Creating Custom Reference Template

[Ethan Harvey]

Hello,

I am currently working with a proprietary neuroimaging dataset. I am following the preprocessing steps from the `t1-linear` pipeline and the registration step is not working as it should (the scans are coming out distorted). This seems to be because the scans in my dataset are much smaller than say for example OASIS-3 scans which are 256x256x179 (the average scan in my dataset is 256x256x20). Would you recommend I use ANTS to build my own template? One template for T1-weighted images and one template for T2-weighted images?

Kind regards,

Ethan Harvey

[ninon....@gmail.com]

Hello Ethan,

The solution will depend on your final objective. If your objective is to have all your T1w images co-registered in a somewhat standard space what you could do is resample the MNI152NLin2009cAsym template to a grid size consistent with that of your images (for example using ANTs) and then run t1-linear using this new template. You would need to do the same for your T2w images. Have in mind that if you want to crop the images at the end of the pipeline, you will need to adjust the cropping parameters. Otherwise just use the --uncropped_image option.

If you really need to be in the MNI standard space, then this would be more complicated. Let us know.

Best,

Ninon

[Ethan Harvey]

Hello Ninon,

Thanks for the quick response. To make sure, you are suggesting resampling the MNI152NLin2009cAsym template on the axial dimension from 193x229x193 to something like 193x229x20 so that the shape is consistent with that of my images. Then using this resampled template to register T1-weighted and T2-weighted images?

Kind regards,

Ethan Harvey

[ninon....@gmail.com]

Hello Ethan,

Yes, this is exactly that. If you have issues registering your T2w images, you might need to use a T2w template (and perform the resampling process).

Best,

Ninon