We also experienced a problem with the FLIRT based motion correction. We fixed it by using mcflirt instead. I can also post our version of the MotionCorrection.sh script if that would be helpful.
-Greg Dear all, I have made changes to two files to make use of FSL's mcflirt instead of HCP's flirt-based mcflirt_acc: * in Pipelines/fMRIVolume : GenericfMRIVolumeProcessingPipeline.sh -> GenericfMRIVolumeProcessingPipeline_mcflirt.sh At line 121: I extract the midvolume as reference image when no SBref is provided (the previous behavior was to use the first volume, which may be problematic); At line 159: I call MotionCorrection_MCFLIRTbased.sh instead of MotionCorrection_FLIRTbased.sh * in Pipelines/fMRIVolume/scripts: MotionCorrection_FLIRTbased.sh -> MotionCorrection_MCFLIRTbased.sh Motion correction is now done with mcflirt (see line 18 if you want to change parameters), and files are renamed and moved around to match the HCP pipeline So, with these two new files, all you have to do is call GenericfMRIVolumeProcessingPipeline_mcflirt.sh instead of GenericfMRIVolumeProcessingPipeline.sh. Hope this helps someone. - Julien Julien Dubois Postdoctoral Scholar California Institute of Technology, Pasadena, CA http://www.emotion.caltech.edu/~jdubois On Sun, Dec 14, 2014 at 11:31 AM, Julien Dubois <[email protected]<mailto:[email protected]>> wrote: > > Thanks Matthew. My data is 2.5mm isotropic, slightly lower resolution > than your 2mm isotropic but not exactly low res. > > I'm curious as to why you guys wrote your own mcflirt (seems like > reinventing the wheel). Was FSL's mcflirt failing on HCP data? > > If no one else is working on a patch like you describe, I'll probably > whip one up for my purposes so I'll keep you in the loop. > > - Julien > > > > > > > On Dec 14, 2014, at 12:45 PM, Glasser, Matthew > > <[email protected]<mailto:[email protected]>> > wrote: > > > > Hi Julien, > > > > You are not the first to report this for lower resolution fMRI data. > > We?re not completely sure why the performance is different in these > cases, > > however we would definitely be interested in a patch for the > > fMRIVolume pipeline that allowed choosing to use mcflirt instead of > > mcflirt_acc for people who mcflirt_acc was not working correctly. > > > > Thanks, > > > > Matt. > > > >> On 12/14/14, 11:11 AM, "Julien Dubois" > >> <[email protected]<mailto:[email protected]>> wrote: > >> > >> Sorry, ignore what I said for the aside: you take the average of 10 > >> volumes starting at volume 11. Better than taking the first 10. > >> Sorry > for > >> my Sunday early morning confusion :-) > >> > >> But the main issue remains, that mc flirt_acc does not correct for > motion > >> as well as mcflirt. Let me know your thoughts on why this may be. > >> > >> Thanks, > >> - Julien > >> > >> > >> > >>> On Dec 14, 2014, at 10:14 AM, Julien Dubois > >>> <[email protected]<mailto:[email protected]>> > wrote: > >>> > >>> Dear all, > >>> > >>> I want to weigh in here: the current default behavior of the > >>> pipeline is not doing a good job at motion correction (in my own data). > >>> > >>> First a quick digression: > >>> The code in global/mcflirt_acc.sh indicates that you are taking > >>> the 10th volume as a reference, in the absence of a supplied reference: > >>> line 29: > >>> ${FSLDIR}/bin/fslroi $input ${output}_ref 10 10 I think that maybe > >>> you meant to take the average of the first 10 volumes (as Greg B. > >>> indicated, and given the following lines of code); so line 29 > >>> should likely be changed into: > >>> ${FSLDIR}/bin/fslroi $input ${output}_ref 0 10 > >>> > >>> But I don't think this is where the issue comes from. Indeed, > >>> running mcflirt with the 10th volume gives a fine result. There is > >>> thus something else in mcflirt_acc that is not functioning > >>> properly (at > least > >>> for data other than HCP), compared to the standard FSL's mcflirt. > >>> > >>> Can you give the rationale for writing your own mcflirt function > >>> (mcflirt_acc.sh), and explain how it differs from FSL's mcflirt? > >>> (I could also go compare with the source code for mcflirt, but I > >>> thought I would ask directly since you probably had a good reason) > >>> > >>> Thanks! > >>> - Julien > >>> _________________________________________________ Gregory Book Senior Technology Manager Olin Neuropsychiatry Research Center, Institute of Living, Hartford Hospital 200 Retreat Avenue Hartford, CT 06106 Tel: 860-545-7267 Fax: 860-545-7797 [email protected]<mailto:[email protected]> http://nidb.sourceforge.net This e-mail message, including any attachments, is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure, or distribution is prohibited. If you are not the intended recipient, or an employee or agent responsible for delivering the message to the intended recipient, please contact the sender by reply e-mail and destroy all copies of the original message, including any attachments. _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
