I am using fMRIPrep on a pediatric dataset. I have decided on using the MNIPediatricAsym:cohort-4 template for spatial normalization. I was interested in using ICA-AROMA (specifically non-aggressive), but I am noticing that it is only supported in MNI152NLin6Asym space. I saw a prior post from a few years back indicating that only MNI152 is supported. Just want to make sure this is still currently the case. I am wondering if there are other recommendations for denoising based on MNIPediatricAsym space. I do see that using the CompCor components are an option, but it would be great to know if there are also new, robust ways to deal with denoising especially for a pediatric dataset.
I am also noticing that antsBrainExtraction uses its own specific templates (ANTs/ANTsR Brain Templates). I was looking at the OASIS and NKI templates and I noticed that the NKI template is based on children <=10 years of age. I just want to make sure that I’m not missing out on other appropriate templates for antsBrainExtraction. If you have any other recommendations for an age range of 8-12 I would love to know.
There are also a few questions I have on resolution and DOF.
When using the MNIPediatricAsym:cohort-4 template, I noticed that there are two resolutions available (1mm and 2mm). My T1w is 1mm isotropic and my BOLD data is 2.2mm isotropic. I am wondering about the pros and cons behind choosing 1mm or 2mm for the template.
Lastly, I want to make sure of the --bold2t1w-dof option. Is this a specification of DOF when registering the bold image to the T1w image in native space? Also want to confirm that 6 DOF is standard practice.
The space you are in does not have much impact on what your denoising strategy should be. Your denoising strategy should be based on the quality of your data, your tolerance for tDOF loss, and the questions you want to answer.
Not worth going up to 1mm isotropic in my opinion. There is only so much signal in your images, sampling up to 1mm will basically just make your images take up 8X more space for little-to-no benefit (and will also make post-processing steps be more computationally intensive).
I believe so, and yes 6 DOF (just rotation and translation, no scaling/shearing) is the default.
Thank you so much for your help, @Steven! You are a godsend.
I just want to confirm whether the documentation mentioning that ‘non-aggressive denoising can be manually performed in the T1w space’ is simply pointing out one example of many possible spaces. So, I should be able to use the AROMAnoiseICs.csv and MELODIC_mixing.tsv files in my fmriprep func output directory to denoise the preprocessed bold data (sub-007_task-rest_space-MNIPediatricAsym_cohort-4_res-2_desc-preproc_bold.nii.gz).