Hello again,
the reason behind using an absolute threshold is that you want to
exclude noise that is generated as a byproduct of segmentation and, even
more so, in segmented, modulated images. Nowadays, segmentation results
contain either pretty high or pretty low values (look at an image
histogram), but this is messed up somewhat by modulation and smoothing.
You therefore want to exclude those very low values as they are likely
spurious. I am not aware of papers that explicitly analyzed the effect
of excluding values below .2 or .1 but that does not mean they don't
exist; however, using .2 means that you, following segmentation,
normalization, potentially modulation, and smoothing, you exclude values
that have a probability of below 20% to be the tissue class you want to
analyze, and I do not see a big issue with that. Certainly not bigger
than using no threshold :)
Cheers,
Marko
R Duke wrote:
> Thank you Marko!
>
> The link you provided offered some nice information. Can you (or anyone
> else) recommend an article that might statistically justify why I would
> use 0.1 vs. 0.25 for my absolute threshold (I obviously get two
> completely different results when I run each value). I now understand
> the importance of using absolute thresholding, but I am unsure as to how
> I might statistically justify using a higher or lower value.
>
> Cheers,
> R Duke
>
> ------------------------------------------------------------------------
> *From:* Marko Wilke <[log in to unmask]>
> *To:* R Duke <[log in to unmask]>
> *Cc:* [log in to unmask]
> *Sent:* Tue, January 11, 2011 2:42:00 AM
> *Subject:* Re: [SPM] VBM8 VBM results off brain? Multiple Regression
>
> Hello,
>
> I would first blame it on the fact that you did not use threshold
> masking during statistical analysis. Entering either an absolute
> threshold (say, 0.2) or an explicit mask will preclude spurious
> activation (or at least, it should help :) Christian's pipeline also
> contains some hints for analysis options at
> http://dbm.neuro.uni-jena.de/vbm/statistical-analysis-for-spm5/
>
> Cheers,
> Marko
>
>
> R Duke wrote:
> > Hello SPM Experts,
> >
> > I am having an issue with my VBM results when using the VBM8 toolbox.
> Specifically, when I estimate and display my results I get a lot of
> clusters that are represented off the brain. This is obviously a
> concern, so I was wondering if someone could maybe identify where I
> might be going wrong? I have included my steps below.
> >
> > Cheers!
> >
> > SPM VBM8 Toolbox
> >
> > Segment Steps in VBM8 Toolbox
> >
> > Estimation Options
> > Tissue Probability Map - TPM
> > Gaussian per class - default
> > Bias Regularisation - medium (0.01)
> > Bias FWHM - default
> > Affine Regularisation - default
> > Warping Regularisation - default
> > Sampling Distance - default
> >
> > Extended Options
> > Low-dimensional
> > SANLM - Denoising
> > MRF - default
> > Clean up - light
> >
> > Writing options
> > GM, WM, CSF
> > Native - yes
> > Normalized - yes
> > Modulated normalized - non-linear only
> > Dartel - no
> >
> > Bias corrected, PVE label - yes for native and normalized
> >
> > Jacobian determinant - default
> > Deformation fields - default
> >
> > GM output files (m0wp101, m0wp102, m0wp103, etc) get smoothed.
> >
> > VBM Steps in VBM8 Toolbox
> >
> > Multiple Regression
> >
> > Select the smoothed files in my vbm analysis (sm0wp101, sm0wp102,
> sm0wp103, etc)
> >
> > Intercept - include intercept
> >
> > Enter my nuisance and covariates of interest
> >
> > Masking
> > Threshold masking - none
> > Implicit - I have used no and yes
> > Explicit - I have used no and yes (TPM - grey)
> >
> > Global calc and global normalization - default
> >
> > Estimate - spm.mat
> >
> > Results - spm.mat
> > Set and select my contrast
> > No masking
> > P-value adjustment - none (0.001)
> >
>
> --
> ____________________________________________________
> 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] <mailto:[log in to unmask]>
>
> http://www.medizin.uni-tuebingen.de/kinder/epn
> ____________________________________________________
>
--
____________________________________________________
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
____________________________________________________
|