Dear Elena and SPMers:
This question may make me unpopular, since it addresses the
validity of using a SVC/mask with VBM, but I am going to ask it
anyways...
A number of groups have been applying SVC to VBM in their
published studies. I understand that, at least in SPM99,
non-stationariness of smoothing could make the SVC approach invalid
(see message below).
John, will you comment?
_______________________________________________________________
Date: Wed, 9 Jan 2002 13:04:17 GMT
Reply-To: Karl Friston <[log in to unmask]>
Sender: "SPM (Statistical Parametric Mapping)"
<[log in to unmask]>
From: Karl Friston <[log in to unmask]>
Subject: Re: svc in vbm
Comments: To: [log in to unmask]
Dear Christopher,
John is right in the sense that SPM99 uses a volume measure that
assumes the smoothness is stationary, The 'statistical' volume is
resel
(resolution elements) per voxel times the number of voxels. In SPM99
the average resels per voxel over the search volume is used If the
Small Volume in the SVC is small it may under-estimate or
over-estimate
the true statistical volume, if the local smoothnes is less than or
more than the average. This is not a problem unless there is
substantial nonstationariness in the smoothness. For VBM this may be
the case.
The way to fix this would be to use the resels per voxel estimate in
the small volume (from the RPV.img). However, SPM99 does not use a
local estimate. The simplest thing to do is to qualify your inference
by making the stationariness assumption explicit and say the inference
are only valid if the averge smoothness in the small volume is roughly
the same as over the entire search volume. This can only be assured if
the Gray matter partitions have been smoothed sufficiently (e.g.
8-12mm
FHWM).
Note that these comments apply to the p value based on height (not
spatial extent).
I hope this helps,
Karl
----- Begin Included Message -----
From [log in to unmask] Wed Jan 9 11:05:44 2002
References:
<[log in to unmask]>
Date: Wed, 9 Jan 2002 11:08:56 +0000
Reply-To: John Ashburner <[log in to unmask]>
From: John Ashburner <[log in to unmask]>
Organization: FIL
Subject: Re: svc in vbm
Comments: To: Christopher Summerfield
<[log in to unmask]>
To: [log in to unmask]
In-Reply-To:
<[log in to unmask]>
The extent statistic is definately a problem because of the
non-stationary smoothness
of the residuals. Smoother regions are likely to produce bigger
blobs. I would imagine
that this non-stationarity would also have negative consequences for
the validity of SVC
for VBM data as it has been implemented in SPM99. Less smooth regions
contain more
resolution elements than average, so more independent t-tests are done
and there is
more chance of getting a false positive result (in these less smooth
regions).
Perhaps someone else can comment.
Best regards,
-John
On Tuesday 08 January 2002 21:56, Christopher Summerfield wrote:
> hi spm
>
> I seem to remember reading somewhere in the list a mail which said
that
> results using a small volume correction in voxel based morphometry
may not
> be reliable - is this the case? if so why? can the procedure be
altered to
> permit svc?
>
> chris summerfield
> psychology
> columbia university
--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john
Robert K. McClure MD, Assistant Professor
Department of Psychiatry, School of Medicine
University of North Carolina at Chapel Hill
Address: Campus Box #7160
Chapel Hill, NC, 27510-7160, USA
E-mail: [log in to unmask]
Office phone: 919-843-6629
|