Fa_template resolution

My DWI data has 2mm isotropic resolution but the fa_template created by “dwiintensitynorm” has 2.2x2.3x2.3mm resolution.
Why this happens and how can I modify it?


This is caused by the way we calculate the unbiased voxel-grid in case of rotated input images. You can find a discussion here. population_template has a -voxel_size option to modify this behaviour but this option is not accessible from dwiintensitynorm.

If you’re after an FA template with specific resolution, I’d convert the DWI images to FA images and create a population template using population_template.

1 Like

Thanks, Max.
I’d like to know more about how “population_template” creates the template. Like what are the steps, what types of registrations are used … . Is there a document that explains those details?

Not a document but population_template prints an overview of what it is going to perform as info messages to the terminal at the beginning of the processing. After initial alignment, the template is iteratively refined (“stages”) by registering each subject to the template, and averaging the transformed subject images to form a new template.

Here is an overview of the registration stages performed by default, for more options see the docs:

Thanks for your response, Max.

1- Would you please elaborate what does initial alignment do?
2- Is it possible to introduce an initial template to start with it? like with FSL fnirt and fnirt commands.
3- After creating the fa_template using the “population_template”, I’d like to register the individual T2 images to that template. What’s the best way to do that?

I’d like to know if someone has an answer to my third question here?

3- After creating the fa_template using the “population_template”, I’d like to register the individual T2 images to that template. What’s the best way to do that?

As always, I encourage distinction between registration and transformation.

If you have FA and T2-weighted images aligned for individual subjects, and you perform registration of subject FA images to a template (and either preserve the transformations estimated during the template generation procedure, or post-hoc explicitly register each subject’s FA image to the template FA image), then you can apply that same transformation to the subject’s T2-weighted data, and it will undergo the same spatial transformation as that estimated to be optimal for moving the subject’s anatomy into the template space.

Performing registration between a subject image of one modality and a template image of another modality may be possible, but is generally best avoided. It is better IMO to do that inter-modality registration in the context of images corresponding to an individual subject, as that way the transformation is rigid only and therefore the complexity involved in measuring spatial correspondence between images with different contrasts is easier to manage.

Hi Max,

Would you be able to tell me how the first step of template building works? what is exactly the initial alignment? i.e. alignment to what?

Thank you,

The -initial_alignment option in population_template performs translation of all images to a common central point calculated using the image intensity (“mass”) or the image centre (“geometric”). After that, the images are averaged which forms the initial template to which each image is registered before updating the template.

1 Like