[MINC-users] mni_autoreg 0.99.6

Giovanna Pellecchia giovanna.pellecchia at camhpet.ca
Tue Sep 15 13:25:24 EDT 2009


> This is not necessarily a bug but a (unnecessary) limitation.
> That part of the code (which is used only with -vr) is identical
> in 0.99.3 and 0.99.6.

Thanks for the feedback - still, if the code is identical, then I do 
not understand why it fails in one case and not the other. Something 
must have changed between the two versions - my understanding is that 
the -vr option is seldom used, so probably it escaped notice so far.

> Can you do a mincinfo on your mnc file? If it says float or
> double, that's the problem. For the purpose of registration,
> you can convert it to unsigned short (use mincreshape -unsigned
> -short). If the problem persists, let me know. I know that
> there was a recent change to convert the image values to float
> internally, but I think this happens later in the code.
>
> bye
>
> Claude
>

I checked my images and they're all signed short. After switching back 
to minctracc version 0.99.3, they work without problems.

Thanks,
Giovanna.


>>> I have encountered what seems to be a bug in release
>>> 0.99.6. When
>>> minctracc is called with the -vr option, it fails with the
>>> following
>>> error:
>>>
>>> minctracc -vr AC_mri.mnc AC_mri.mnc test.xfm
>>> COG of v1: -1.718248 29.835024 2.200544
>>> COG of v2: -1.718248 29.835024 2.200544
>>>     [trans] = 0.000000 0.000000 0.000000
>>>
>>> Error in minctracc in file segment_table.c, line 34
>>> Currently an unsupported data type (8).
>>>
>>> Version 0.99.3 on the same system works correctly.
>>> The system is ubuntu ibex (2.6.27-14-generic #1 SMP Fri Jul
>>> 24 23:49:25
>>> UTC 2009 x86_64 GNU/Linux), but I do not know whether this
>>> problem is
>>> restricted to the x86_64. All the minc tools have been
>>> compiled from
>>> the latest source packages from the bic, except for HDF5:
>>> libminc: 2.0.18
>>> netcdf : 3.5.0
>>> HDF5   : 1.6.9
>>>
>>> Also, with the mni_autoreg version 0.99.6 the MRI to PET
>>> coregistrations seem to be quite off their targets. This is
>>> corrected
>>> in the 0.99.3 version (this might or might not be related
>>> to the bug
>>> above).

_________________________
Giovanna Pellecchia

Research Analyst
Toronto Western Hospital &
CAMH-PET Imaging Center
250 College Street
Toronto, Ontario
Canada M5T 1R8
Ph: 416-535-8501x7395
Fax: 416-979-3855
Email: giovanna.pellecchia at camhpet.ca



More information about the MINC-users mailing list