Hi Erdem,
regarding reliability and accuracy of spm12's segmentation with regard
to total intracranial volume, you may also want to look at Malone et
al., 2015 (http://www.ncbi.nlm.nih.gov/pubmed/?term=25255942).
Cheers,
Marko
Erdem Pulcu Ph.D. wrote:
> Dear Guillaume,
>
> Actually, this is another issue that I am not too confident with. I
> initially started off this analysis with SPM8 and back then I was using
> get_volumes to calculate intracranial volume. After communicating with
> John Ashburner I have updated to SPM 12 and started this analysis from
> scratch and use tissue volumes from the GUI. But comparing the
> intracranial volume from SPM8 and SPM 12, although they are highly
> correlated (r =. 99), but consistently SPM 12 gives lower volumes around
> 100-120 mL for each subject. Do you think there might be a special
> reason for this? I am not too worried about the reliability of the
> results because of the high correlation, but I still wanted to ask
>
> Please let me know what you think,
>
> All the best,
>
> Erdem
>
> > Date: Thu, 21 May 2015 20:01:55 +0100
> > From: [log in to unmask]
> > To: [log in to unmask]
> > CC: [log in to unmask]
> > Subject: Re: [SPM] VBM12, SnPM Nonstationary cluster issues
> >
> > Dear Erdem,
> >
> > There is now an option in SPM12's batch to compute total tissue volumes
> > from segmentation results:
> > SPM > Util > Tissue Volumes.
> >
> > Best regards,
> > Guillaume.
> >
> >
> > On 21/05/15 16:18, H. Nebl wrote:
> > > Dear Erdem,
> > >
> > > See below for additional comments.
> > >
> > > 1) I don't think there has been any change for the non-stationary
> cluster correction between e.g. SPM12 and SPM8. If you want to use this
> correction you do not have to set up any new models, and there's no need
> to turn to SnPM, it's just modifying spm_defaults and loading the
> already estimated models via "Results", the results tables will reflect
> non-stat. corrected cluster p values then (which is not explicitely
> stated within the table though). Non-stat. cluster correction is usually
> considered a very good idea for VBM.
> > >
> > > 2) SnPM13 should be compatible with SPM12 (at least it was with the
> beta version SPM12b) and can be obtained via registering at
> http://www2.warwick.ac.uk/fac/sci/statistics/staff/academic-research/nichols/software/snpm/snpmreg
> .
> > >
> > > 3) There's no need to calculate anything manually if you go with
> the implemented correction, see 1)
> > >
> > > 4) No, just go with the correction.
> > >
> > > 5) When going with SnPM VBM data could be treated like an fMRI data
> set, instead of beta/con images from single-subject level you would
> select the s(m)wc1 files. Likely there are a few aspects that need some
> adjustments like masking, covariates (e.g. TIV) according to specific
> needs of VBM analyses, but the SnPM GUI should be similar to that within
> SPM.
> > >
> > > 6) This MarsBaR version
> http://sourceforge.net/projects/marsbar/files/marsbar/0.44/ should be
> compatible with SPM12. You can indeed use MarsBaR to extract GM values
> from a second-level VBM model, similar to extracting (cluster-averaged)
> beta/con estimates from a second-level fMRI model. There should be no
> difference in methodology between different toolboxes / scripts, it's
> really just computing the average across voxels. Of course some of them
> might be easier to handle, e.g. MarsBaR has a GUI, whereas get_totals.m
> is a simple m file without any further requirements (see
> http://www.nemotos.net/?p=292 for instructions). Just make sure that the
> dimensions and voxel resolution of the ROI mask files match those of the
> s(m)wc1 files, and check what the extracted values mean (with MarsBaR
> you get the average, if you want to convert it into volume [mm^3] you
> have to multiply by number of voxels and voxel size in mm^3).
> > >
> > >> any additional comments
> > > I wouldn't use the term "VBM12" (which the subject implies) except
> you really rely on the preliminary VBM12 version. If you did not, then
> your preprocessing is based on the improved unified segmentation
> algorithm with 6 tissue probability maps ("Segment"), the "old" unified
> segmentation algorithm with 3 tissue probability maps ("Old Segment"),
> or the SPM12-compatible version of the VBM8 toolbox. You should make
> sure about that in a paper as the algorithms differ.
> > >
> > > Best
> > >
> > > Helmut
> > >
> >
> > --
> > Guillaume Flandin, PhD
> > Wellcome Trust Centre for Neuroimaging
> > University College London
> > 12 Queen Square
> > London WC1N 3BG
--
____________________________________________________
PD Dr. med. Marko Wilke
Facharzt für Kinder- und Jugendmedizin
Leiter, Experimentelle Pädiatrische Neurobildgebung
Universitäts-Kinderklinik
Abt. III (Neuropädiatrie)
Marko Wilke, MD, PhD
Pediatrician
Head, Experimental Pediatric Neuroimaging
University Children's Hospital
Dept. III (Pediatric Neurology)
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/
____________________________________________________
|