hi everybody,
just for your information: no more help on this needed, i found the problem.
after I normalized the signal amplitude and std.dev., all the zero
voxels outside the brain were no longer zero, but had some finite value.
that screwed up the implicit masking (which looks for zeros): the mask
was one for the whole bounding box, and that took too much memory (while
creating a variable in spm_est_smoothness), and matlab got out-of-memory.
now i only normalize the voxels in the original image that are non-zero,
and now the mask is OK again, and so is my regression model.
I knew it: it was something trivial that i did not think of... :-)
bye,
HS
--
Dr. rer. nat. Hartmut Schuetze
Universitaetsklinik f. Anaesthesiologie und Intensivtherapie
Universitaetsklinik f. Neurologie II
Leipziger Str. 44
39120 Magdeburg
Phone: +49 391 67 15145
Fax : +49 391 67 13501
|