Hi Katia,
We use the following to get around a similar problem:
Around line 20 of spm_get_data.m, just after
try
v = cat(2,V{:});
end
add the following:
else
V2 = spm_vol(V(1).fname);
if spm_type(V2.dim(4), 'swapped') ~ = spm_type(V(1).dim(4), 'swapped')
V = spm_update_vol(V);
end
The above snippet refers to spm_update_vol.m which you can create as
follows:
function [V] = spm_update_vol(V)
% changes the swap info to fix byte ordering problem
% Yuanyuan Jia 11/05
V2 = spm_vol(V(1).fname);
if spm_type(V2.dim(4), 'swapped') ~=spm_type(V91).dim(4), 'swapped')
[V.dim] = deal(V2.dim);
end
Hope this helps. We have found it to be useful when we transfer data
from Linux to Mac OSX.
-Skip
On Apr 6, 2006, at 6:21 PM, Katia DAUCHOT wrote:
> Hello,
>
> can anybody please help with the following problem:
>
> I've generated SPM2 data (pre-processing, model generation and
> estimation, contrasts etc) using
> SPM2 under Matlab6.5, PC Linux. Everything worked out fine.
>
> Then I transferred the result directories to my PowerBook G4 (osX,
> Tiger) on which I installed
> Matlab7.1 and the recent SPM2 with the updates.
>
> All the transferred *img files look OK, i.e. I can look at them using
> the "display" function, and both
> the header and the intensity values are identical on either platform (
> PC or Mac).
>
> However, when I compute "results" (hit "results", select a contrast,
> and go through the whole
> procedure of choosing a threshold, the extent etc), the outcome is
> correct on the PC but wrong on
> the Mac. The resulting file shows numerous disperesed separate voxels,
> among the majority
> "empty" ones (see the attached file).
> The whole thing looks much like a format problem (typically appeared
> years ago, when we passed
> from Solaris Unix to PC and vice-versa: e.g. reading double float
> files as if they were simple float,
> or the big/little endian error). However, as I said, with "display"
> all files look OK and identical on
> either platform.
>
> Here's some additional information on all this:
>
> 1. SPM doesn't show any error msg when this happens.
> 2. Voxels in the volupme don't seem to have any value attributed (when
> clicked upon, no value
> appears).
> 2. The problem is the same, whatever SPM.mat file, contrast or
> threshold chosen, within my set of
> data.
> 3. It is not specific to one particular Mac or PC machine. The data
> being transferred to a different
> PC, with SPM2 running under Matlab6.5, the problem disappeared. The
> data transferred to a
> different Mac (recent GM running SPM2 under Matlab7.1) the same pb
> re-appeared.
> 4. As I said, all the *img files look identical and present no visible
> pbs, on either platform (Mac, PC
> Windows or PC Linux).
>
> As we used SPM2 in all cases (we even tried to homogenize the
> versions, between the platforms as
> well as using the latest updates), we suspected that the pb be related
> to the Matlab-version. I tried
> then to re-install Matlab6.5 on my Mac. However this requires XDarwin.
> And XDarwin seems to be
> poorly compatible with Mac osX Tiger.
> Trying to install it we couldn't make it work, nor any version of
> Matlab with it.( At the end, had to
> re-install X11 and Matlab7.1 from scratch.) And there seemed to be a
> conflict between the
> different X servers (X11 and XDarwin). I hear other people had
> problems installing XDarwin under
> Tiger, too.
>
> We're running out of ideas...
>
> Can someone please help?
>
> Thanks so much in advance
>
> Katia Dauchot
>
>
Glenn T. Stebbins, Ph.D.
Associate Professor of Neurological Sciences
Tel: 312-563-3854
Fax: 312-563-4009
email: [log in to unmask]
|