Hi,
1) Yes this sounds sensible
2) You could compare a number of different things such as the extent of
activation to your task both with and without inclusion of the motion
parameters. The difference between these is either false positives (found
without motion regressors included) or false negatives (with motion
regressors included). You cannot easily tell between these two however -
but you're safer with motion regressors because you know your remaining
activation is not related to motion. You could compare cluster size in terms
of voxel numbers to obtain a % difference. The marsbar toolbox contains some
extra features to help with further processing tasks not available in SPM
which might be useful.
You will probably also find the paper below useful.
Lund TE, Nørgaard MD, Rostrup E, Rowe JB, Paulson OB., Motion or
activity: their role in intra- and inter-subject variation in fMRI.
Neuroimage. 2005 Jul 1;26(3):960-4.
David
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On
Behalf Of Krishna A. Dani
Sent: 28 April 2008 18:28
To: [log in to unmask]
Subject: [SPM] Volterra Expansion of Motion Parameters
Dear SPM,
I wonder if you could advise? We are current performing sequences on acute
patients. The data is fairly noisy and we would like to determine how much
is related to motion (as opposed to physiological parameters). I am keen
to produce 'maps' where voxels where there is a significant impact on
signal from motion are highlighted, similar to the approach described by
Friston (1996) and used in epilepsy by Lemieux et al (MRI 25, 2007, 894-
90; Modelling large motion events in fMRI studies in patients with
epilepsy).
I have used SPM2 to do this, and have used volterra expansions of motion
parameters to generate 24 regressors, which i have included in the model
at the preprocessing stage. At the 'results section I have then defined an
new F contrast with no masking and have included all 24 regressors and the
constant (excluding only 1 regressor which defines the experimental
model). I have assumed the map produced represents all voxels where motion
has a significant impact on the signal.
May I ask
1) Is this the best way / correct way of doing this?
2) How does one determine the % of voxels affected?
Many Thanks in anticipation.
Regards
Krishna
Dr Krishna A Dani
Clinical Research Fellow
Department of Neurology
Division of Clinical Neurosciences
University of Glasgow
UK
|