Dear Stefan and SPMers,
A couple of questions about the problem with using mean voxel intensity
as a baseline to measure % signal change, rather than mean volume intensity.
In your example below, is it the case that voxels A and B have equivalent
signal changes from baseline (i.e., 'the same grey matter value at A and
B'), but wind up with different % signal changes because their baseline
intensities differ (i.e., 'the mean intensities of A and B will differ at
the point of making inferences')? If so, then your example appears to
assume that a change of 10 units from a baseline of 500 (say, for Voxel A)
is equivalent to a change of 10 units from a baseline of 1000 (say, for
Voxel B).
Might it not be, however, that a deviation of 10 units from a baseline
of 500 is a larger % signal change than a deviation of 10 units from a
baseline of 1000? In other words, given that the mean voxel intensity is
500 for voxel A and 1000 for voxel B, might a change of 10 units be harder
to come by for Voxel A than for Voxel B and, therefore, indicate a larger
change in activation?
For example, suppose that Voxel A is surrounded by white matter, while
Voxel B is surrounded by gray matter. The functional data are smoothed,
which allows the activity of Voxels A and B to be influenced by the activity
of their neighboring voxels. Let's say that Voxel A's mean activation is
only 500 due to smoothing near the white matter, while Voxel B's mean
activation is 1000 because it is surrounded by gray matter. Now, imagine
that a stimulus is presented, which activates Voxel A, Voxel B and Voxel B's
neighbors. The activation of Voxel A would be due entirely to itself since
it is completely surrounded by white matter. The activation of Voxel B,
however, would be due both to itself and to the activation of its neighbors
through smoothing. If Voxel B's neighbors are also responsive to the
stimulus (which is likely under the assumption that functional areas
comprise several voxels), then smoothing might make a change of 10 units
relatively easy to come by for Voxel B In contrast, a change of 10 units
would be harder to acheive for Voxel A, since it would receive no help from
its white-matter neighbors. In this example, it seems to me using mean
voxel intensity as a baseline against which to measure % signal change might
be appropriate, since the same factors that influence a voxel's mean
intensity (e.g., presence/absence of surrounding white matter) also
influence that voxel's responsiveness to a stimulus (e.g., whether the voxel
increases its activation by 10 vs. 5 units).
I'd be interested to know what others think about this issue.
Daniel Weissman, PhD
Center for Cognitive Neuroscience
Duke University
Durham, NC 27705
A simple
> case would be e.g. two activations at voxels A and B, where A and B both
> are in grey matter, but the neighbouring voxels are of different image
> intensities. Given partial volume effects and some smoothing, the mean
> intensities of A and B will be different at the point of making
> inferences, although the signal change should be related to the same
> grey matter value at A and B.
>
>
> Stefan
> --
> Stefan Kiebel
> Functional Imaging Laboratory
> Wellcome Dept. of Cognitive Neurology
> 12 Queen Square
> WC1N 3BG London, UK
> Tel.: +44-(0)20-7833-7478
> FAX : -7813-1420
> email: [log in to unmask]
>
>
|