[MINC-users] Failure of nu-estimate on older t1-weighted brain data.

[Alex ZIJDENBOS]

I'm not sure what could be going on here; Najma sent me some of the
data she used and I ran it through nu_correct v1.10 just fine. I'm not
sure why essentially the same version of n3 would behave differently
on these data. Are you using a standard install, or a quarantined
version? The CachedArray error seems to indicate a failure in creating
a temporary file - that could be a machine resource problem (out of
/tmp space?).

As for the masking, it seems you are trying to mask a native file with
a stereotaxic mask, which is likely to produce unexpected results ;) I
thought N3 checked at least that the sampling lattice of the source
file and mask are the same - it *should* exit gracefully on something
like this.

-- A

On Tue, Feb 28, 2006 at 07:18:25PM -0500, Najmeh Khalili M. wrote:
> 
> Greetings,
> 
> I am trying to perform N3 on t1-data of healthy older brains and I
> encounter two error messages. (nu_estimate, version 1.10)
> 
> First, I got the following error for N=50 data sets:
> 
>  Assertion failed at line 742 in file templates/CachedArray.cc
>  nu_estimate_np_and_em: crashed while running volume_stats (termination
> status=256)
> 
> Next, I used an ICBM mask; succeeded in non-uniformity correction of N=45
> data sets, but failed in 5 with :
> 
>   Direction cosines of mask.mnc and /tmp/native_t1.mnc do not match
>   Failed to shrink mask volume.
>   nu_estimate: crashed while running nu_estimate_np_and_em (termination
>   status=65280)
> 
> The data dimension in the failed cases is identical to those that have
> succeeded. The data quality and the brain structure in failing cases is
> not visually different. I have not seen this problem in younger brains.
> 
> Has someone else experienced such errors and can any one give an
> explanation of what the cause of this might be, please?
> 
> Best,
> Najma
> 
> 
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users