Dear all,
I am currently running multirun ICA-fix on HCP-style data (3T, resting state, TR=0.8, 4 runs of 5min). I was wondering if “HCP_Style_Single_Multirun_Dedrift.RData” is the correct option for the “–hcp_icafix_traindata” parameter. Should I not be using “HCP25_hp2000” trainingdata instead, and is this option available in qunex?
Thank you!
Best regards,
Thomas
Hi Thomas,
There are two ICAFIX implementations, the old/legacy one called hcp_fix (HCPpipelines/ICAFIX/hcp_fix at master · Washington-University/HCPpipelines · GitHub), and the “new” one called hcp_fix_multi_run (HCPpipelines/ICAFIX/hcp_fix_multi_run at master · Washington-University/HCPpipelines · GitHub). QuNex will by default use the modern hcp_fix_multi_run. If for some reason you need hcp_fix (e.g., reprocessing a couple of sessions for an old study that was already processed with hcp_fix or something like that), you need to explicitly specify this through the hcp_icafix_bolds parameters. By defaults hcp_icafix_bolds will be specified as --hcp_icafix_bolds=fMRI_CONCAT_ALL:<BOLD1>,<BOLD2>.... So it will execute hcp_fix_multi_run with their concatenated/grouped name fMRI_CONCAT_ALL. Set this as you see fit for your study. To use hcp_fix (old, single run version), do not specify the group name, just list the bolds, e.g. --hcp_icafix_bolds=<BOLD1>,<BOLD2>.... It is heavily advised to use the newer, hcp_fix_multi_run as it gives better results.
The training data to use depends on which version above you are using HCP_hp<high-pass>.RData should be used for the old hcp_fix, while HCP_Style_Single_Multirun_Dedrift.RData is to be used for hcp_fix_multi_run.
FSL somewhat recently introduced a Python version of fix called pyFix (the two versions above are based on MATLAB and R). This Python version seems to be giving even better results. It should be integrated into HCP and then into QuNex soon.
Best, Jure
Dear Jure,
Thank you for the explanation, that really clears things up!
I will use the multi_run version of qunex then. Also, the python version seems interesting, so I will definitely look into that.
Thanks again!
Best,
Thomas