Hi Phil,
that is indeed what I meant.
I should have said 'additive' rather than linear with regard to the commutive order of operations. As smoothing is only adding (a portion of) voxel values together, and the GLM is fitted to all voxels independently, the 'removal' of any component reflected in a voxel carrying components of its neighbours should be nearly identical to 'cleaning' the neighbour first (using GLM) and then smoothing. In other words, when movement is removed from a neighbouring voxel, the 'cleaned' time series smoothed from that neighbour (partially) added to the current voxel, or whether you sum up neighbouring voxels and then remove the component should be identical, right? This assumes the modeling (and subsequent removal) of movement is fully linear (GLM is).
So when the result is the same, why bother?
I have seen nuisance regressors (such as movement) added to GLMs of seed rsfMRI analyses frequently... This of course assumes you are planning to do rsfMRI with the GLM seed method.
As you will need to subtract a regressor coefficient times a (movement) beta from your signal voxel by voxel, I think you will need to write your own batch module and add that to the matlabbatch system of SPM. But I never went down that road (it is said to work though), so I'd actually do this all in my own matlab script, and call portions of the matlabbatch system SPM used by loading jobs and calling spm_jobman repeatedly.
Having said all that, it is highly unlikely you remove substantial movement by simply linearly regressing out (realignment) movement parameters in a GLM, irrespective of the order. Movement effects in fMRI data are highly non-linear and very different from voxel to voxel (think of spin-history artifacts, a slice being excited, rotated a bit, and then read, etc etc). After all your efforts you would still dont know at all whether your correlations over voxels is due to movement (not by a long shot).
rsfMRI done that way is only fully free of correlated movement when obtained from a cadaver. But that would rule out much cognition ;-)
A fully MR physical model correcting for movement taking into account k-space progression during EPI readout, a full 3D reconstruction of the excited slice and receive coils in space, etc etc, could come close to really removing most motion. But although I've seen efforts i havent seen anything convincing yet (when I havent missed anything).
Good luck,
Bas
--------------------------------------------------
Dr. S.F.W. Neggers
Division of Brain Research
Rudolf Magnus Institute for Neuroscience
Utrecht University Medical Center
Visiting : Heidelberglaan 100, 3584 CX Utrecht
Room B.01.1.03
Mail : Huispost B01.206, P.O. Box 85500
3508 GA Utrecht, the Netherlands
Tel : +31 (0)88 7559609
Fax : +31 (0)88 7555443
E-mail : [log in to unmask]<mailto:[log in to unmask]>
Web : http://www.neuromri.nl/people/bas-neggers
: http://www.brainsciencetools.com<http://www.brainsciencetools.com/> (CEO)
--------------------------------------------------
________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Philip R. Baldwin [[log in to unmask]]
Sent: Wednesday, October 31, 2012 6:48 PM
To: [log in to unmask]
Subject: Re: [SPM] is there a batch module for regressing out head motion for preprocessing
Hi Bas and Eugenio,
Thanks for your suggestions.
I am sympathetic to Bas’s point of view. The projection of the signal into the subspace orthogonal to the motion is an operation on the temporal components (does not blend neighboring voxel values), whereas the smoothing operation takes place on the spatial components. Although linear operations are generally not commutative, in this case the smoothing (transformation to the fourier basis) does not affect the time components and the projection along the motion.
Mathematically, I believe one could write
S’ = (1 – Mhat Mhat) * S * H
Where S, the original signal, is time by voxels
S’, the reduced signal
Mhat is the normalized motion
H the spatial filter
Notice it does not matter the order of the projection on the left or the filtering on the right.
On the other hand, the convention I see in the literature is to regress before smoothing, so I will have to be consistent with what is published.
Can anyone give me some pointers to creating a new module to the batch script editor?
Thank you! Phil Baldwin , Houston, TX
------------------------------------------------------------------------------
De informatie opgenomen in dit bericht kan vertrouwelijk zijn en is
uitsluitend bestemd voor de geadresseerde. Indien u dit bericht onterecht
ontvangt, wordt u verzocht de inhoud niet te gebruiken en de afzender direct
te informeren door het bericht te retourneren. Het Universitair Medisch
Centrum Utrecht is een publiekrechtelijke rechtspersoon in de zin van de W.H.W.
(Wet Hoger Onderwijs en Wetenschappelijk Onderzoek) en staat geregistreerd bij
de Kamer van Koophandel voor Midden-Nederland onder nr. 30244197.
Denk s.v.p aan het milieu voor u deze e-mail afdrukt.
------------------------------------------------------------------------------
This message may contain confidential information and is intended exclusively
for the addressee. If you receive this message unintentionally, please do not
use the contents but notify the sender immediately by return e-mail. University
Medical Center Utrecht is a legal person by public law and is registered at
the Chamber of Commerce for Midden-Nederland under no. 30244197.
Please consider the environment before printing this e-mail.
|