[MINC-development] nu_correct on linux

[John G. Sled]

Dear Richard,

The mask that is used by N3 is the intersection of that part of the
volume with values greater than 1.0 and the part of the mask with 
values greater than 0.  The N3 algorithm is likely to quit after 
a single iteration if only a few voxels meet this criteria.
I suggest that you check the image data to see that the real values
are greater than 1.0.

with regards,

John Sled
 


On Tue, Jan 06, 2004 at 09:42:11AM -0000, Richard Boyes wrote:
> dear peter,
> 
> thanks for your feedback - unfortunately it crops up with a different
> error, this time:
> 
> rboyes@gyrus:~/cvs/bias-correct-regress/custom-sw/regress/data/junk
> rawtominc -byte -obyte -clobber -input ./bmask00000.img -coronal -xstep
> 0.9766 -ystep 1.8000 -zstep 0.9766 -range 0 255 -real_range 0 255
> ./bmask00000.mnc 120 256 256
> rboyes@gyrus:~/cvs/bias-correct-regress/custom-sw/regress/data/junk
> nu_correct -clobber -mask ./bmask00000.mnc ./00000.mnc
> /correct00000.mnc  -stop 0.0005 -fwhm 0.075 -distance 100 -iterations
> 350 -mapping_dir . -shrink 3
> Transforming slices:.........................................Done
> No points were chosen by the specified criteria
> resample_labels: crashed while running extracttag (termination status=256)
> nu_estimate_np_and_em: crashed while running resample_labels (termination
> status=256)
> nu_correct: crashed while running nu_estimate_np_and_em (termination
> status=256)
> 
> i've tried running nu_correct without the mask (below), but i'm getting the
> problem of 1e-27 field change again. any other ideas, or can i send
> you any other information?
> 
> rboyes@gyrus:~/cvs/bias-correct-regress/custom-sw/regress/data/junk
> nu_correct -clobber ./00000.mnc ./correct00000.mnc -stop 0.0005 -fwhm
> 0.075 -distance 100 -iterations 350 -mapping_dir . -shrink 3
> Processing:.........................................Done
> Not implemented yet in cache_volume_range_has_changed()
> Not implemented yet in cache_volume_range_has_changed()
> Number of iterations: 1
> CV of field change: 7.88338e-27
> Transforming slices:.........................................Done
> 256+0 records in
> 256+0 records out
> Not implemented yet in cache_volume_range_has_changed()
> Not implemented yet in cache_volume_range_has_changed()
> Transforming
> slices:.....................................................................
> ...................................................Done
> 
> 
> thanks, richard.
> 
> 
> > -----Original Message-----
> > From: minc-development-admin@bic.mni.mcgill.ca
> > [mailto:minc-development-admin@bic.mni.mcgill.ca]On Behalf Of Peter
> > NEELIN
> > Sent: 05 January 2004 03:02
> > To: Richard Boyes
> > Cc: MINC
> > Subject: Re: [MINC-development] nu_correct on linux
> >
> >
> > On Tue, 23 Dec 2003, Richard Boyes wrote:
> >
> > > when i try nu_correct, it jumps out straight away,
> > > saying the field change is 1E-27. this has happened
> > > for multiple images with bad bias fields, which doesn't
> > > seem correct.
> >
> > If I recall correctly, nu_correct had some difficulties with masks that
> > had ranges that did not correspond to its expectations (is that correct,
> > John?). This is just a guess, but try using the -scan_range option for the
> > mask images as well:
> >
> > #Mask file - no need to swap bytes
> > rawtominc -byte -obyte -clobber -input ./bmask00000.img -coronal \
> >    -xstep 0.9766 -ystep 1.8000 -zstep 0.9766 \
> >    -orange 0 255 -scan_range ./bmask00000.mnc 120 256 256
> >
> > Alternatively, replace the -orange option with
> > "-range 0 255 -real_range 0 255" and do not use -scan_range.
> >
> >             Peter
> > ----
> >             Peter Neelin (neelin@bic.mni.mcgill.ca)
> >
> > _______________________________________________
> > MINC-development mailing list
> > MINC-development@bic.mni.mcgill.ca
> > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-development
> 
> _______________________________________________
> MINC-development mailing list
> MINC-development@bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-development
>