fmriprep extract time series for adolescents (MNIpediatric)

, , ,
1

Hi all,

I’m a trainee & fmri novice so I’d really appreciate any assistance!
My goal is to extract time series from an adolescent task-based scan for network analysis in R.

For fmriprep, I see that there is a pediatric MNI atlas I can use as the output space. However, I’m not sure how well this will work with other tools.

  1. First, I was going to --use-syn-sdc for fieldmap-less distortion correction, but I believe I read this functions in a standard MNI atlas. If so, would this introduce some error?

  2. Also, I’ve seen a previous suggestion to follow this tutorial to extract time series. However, I do not see the MNIpediatric atlas within nilearn’s database, only the standard MNI152. So, I’m not sure how well this approach would work with my adolescent dataset if I use the MNI-pediatric atlas in fmriprep.

  • A. Alternatively, I’ve had thoughts of using each subject’s T1 space as the output space, but I can’t seem to find a workflow I could follow to extract time series downstream if I did that.

  • B. Another thought I’ve had was using the fmridenoise tool and then extracting time series, but I’ve been warned about potentially re-introducing artifacts (https://onlinelibrary.wiley.com/doi/epdf/10.1002/hbm.24528) if I use that approach.

Thank you for any advice!

2
  1. No, using syn-sdc will not cause an error if your output space is different. The SDC calculation is performed in the standard MNI space, but the resulting fieldmaps are warped to whatever output space you specify.
  2. Nifti files are themselves timeseries. They are four dimensional images (x,y,z,time). This is true for any space (native space or standardized). To get a voxel’s timeseries you can just load the image using nilearn. If you need to extract them in specific regions, like in the tutorial, that is a different story, since these regions will have to be in the space of your outputs. Someone else can probably help you better in regards to adapting the pediatric template, but I would see how MNI152 would work for your data. In any scenario, you will need to clean your data. Nilearn has functionality to do that. fmridenoise is still in beta. You might also find this helpful https://github.com/SIMEXP/load_confounds.
1 Like
3

Thanks so much for the info & resources to look into

  1. Makes sense, thanks!

  2. I am trying to extract series for specific ROIs. If I were to use MNI-Pediatric in fmriprep, and then use MNI152 in nilearn for cleaning and ROI extraction, would the data just be resampled across the spaces and good to use? Or, would it better to just stick with the default MNI space in fmriprep in the first place?

4

There probably is a way to load in a non-standard parcellation like the MNI pediatric, but I don’t know best practices for doing so off the top of my head. You would want to use the same space in nilearn and fmriprep, so for now you can try just doing everything in MNI152.

1 Like