There seems to be some confusion on the use of the SVC in the pickatlas,
so here goes:
The pickatlas automatically implements the SPM small volume correction
(SVC). The SVC does not correct for multiple comparisons. It just
reduces the number of voxels/resels that go into a multiple comparisons
correction. Instead of 10,000 voxels going into the computation, only
the number of voxels in the ROI mask go into the computation (e.g. 350
voxels, thereby improving the resulting P-values in a multiple
comparisons computation). The FWE and FDR options can still be used
with the pickatlas. If you choose not to use either correction with the
pickatlas (FWE or FDR)then you will be looking at results that have not
been corrected for multiple comparisons - you may be perfectly happy
with this uncorrected P at a sufficiently stringent alpha cutoff.
Recognize, however, that even with a pickatlas ROI mask, you will still
have a lot of voxels going into the results. If no correction is
applied when using the pickatlas, the P-values displayed by SPM will be
exactly the same as looking at the data without having applied any
pickatlas ROI mask.
When using the pickatlas there is no need to additionally apply the
SPM SVC, as this is automatically applied by the pickatlas.
Regarding the generation of the subtraction mask:
Using Pickatlas version 2.1, click on advanced mode, then select in
this order:
Frontal lobe, then Brodmann areas 9, 44, 45, 46, 47
Highlight all of them in the right advanced pane, then click on
Subtract. Click on commit, then Done.
More complex combinations can be created by performing multiple
unions/subtractions/intersections of subregions and user defined
spheres.
Hope this helps.
------------------------------------------------------------------------
----------
Joseph A. Maldjian, MD
Professor, Dept. of Radiology and Biomedical Engineering
Director Advanced Neuroscience Imaging Research Core (ANSIR)
Wake Forest University School of Medicine
Office: 336 716-2815
fax: 336 716-2870
email: [log in to unmask]
website: www.ansir.wfubmc.edu/maldjian.htm
------------------------------------------------------------------------
----------
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
On Behalf Of Kerrie Tainter
Sent: Friday, August 04, 2006 12:55 PM
To: [log in to unmask]
Subject: Re: [SPM] wfu_pickatlas & FWE
this is just the kind of newbie question that might go unanswered (:*}),
so even though I am a newbie myself, I will try to help. Any of you
gurus, please correct my mistakes.
Kerrie
>From: Ryan Dobbs <[log in to unmask]>
>Reply-To: Ryan Dobbs <[log in to unmask]>
>To: [log in to unmask]
>Subject: [SPM] wfu_pickatlas & FWE
>Date: Fri, 4 Aug 2006 16:37:13 +0100
>
>Yesterday I was informed not to use FWE when using wfu_pickatlas
>becasue it already performs a SVC. However, I had a couple questions
>for
>clarification:
>
>1. Can someone explain the rationale behind not using FWE and SVC.
>I'm pretty sure I understand why, but my boss said "that's how we've
>always done it," and is giving me a hard time about the info I was
given.
FWE corrects the p-values for situations where large numbers of
comparisons are made, as is the case when every voxel in a large image
volume is tested for significance. In the case where you are sampling a
subregion, the number of comparisons are fewer (by orders of magnitude)
and the correction applied by FWE is overly strict. the SVC (small
volume correction) is a correction to estimates that can be applied in
the case of sampling subregions, and is more appropriate to the WFU
Pickatlas situation.
For a better explanation of why you need to make corrections in the
first place,
you'll need to do some reading . Try the book chapters on the SPM web
site.
>
>2. If i don't use FWE with wfu_pickatlas, do i choose none @ .001?
>
I believe this is correct, any comments from others?
>3. How do I create a mask using wfu_pickatlas? I've just been doing
>it manually, and no one here seems to be able to help me. I want one
>for [frontal - 9_44_45_46_47], 21_22_39, 44_45_47, 9_46,
9_44_45_46_47, etc.
>I was told to use the toolbox to create one, but I have no idea what
>i'm doing. Is this explained in the user's guide?
>
Lots of this is explained in the users guide, but basically you select
regions (that may be predefined in the pickatlas) to include in the mask
that you are building, and this mask is then used to sample the image.
You can concatenate multiple regions easily.
I am more interested in creating my own custom atlases/masks, and I
noted in the manual that all atlases are created from ANA7.5 image
formats, so conceivabling a segmentation program could be used to
created very specialized masks, which would be incorporated into a
custom atlas.
Hope this helps.
>Thanks
>Ryan
|