Print

Print
Hello all,

I am having some issues running my group level analysis on 6 participants in two different groups. It its a small sample because I am testing my processing pipeline as a new FSL user.

Here is my command:

melodic -i FRA_1_denoised_data.nii.gz,FRA_3_denoised_data.nii.gz,FRA_4_denoised_data.nii.gz,FRA_6_denoised_data.nii.gz,FRA_8_denoised_data.nii.gz,FRA_12_denoised_data.nii.gz -o groupmelodic6.ica -v --nobet --bgthreshold=10 --tr=3.00 --report -d 30 --mmthresh=0.5 --nl=gauss -a concat --Sdes=ICA_GLM6.mat --Scon=ICA_GLM6.con —Oall


Note: All in purple is the list of my preprocessed (including motion scrubbed, denoised and "featregapply-ed") resting state functional connectivity data


This is the abbreviated version of my error message:

…

No convergence after 500 steps

Starting PCA  ... done

Start whitening using  11 dimensions ... 

  retaining 46.5897 percent of the variability 

 ... done


Starting ICA estimation using concat


  Step no. 1 change : 0.831782

  Step no. 2 change : 0.937099

…

  Step no. 492 change : 0.880205

  Step no. 493 change : 0.503637

  Step no. 494 change : 0.852382

  Step no. 495 change : 0.987818

  Step no. 496 change : 0.960082

  Step no. 497 change : 0.886395

  Step no. 498 change : 0.883855

  Step no. 499 change : 0.851147

  No convergence after 500 steps 


No convergence -- giving up 


It is absolutely possible that I made a mistake on a previous step before tying to do group level analysis.


Let me know if you have any suggestions as I am a little lost. I read somewhere in the email list about using binary masks but I don’t know where that needs to be put in my analysis or if they require thresholding or not.


Thank you for your time.


Noelle