Dear Simone,
>I'm not interested in the true camera resolution, but in my resolution after
>reconstruction of the images (which I named the 'raw' images), which I hoped
>to find out using SPM.
[...]
>For our single subject PET data we are interested in the resolution of our
>'raw' images (e.g. nearby camera resolution, after reconstruction). We tried
>to obtain this raw FWHM by doing a statistical analyses on the images which
>are realigned but not normalised or smoothed (avoiding the introduction of
>extra smoothing by one of them) and read the FWHM from the spm output.
In addition to Phil's and John's comments just my 2p -
There are several factors influencing the resolution of your "raw" images.
You can work out the capabilities of your machine (the theoretical maximum
resolution or minimum FWHM) from the sinograms, the "raw data" output.
What you're really interested in, though, is the FWHM of your
_reconstructed_ images. This will be affected by many things, e.g. the
filter used during reconstruction (rapm, Hanning...), as pointed out by
Phil, and the scatter medium (air, water, brain tissue...), among others,
and it will vary across the field of view (FOV) as well. The only way to
work this out, I'm afraid, is the way Phil suggested - taking a point or
line source, putting it into a suitable phantom and get down to measuring,
using the same reconstruction parameters as for your final images. You will
usually want to acquire data for points distributed throughout the FOV.
A good start will be a search on papers containing "performance" and "PET".
In addition to the reference Phil mentioned, you may want to have a look at
Spinks, T.J.
Physical performance of a positron tomograph with retractable septa.
Phys. Med. Biol. 1992, 37(8):1637-1655
which I found quite readable.
I'm afraid for the reasons John mentioned you cannot use SPM to work
smoothness out for you - if you resampled the images all into the same
space, you would have additional smoothing through interpolation, similarly
to the normalisation/ smoothing procedures which you already identified as
confounds.
Moreover, even if your subject had kept perfectly still and you did the
stats on a single subject, the FWHM value SPM would give you would depend
on your reconstructed image resolution _and_ the smoothness of the object
(brain) imaged, and it would also be averaged across the image matrix - so
unfortunately, it would not give you the answer you want.
So, I'm afraid you won't be able to avoid the trip to the physicists
(unless you only want an approximation and can find something on your type
of machine _and_ your exact reconstruction parameters in the literature
which is unlikely).
Sorry about the bad news,
Alexander
--------------------------------------------------------------------------------
Dr Alexander Hammers, MD
Epilepsy Research Group
National Hospital for Neurology and Neurosurgery/ Institute of Neurology
33 Queen Square
London WC1N 3BG
and
MRC Cyclotron Unit
Imperial College School of Medicine
Hammersmith Hospital
DuCane Road
London W12 0NN
Telephone +44-(0)20-8383-3162
(ext. 2651)
Fax +44-(0)20-8383-2029
Email [log in to unmask], [log in to unmask]
|