Re: [HCP-Users] Minimal Preprocessing Pipeline (on my own data): Flipped EPI (AP)

[Glasser, Matthew]

What I am most concerned about for low resolution T2w is the accuracy of the cortical surface models and the myelin maps.  Also, the T2w is used as a part of the bias field correction, so you’ll want to have a look at the ${StudyFolder}/${Subject}/T1w/T1w_acpc_dc_restore.nii.gz images to make sure they don’t look “crappy” relative to the ${StudyFolder}/${Subject}/T1w/T1w_acpc_dc.nii.gz images. I would focus your attention on the initial rigid alignment done in epi_reg (before BBR and the distortion correction).  This step we have found is sometimes not robust, particularly when acquisition protocols deviate from the recommended ones.  It isn’t BBR based and doesn’t yet have the data in rough alignment (generally pipeline failures occur at initial stages, if the initial stages work, generally everything else is more robust).   I understand you are working with legacy data, it’s just important to note that the amount of work required to make legacy data work with the HCP Pipelines is not indicative of the amount of work required when following the image acquisition guidelines.  We’re still willing to provide suggestions as to what to look at when getting the legacy data to work, but cannot fully support legacy data (e.g. you have to hack some on your own and figure it out).  Also, not all pipeline changes required when working with legacy data may be generally applicable to all data (so we’ll consider this when considering patches that come from trying to get legacy data to work).   Peace, Matt. From: Julien Dubois <jcrdub...@gmail.com> Date: Tuesday, April 21, 2015 at 11:22 AM To: Matt Glasser <glass...@wusm.wustl.edu> Cc: "hcp-users@humanconnectome.org" <hcp-users@humanconnectome.org>, "tbbr...@wustl.edu" <tbbr...@wustl.edu> Subject: Re: [HCP-Users] Minimal Preprocessing Pipeline (on my own data): Flipped EPI (AP) Also, a low resolution T2w image is problematic (minimum tested is 1mm isotropic and we know some things don’t work well when it gets worse than 1mm though that is probably not the source of the current issue).   can you elaborate on what things may not work well? this is a different topic though. Note that we have started acquiring the higher res T2 from the time when I decided to use the HCP pipelines. It is possible that because this data doesn’t meet the minimum acquisition requirements (at least 1mm isotropic T1w and T2w, field map, fMRI—if using MB then saving SBRef) you’ll have to hack the pipelines somehow to get them to work more robustly.  You’ll need to look into what is happening in the EPI to T1w registration, focusing on the files output by this module: Point taken for the SBRef. Our MB images look fairly good (eyeballing them, and compared to HCP unprocessed data) so I thought it'd be ok without the SBRef. Going forward we'll definitely include the SBRef acquisition in our data. https://github.com/Washington-University/Pipelines/blob/master/fMRIVolume/scripts/DistortionCorrectionAndEPIToT1wReg_FLIRTBBRAndFreeSurferBBRbased.sh Especially focusing on line 255 (check that the inputs to epi_reg look okay and the outputs).  epi_reg is an FSL script, but you might have to either hack it or improve its inputs to get it to work.  These days, we recommend using spin echo phase reversed images that match your fMRI acquisition’s geometry and echo spacing to compute field maps (they are faster to acquire) and you don’t have to rely as much on epi_reg to handle everything in a black box.   The output of epi_reg is indeed flipped. I'll dig into it and see how I can fix that.  Of course, we definitely recommend users ask about their MRI protocols on the HCP Users list before acquiring them, so we can prevent these kind of issues ahead of time.   Thanks for this recommendation. This data was acquired prior to the release of the HCP preprocessing pipelines. I realize that not all of the optimizations that you have implemented will work well with this data, however I do like the innovation of the CIFTI format (in most cases, people who work with surface data completely ignore subcortical structures, and I want to work with both) and running the HCP preprocessing pipelines seemed like the easiest way to go to end up with that format.  Thanks for all the good work and responsiveness. - Julien   The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail. _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users