Dear Aggie,
personally I use a much short pipeline for pediatric data and in short the largest difference is that I use the pre-estimated TOM8 data from the NIH sample based on about 400 subjects between 5-18 years. Because of this large number of subjects the estimates are more stable and reliable than estimates from smaller own pediatric samples. I would suggest to estimate the regression parameters for your TOM8 template only if you have enough subjects (I would rather use > 100 subjects for template creation here) and/or age of the subjects is < 5 years. Anyway, this is my suggestion:
1. Create TOM8 template using age/gender from your sample
2. VBM8 Estimate & Write | Writing options | Grey/While matter | DARTEL export | affine
3. Run DARTEl | Create template
Now you can continue with your steps from 6, but without MNI registration because your template is roughly registered to MNI due to use of affine registration (which includes scaling in contrast to rigid registration that is used be DARTEl default). Furthermore, estimation of segmentation can be skipped, because you already did this before.
4. VBM8 “write already estimated segmentations”
a. Input TOM template for TPM.nii file and DARTEL template generate in previous step (Template_1.nii,1)
b. Writing options: Grey/While matter | modulated normalized | non-linear only
Maybe Florian Kurth can add something more, because he created a more complicated pipeline for asymmetry analysis and has more practical experience.
Best,
Christian
On Fri, 7 Dec 2012 21:08:32 +0000, Aggie McMahon <[log in to unmask]> wrote:
>Hi all!
>
>I have a few questions about what I’ve done in my VBM8/DARTEL analysis of a pediatric sample. I outlined below what I did and the questions I had along the way. If there aren’t too many mistakes in this pipeline, I believe it would serve well as a step-by-step reference after some editing. I put ***'s by my questions because I couldn't bold things in the plain text format of this message. I’d love any feedback, thanks SPMers!!!
>
>Pipeline for using DARTEL with a pediatric sample:
>
>1) Segment data using “New Segmentation” tool in SPM8
>2) Create template using the TOM8 toolbox
> a. “Estimate regression parameters”
> i. GM, WM, CSF, & T1 images inputted from previous step
> ii. Enter age and gender vectors
> iii. ***What are BKG1,2,&3?
> b. “Create new template”
> i. Using TOM.mat file generated in previous step
> ii. ***Low-dimensional normalization?
>3) “New segmentation” using TPM_age.nii,1-6 generated in previous step
>4) “Initial Import” module in DARTEL
> a. Input *_sn.mat parameter files
> b. ***Does new segmentation generate these files as well? I’m only asking because I ran the “Segmention” and “New Segment” options together so I don’t know how to distinguish which came from which. If New Segmentation doesn’t generate them than this means I am more confused than I had initially anticipated.
>5) Run DARTEL (create Template) – input images from previous step, rc1* and rc2* in each their own “Images” channel
>6) VBM8 “estimate and write”
> a. Input TOM template for TPM.nii file and DARTEL template generate in previous step (Template_1.nii,1 ***Was I supposed to select Template_6?)
> b. ***Since the data has already been segmented using the TOM template, could we avoid this step by selecting “ Write already estimated segmentations”?
> c. Writing options:
> i. Under GM&WM
> ii. DARTEL export – affine option
> d. CSF: Modulated normalized – Non-linear only; DARTEL export - affine
>7) Normalize to MNI space – I understand this is a debatable step, especially since we’re going through the trouble to have a custom template, but we need some frame of reference
> a. John Ashburner’s script http://brainmap.wisc.edu/assets/normalize_DARTEL.m
> b. ***Would the use of the DARTEL tool “Normalise to MNI space” be for altering the template prior to running VBM?
>8) Group statistical analysis
>
>Unless otherwise noted, I used the default options. This is the first time I’ve used SPM so I’m concerned there are “unknown unknowns” about what I could be doing wrong. My biggest hesitation with this is I have a small sample for the template (control n=21). I wasn’t sure if the TOM toolbox, by using the external imaging data from the NIH study, eliminates the need for a large sample size to create a custom template. If I would need at least 50, does anyone else know of an appropriate template that works with VBM8? My cohort is pretty specific (adolescent, females only).
>
>Thank you,
>Aggie
|