Hi folks, I have a technical question about processing data using the HCP 3.5 pipeline scripts. We recently set up an HCP MRI data collection protocol for one of my new R01s. I just learned over the weekend that - although we did a very careful job ensuring that all the sequences being used for the primary experimentation were correctly configured - we made one mistake. We added a resting state scan after the rest of it was set up and mistakenly pulled the wrong EPI sequence for it. As a result, the voxel sizes and phase encoding for resting state do not match the spin echo sequences we were planning to use for TOPUP unwarping of our other EPI data. Specifically, the spin echo sequences are L/R and R/L encoding at 2.1 mm isotropic resolution, while the EPI resting state data was collected using A/P at 3.0 mm isotropic.
So, we can process the primary fMRI task data just fine. But, will it still be possible to process the resting state data using the HCP scripts and approach? We have to use TOPUP for distortion correction as we did not collect a gradient echo fieldmap scan. I had understood that TOPUP didn't really care that much as long as you could create a valid fieldmap. However, I've been primarily an SPM user for my MRI research until we started using the HCP approach a year ago. So my FSL knowledge remains comparatively limited and I just don't know. When I run the HCP scripts to try this out, the primary thing that seems to go wrong is it throws an error in from the TopupPreprocessingAll.sh script (line 251) when it tries to multiply the Scout by the Jacobian.nii.gz b/c they're different voxel sizes. Otherwise it runs through to the end (we've disabled Jacobian modulation option (so the OneStepResampling script never encountered the same problem). However, visual inspection of the results for one sample dataset doesn't look like the later processing stages got things exactly right... There's some odd distortions down near the OFC where you'd expect the most amount of unwarping to be done. I got to wondering whether there was a simple L/R vs A/P incompatibility, i.e, that this won't work at all. Or maybe more complicated... if the different voxel sizes between the fieldmap and EPI might've put the origins of the images at different places and ended up confusing the OneStepResampling step. Or something... I'm testing this on all these datasets now to see whether the results across multiple subjects look visually OK or not. But as I do this, it'd be great if someone could tell me a simple "Yes" or "No" if I'm wasting my time with this. Even if this is possible but it requires some modifications/recoding to the HCP scripts (e.g., to resample the fieldmaps to 3x3x3 resolution after they're calculated??). I'm happy to do the latter if someone lays out for me what changes might needed. Thanks, Mike Michael C. Stevens, Ph.D. Director, Clinical Neuroscience & Development Laboratory Director, Child & Adolescent Research, The Institute of Living Associate Professor of Psychiatry (Adjunct), Yale University School of Medicine 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 HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users
