Hi Alfonso:

I too enjoyed your recent course in Cincinnati.  I am running an analysis of children age 3-5 years, and hope to use a pediatric template in the CONN preprocessing pipeline.  Can you please advise on how to do this and/or if you think it is necessary? 

Thanks and cheers,
John

Hi John,

I do not have enough experience with pediatric samples so I am going to ask around to get a better answer (and other CONN users please feel free to chime in). My impression is that for this age-range the best choices are to either use an age-appropriate pediatric template or to use a group-specific template (e.g. using DARTEL or freesurfer), and that for any of these choices (unless you have a EPI-template) you would want to use "indirect normalization" (i.e. normalize your structural volumes and apply the same transformation to your already-coregistered functional data). 

Regarding the specifics of how to do this in CONN, if you would like to use your own age-appropriate template (I am going to assume here that you have a T1-template, but let me know if your template is defined differently), the simplest way to do this would be by using SPM8 instead of SPM12 (this is because the default normalization in SPM12 uses tissue probability maps instead of a T1 template). After importing your data into CONN, simply type the following command in Matlab's command window to run all preprocessing steps:

 conn_setup_preproc({'functional_realign&unwarp', 'functional_slicetime', 'functional_coregister', 'structural_segment&normalize', 'functional_art', 'functional_smooth'}, 'applytofunctional', true, 'structural_template', '/template.img');

(changing './template.img' to the location of your structural template file). That will run all standard preprocessing steps including the normalization to your own template (it is also possible to do the same using the GUI or using SPM12 but both of these options would require a somewhat more involved procedure).

Hope this helps and let me know if your template is defined differently and/or if you would like me to elaborate on some of the alternative options
Alfonso

EDIT: I corrected the conn_setup_preproc line above (added a missing argument)

Originally posted by John Hutton:

Thank you very much, Alfonso.  The good news, perhaps, is that I actually have gray/white/CSF pediatric tissue probability maps, as well as a pediatric T1 template.  Can I use these maps in CONN/SPM12, so that I don't need to switch to SPM8?  

Thank you and apologies for the novice-level questions!

Best,
John

Hi John,

Yes, definitely. There are only a couple of subtleties to consider when using alternative tissue probability maps in SPM12:

a) in SPM12 the tissue probability maps should be entered as a single 4d-file and explicitly add-up to one (in SPM8 the "new segment" procedure also accepted tissue probabiity maps but they were defined there only for grey/white/csf areas and SPM would implicitly consider a fourth tissue class corresponding to any other tissue type), so if you have grey/white/CSF tissue probability maps and you want to use these in SPM12 you should create an additional "other" probability map with probability values 1-p1-p2-p3 (assuming your original files are scaled between 0 to 1 representing tissue-class probabilities) and merge all four classes into a single map. You can do this, for example, using the following commands:

 spm_imcalc({'grey.nii','white.nii','csf.nii'},'other.nii','max(0,1-i1-i2-i3)');
 spm_file_merge({'grey.nii','white.nii','csf.nii','other.nii'},'TPM.nii');

 b) when using segmentation with non-default tissue probability maps you need to specify the number of gaussians associated with each tissue class. The recommended values for a grey/white/CSF template are [2 2 2 4] (respectively for grey/white/CSF/other) but you may want to play with these values if you are using a new template to find what works best

In terms of how to use these alternative tissue probability maps in CONN, you could use the 'interactive SPM batch editor' option in 'structural tools' and 'functional tools' to manually edit the default values in the normalization/segmentation steps but that is probably a bit cumbersome, so I am attaching a patch (this patch is for version 15e, copy it into the CONN distribution folder overwriting the file with the same name there) that allows you to use alternative tissue probability maps using a similar syntax as I was using in the previous-post examples. In particular, if you want to use the indirect normalization pipeline you could load your raw functional/structural data in CONN and then simply run:

conn_setup_preproc({'functional_realign&unwarp', 'functional_slicetime', 'functional_coregister', 'structural_segment&normalize', 'functional_art', 'functional_smooth'}, 'applytofunctional', true, 'tpm_template', 'TPM.nii', 'tpm_ngaus', [2 2 2 4]);

This will run all preprocessing steps (including the normalization using the new tissue probability maps). If you prefer to use the direct normalization pipeline simply run instead:

conn_setup_preproc({'functional_realign&unwarp', 'functional_slicetime', 'structural_segment&normalize', 'functional_normalize', 'functional_art', 'functional_smooth'}, 'tpm_template', 'TPM.nii', 'tpm_ngaus', [2 2 2 4]);

Let me know if you run into any issues and/or if you would like me to clarify any of the above. 

Hope this helps
Alfonso
Originally posted by John Hutton:

Dear Alfonso and colleagues,

Sorry to post in such an old topic, but I have an additional question regarding pediatric templates so I figured I might as well ask it here. It may help someone in the future too. 

So, we're also analyzing data from children (7-9 years), and we want to combine VBM and rsfMRI. To do this, we would like to make sure structural and functional data are all in the same space, more specifically, normalized to DARTEL template that we have created using age-appropriate TPM's from the template-o-matic toolbox. In your first response to this topic, you mentioned that a DARTEL template would be a good approach for normalization in infants, so I'm trying to implement it in conn toolbox. 

From your previous responses to John, I get that SPM8 requires a T1 template, and SPM12 a TPM. How can we go about doing indirect normalization to a DARTEL template? 

(hope this is not a trivial question, I'm not much of a SPM user). 

Best regards
Paulo

Hi Alfonso and Conn Experts,

Concerning the command for the use of a pediatric template:  conn_setup_preproc({'functional_realign&unwarp', 'functional_slicetime', 'functional_coregister', 'structural_segment&normalize', 'functional_art', 'functional_smooth'}, 'applytofunctional', true, 'structural_template', '/template.img');

I have difficulties figuring out how to perform these preprocessing steps for only a subset of subjects. Should i create an array - subset of subjects (ex: nsub = {'Subject 1', 'Subject 3', 'Subject 10'}) and then stating something like this:

conn_setup_preproc({'functional_realign&unwarp', 'functional_slicetime', 'functional_coregister', 'structural_segment&normalize', 'functional_art', 'functional_smooth'}, 'applytofunctional', true, 'structural_template', '/template.img', 'subjects', nsub) ?

Cordially,

Jules