James-
Sorry for the long winded answer. If your system is not a ECAT system, you
can probably ignore this.
If your scanner is an ECAT system, there might be a small difference in the
output between 16-bit and 32-bit data. Depending on the setup, some ECAT
scanners generate a separate Calibration scaling factor for each slice in
the image. This allows the 16-bit integers of the ECAT format file to
contain a tiny bit more precision. With SPM, a single scaling factor must
be applied to ALL slices. Therefore, the only way to retain the inherent
precision is to convert the files to 32-bit real numbers (which doubles the
file size).
To check what your ECAT scanner does, convert the image with MRIcro (choose
"Convert ECAT to Analyze" from the "Import" menu) and then select "Display
Foreign Header" from the "Import" menu. If you see a listing of different
calibration values for each slice, you should convert your files to 32-bit
real values.
If you use MRIcro to convert your files, you can either choose to convert
ECAT files to 32-bit reals (the default) or retain the 'raw' 16-bit data.
The latter choice is only recommended if your ECAT scanner is set to apply
the same calibration factor to all slices. This is described in the
"Exceptions" section of the web page:
http://www.psychology.nottingham.ac.uk/staff/cr1/faq.html#lossless
Having said that, the variation in the calibrations may be tiny, and may be
statistically insignificant.
-c
At 05:23 AM 1/25/02 -0600, Patterson, James wrote:
>Hello,
>
>I have a dataset with FDG images that came off the PET scanner in two
>different methods. Some are floats, others are signed 16-bit. SPM doesn't
>seem to mind...did a test analysis with mixed datatypes, and didn't
>complain. Could someone comment on how SPM does this, and whether I should
>convert all to one format or another? Also could someone remind me where I
>cam find the scale factor in Matlab, I would like to take a peek at it and
>see the differences there between the two types.
>
>Thanks,
>
>James
>--
>===========================================
>James C. Patterson II, MD/PhD
>Director of PET Neuroimaging Research
>PET Imaging Center, Biomedical Research Foundation
>Assistant Professor of Psychiatry
>Louisiana State University Health Sciences Center
>Shreveport, Louisiana
>318-675-6040 - Psychiatry
>318-675-6148 - fax
>318-675-4000 - PET Center
>318-675-4020 - fax
>===========================================
__________________________________________
Dr. Chris Rorden
Department of Psychology
University of Nottingham
Nottingham NG7 2RD, UK
Tel: +44 [0]115-951-5294
Fax: +44 [0]115-951-5324
http://www.psychology.nottingham.ac.uk/staff/cr1/
|