Hi Stef,
thanks for your mail. Let's see:
> There are three sets of parameters online - derived from different
> strategies for spatial normalisation if I understand it correctly.
Yes.
> Given that our sample is quite homogeneous (age range: 62.04 - 90.48
> months; 53% female, all scanned using a 3T Siemens Trio), which
> mw_com_info.mat should we use?
I would suggest to use the "unified segmentation" dataset which seemed
like the best compromise overall. The "affine only" was used as a
reference point from earlier days.
> Does the choice affect later steps in the pre-processing?
In terms of obtained results, it really should otherwise I would have
wasted a lot of time ;) You could argue that the likely ensuing DARTEL
procedure will effectively supersede the initial choice, but that would
be up to you to find out.
> Further, the batch mode of mw_com_gen states, that one
> should make sure that the requested output is available with the
> selected data file. Could you please explain which data file is
> necessary to obtain a certain output?
I was not referring to a file, but to information, as in age, gender,
and field strength for each subject. This remark refers to you trying to
enter, e.g., subject age outside of the specified range, which would not
work.
> My layman understanding would be
> that it would only possible to get Dartel/Shoot template output if the
> mw_com_info.mat saved in the com_parameters_Dartel folder was used. Does
> that mean that TPM and T1 can only be generated using either
> com_parameters_affine/mw_com_info.mat or
> com_parameters_unified-segmentation/mw_com_info.mat as initial input?
Yes, you will likely want to generate a "standard" TPM (one 6 class TPM)
using the "unified segmentation" dataset and then a DARTEL TPM (6
iterations of GM and WM) using the DARTEL dataset described in the
second publication. These should then be entered into your processing
stream at the appropriate places.
> After reading both papers, I was wondering whether it is generally more
> recommended to use CAT12 for segmentation and pre-processing when
> dealing with unusual (i.e. 6-year-old children) populations compared to
> the implementations in SPM - especially when handling structural data. I
> am not sure which preprocessing pipeline is more appropriate in our
> case. I am weighing up two options (after creating both Dartel template
> and TPM using COM): Would it be better to use the SPM12 segmentation
> that - to my understanding - relies more on tissue priors given the fact
> that we will create an age-appropriate TPM in COM followed by Run Dartel
> (existing Templates) using the COM Dartel template? Or would it be
> better to use the CAT12 segmentation specifying sample-specific COM TPM
> and template to assist the spatial normalisation?
That, as John likes to state, depends on your definition of "better" and
can likely only be answered empirically. You (in my opinion) correctly
describe the problem, but what the answer is is hard to say.
> Would you recommend to
> always run an initial round of SPM segmentation to obtain the
> bias-corrected image and segment the obtained bias-corrected images
> afterwards as described in the referred publications?
No, that was only done in this case as I had data from different sources
and wanted to have equal starting estimates for the final processing
round. In a homogeneous sample, I would suggest to rather play with the
bias estimate options.
> Thanks for sharing your expertise with us!
For what it's worth ;)
Cheers
Marko
> On 18/05/2018, 16:08, "Marko Wilke" <[log in to unmask]>
> wrote:
>
> Hello Stef,
>
> as it seems no-one has answered your query, I will at least have a
> stab
>
> at the pediatric aspects of it:
>
> > 4. Last but not least, we would like to create a study specific
> sample
>
> > using the Dartel algorithm. How many structural scans are
>
> > recommended to use in order to run this procedure
> successfully? In
>
> > total, we have around 100 structural scans, but some images
> are only
>
> > rated as moderately good in terms of their quality. So when
> creating
>
> > the template, should we only use the high quality images or
> would it
>
> > be better to include all images that will be used in the
> analysis?
>
> > Related to that, if we use a study-specific template, will it
> still
>
> > be possible to perform ROI analysis? If so, what steps would be
>
> > necessary to do that and does that have any implications on the
>
> > pre-processing steps?
>
> One, you should always use all contributing subjects if you want to
>
> generate your own template. This has actually been discussed once or
>
> twice before on this list (which using the archive search function at
>
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?S1=spm may have revealed ;)
>
> Two, more images are usually better, but if you look at the
> simulations
>
> shown in Figures 5-8 in a recent paper
>
> (https://www.frontiersin.org/articles/10.3389/fncom.2017.00005/full) you
>
> will see that the quality of the generated tissue maps continues to
>
> increase way after reaching 100 subjects. Hence, in your case you may
>
> want to consider looking into alternative methods of generating a
>
> template matched to your pediatric population (as a hint, also try
>
> https://doi.org/10.1016/j.dib.2017.12.001).
>
> Three, using standard ROIs with a custom template is always a tricky
>
> question. The larger and more general the regions are, the less
> relevant
>
> this will be, but an issue it is. There is a number of things you can
>
> do: you could manually define your own regions of interest on your own
>
> template, you could match the ROI to your template (e.g., the original
>
> GM map forming the basis for the standard ROI to your final DARTEL GM
>
> template), you could use larger ROIs and assume that they will broadly
>
> match the regions you are interested in, or you could use functionally
>
> and/or diffusion-MRI-defined ROIs from your own population. Or you
> could
>
> come up with an approach I did not yet mention, of which certainly
> there
>
> are many. In other words, it will depend on the very question you
> have,
>
> on your data, and your requirements for spatial specificity.
>
> > Thank you for taking the time to consider this request. I am looking
>
> > forward to your thoughts on the issues mentioned.
>
> Not sure I have helped but I tried :)
>
> Cheers
>
> Marko
>
> --
>
> ____________________________________________________
>
> Prof. Dr. med. Marko Wilke
>
> Facharzt für Kinder- und Jugendmedizin
>
> Leiter, Experimentelle Pädiatrische Neurobildgebung
>
> Oberarzt der Abteilung Neuropädiatrie
>
> Universitäts-Kinderklinik
>
> Marko Wilke, MD, PhD
>
> Pediatrician
>
> Head, Experimental Pediatric Neuroimaging
>
> Consultant in Pediatric Neurology
>
> University Children's Hospital
>
> Hoppe-Seyler-Str. 1
>
> D - 72076 Tübingen, Germany
>
> Tel. +49 7071 29-83416
>
> Fax +49 7071 29-5473
>
> [log in to unmask]
>
> http://www.medizin.uni-tuebingen.de/kinder/epn/
>
> ____________________________________________________
>
--
____________________________________________________
Prof. Dr. med. Marko Wilke
Facharzt für Kinder- und Jugendmedizin
Leiter, Experimentelle Pädiatrische Neurobildgebung
Oberarzt der Abteilung Neuropädiatrie
Universitäts-Kinderklinik
Marko Wilke, MD, PhD
Pediatrician
Head, Experimental Pediatric Neuroimaging
Consultant in Pediatric Neurology
University Children's Hospital
Hoppe-Seyler-Str. 1
D - 72076 Tübingen, Germany
Tel. +49 7071 29-83416
Fax +49 7071 29-5473
[log in to unmask]
http://www.medizin.uni-tuebingen.de/kinder/epn/
____________________________________________________
|
