MMstats_1
      re-scaling spatial maps ... 
      thresholding ... 
      alternative hypothesis test at p > 0.5
      saving thresholded Z-stats image:  groupmelodic.ica/stats/thresh_zstat1
      creating report page ... 

3. the size of the file ./groupmelodic.ica/report/t1.png is 0 bytes

4. after ./groupmelodic.ica/report/IC_1.html is created, 
    two files ./logs/feat5_stop.o27061 and ./logs/feat5_stop.e27061 appear

5. the last file created in the gica directory, ends in these lines:
    $ tail -2 ./logs/feat1 

    /usr/local/bin/fsl502fix/bin/fsl_sub -T 1 -l logs -N feat5_stop -j 31777  /usr/local/bin/fsl502/bin/feat /user/EP2D_REST06_rest_ica.ica/design.fsf -D /user/rest_ica.gica -stop
    27061

So in the report does not show errors in the log page, and suggests that the HTML information for component #1 has been generated. But that is not true. The png file that would contain the temporal mode of the component, is empty and there is no power spectrum plot.

This problem has been reported by others (in February) although a solution has not appeared on the list. Has it been solved? Are other people experiencing these problems as well?
The only workaround we have at the moment is run the analysis with the new software, let it crash, and then re-do the ICA bit from the command line with the old software -- hardly the way to go...

Many thanks for your help,
Alle Meije Wink