Obtaining age-specific tissue priors

[dsodums]

We are interested in using age specific segmentation with the Rockland dataset as suggested on the wiki since it does not have T2 FLAIR images. I understand we need to provide subject ages, but I'm a bit at a loss as to how we provide the tissue priors. Would we run this separately and create our own study averages, i.e. taking a representative sample from the dataset for each decile, segmenting using FAST, and then averaging the outputs to be used as priors? Or would we obtain these files from a publicly available dataset (i.e. MNI template but for age deciles)?

Any suggestions would be greatly appreciated!

[justinliuyuwang]

Hello!

We actually provide tissue priors that you can use, in the Cam-CAN branch. The templates can be found in templates/age-specific and the CSV files used to refer the pipeline to the appropriate age-specific priors can be found in bb_structural_pipeline/FAST_age_spec. The init_vars references to these files are on Lines 102 and 103 in init_vars. Lastly, the code that runs the age specific segmentation can be found on Lines 101 to 158 in /bb_structural_pipeline/bb_struct_init.

Unfortunately, the main branch does not yet have a toggle for age-specific segmentation so you will have to copy the above-mentioned features from the Cam-CAN branch over to your current working branch.*

• EDIT: Currently, the preferred course of action for using age-specific tissue priors is not to copy the above mentioned features from the Cam-CAN branch, but to use the Cam-CAN branch instead of the main branch. Details (including turning off fieldmap processing) discussed by @dsodums here.

[dsodums]

Thank you!

I took a stab at copying over the various functions but ran into an issue where the labelled_GM.nii file for some reason has an extra "_" in it:

labelled_GM_.nii.gz

This has downstream issues for the FC generation, any suggestion on how I can fix this?

I took the following steps to copy functions from Cam-CAN branch to the main one. If its better to start from scratch by using the Cam-CAN branch and I'm open to that as well...

1. Copied templates/age-specific folder to main branch templates/
2. Copied bb_structural_pipeline/FAST_age_spec into main branch bb_structural_pipeline/
3. Added Lines 102 and 103 in init_vars
4. Replaced /bb_structural_pipeline/bb_stuct_init in main branch with /bb_structural_pipeline/bb_struct_init from Cam-CAN branch
5. Copied rename_pve.py from Cam-CAN branch /bb_structural_pipeline to main branch
6. Copied tvb_struct_parcellation.sh from Cam-CAN branch /bb_structural_pipeline to main branch

[dsodums]

Hi again,

I thought it might be easier to use the actual Cam-CAN branch rather than copying functions from it to the main one. I'm noticing though that since the Cam-CAN dataset has fieldmaps, and Rockland doesn't, that its causing problems.

Would you recommend I continue working from the main branch, or edit the Cam-CAN one? If I should keep using the main branch, any suggestions on how to fix the issue where its adding the "_" to the labelled_GM.nii.gz file?

Thanks again for all the help!

[justinliuyuwang]

Yes! If it is only FEAT that is failing, then I believe:

echo "set fmri(regunwarp_yn) 1" >> $1/fMRI/rfMRI_$2.fsf in line 48 of Cam-CAN's tvb-ukbb/bb_functional_pipeline/bb_prepare_rfMRI

should be changed to:

echo "set fmri(regunwarp_yn) 0" >> $1/fMRI/rfMRI_$2.fsf

[justinliuyuwang]

Please try that, and in the meantime, please don't discard the changes you've made to your copy of the main branch. In case the Cam-CAN branch doesn't fit your dataset well, I'll see how we can change your copy of the main branch to accept age-specific templates.

[dsodums]

Amazing, thanks for the quick help! I'll try that and will let you know how it goes. Sounds good for keeping the changes to the main branch :).

[dsodums]

OK that worked! Thanks again for the help, much appreciated.

[justinliuyuwang]

Fantastic! No problem at all :)