Does anyone happen to have any HCP-esque (highly accelerated, <1 second TR) single-echo fMRI data with phase reconstruction enabled that they’d be willing to share openly?
For context, I’m trying to curate some complex-valued fMRI test datasets for fMRIPost-Phase, and the datasets I’ve been able to find are all either multi-echo or 7T.
We do have a lot of complex reconstructed data in cneuromod (not all sub-datasets), but the TR is 1.49s for the single-echo, we also have complex multi-echo (not released yet).
For example mario/sub-01/ses-001/func at main · courtois-neuromod/mario · GitHub .
3 (soon 4) of our participants asked to release their data with CC0 on CONP, while the rest is unfortunately behind a DTA.
Out of curiosity, what do you plan to do with preprocessed phase (or complex) data?
That’s awesome, thanks!
I’m mostly going to try to use these data for testing fMRIPost-Phase to make sure it works well across common protocols for complex-valued BOLD. The main steps I’ll want to test with single-echo data are NORDIC, highcor confound extraction, complex ICA with GIFT, phase regression, and phase jolt estimation.
That’s great, looking forward to that package!
Unless I am mistaken, it seems to me that NORDIC (which is the main reason we collected phase data), should be applied before preprocessing, as interpolation would smooth the noise it is trying to model. I was planning to package it as a BIDS-app, for cneuromod and for other upcoming projects
I am planning to add it to fMRIPrep (see Add optional NORDIC denoising step · Issue #3308 · nipreps/fmriprep · GitHub), maybe as a subworkflow saved in niworkflows. However, with fMRIPost workflows, the two main use-cases involve either using preprocessed data from fMRIPrep or using a combination of transforms from fMRIPrep and the raw data from the BIDS dataset (e.g., if you run fMRIPrep in minimal mode).
Phase processing will require access to the raw data no matter what, since fMRIPrep is probably never going to natively process phase data, so fMRIPost-Phase needs to basically re-do any initial processing before the motion correction, SDC, etc. transforms. In this case that would be NORDIC + slice timing correction.
BTW the NORDIC licensing is pretty restrictive, so I’m hoping to get in touch with the dev (Steen Moeller) about re-licensing at some point.
MANAGED BY INCF