Dear Radhika,

I agree with Steve about the RAM/swap space issue.
As it is only a couple of things that fail, and they are the most memory
intensive processes, then it does seem to point to this.  It also explains
the slow running of FLIRT.  A rule of thumb for FLIRT is that it takes
about 30 minutes to do a high-resolution registration on a PIII 1Ghz+ with
256MB ram.  Of course it does depend on many other things, but this should
give you an idea.

a)
As for your registration problem, it is caused by the fact that your
input image is 180 degrees out from the avg152 (it is upside down when you
look at it in slices).  This can often be fixed by enlarging the search
space (using the full +/- 180 degrees in each angle), although a better
way is to provide it with an initial "flipping" matrix to get the orientation
close to begin with.  This flipping can be done by avwswapdim.  Here's what
I did to get a nice registration:

avwswapdim subject_brain x -y -z > swap.mat
flirt -ref /usr/local/fsl/etc/standard/avg152T1_brain -init swap.mat -in subject_brain -out subject_brain2av -omat
subject_brain2avg.mat

This wrote the image subject_brain2avg
The matrix file, subject_brain2avg.mat, was:
1.03258  0.0526024  -0.0396348  -35.5067
0.0497804  -1.21218  -0.163621  275.445
-0.0463189  0.203971  -1.14437  193.038
0  0  0  1

(if you don't have avwswapdim then download avwswapdim.cc and Makefile
from www.fmrib.ox.ac.uk/~mark/files/avwswapdim.cc and www.fmrib.ox.ac.uk/~mark/files/Makefile, put these files in
$FSLDIR/stable-src/avwutils , cd to this directory and type "make install-stable")

b) We check the field sizeof_hdr to see if the byte ordering is native to
the machine or not and then reorder this on reading.  Hence the FSL utilities
should be able to read either byte-ordering without trouble.  On writing, they
use the native format, regardless of whether the input needed byte reversal.
Hence the output images should be correct for your machine.  However, the data
type used might well be changed.  It is not unusual for images that start as
8 bit to end as 32 bit float or something, depending on what processing is
required.  All other fields should be correctly copied, so I can't think of
any other reason why you couldn't read our output images.  If you want a good
look at the reading/writing code, see the source in $FSLDIR/stable-src/avwio/

c) As Steve says, this image is stored as bytes (8 bit).

Cheers,
        Mark


Stephen Smith wrote:
>
> HI - looks like you just don't have enough RAM/swap space probably. How
> much do you have? I'm not sure exactly how much you need as it's machine
> dependent, but you probably need at least 128MB ram and maybe 300MB swap -
> maybe more.
>
> a) I'm away at a conference so can't look right now - maybe Mark will get
> a chance. but you could try running BET first, also try enlarging the
> search space under advanced options if the images are not already roughyl
> in the same orientation.
>
> b) if you look at the source for example in fsl/stable-src/ss/avwio.c (or
> something like that) you will see the test for byte order - AVW supports
> either byte order so you just have to test the first number in the header.
> hopefully the source code is clear.
>
> c) I think mni152 may be 8bit unsigned char - can't rememebr right now -
> so that would explain it!
>
> ttfn :)
>
> On Wed, 5 Jun 2002, Radhika Srinivasan wrote:
>
> > Hi,
> >
> > I just started using FSL and downloaded the FEEDS routines to evaluate my
> > FSL installation. I get a few errors as appended below. It also takes a very
> > long time for each step and does not seem to finish the evaluation. Could
> > you please advise on whether this means that the download is not right. I
> > have attached the errors below. What is the average time for the completion
> > for the FSL functions (eg. FLIRT)? I would also appreciate some advise on
> > the following..
> >
> > a) I was not able to get FLIRT to work with my T1 weighted dataset. The
> > program completes but the result does not seem right. I used the default
> > FLIRT GUI parameters. The images are in the URL
> > ftp://ftp.mrsc.ucsf.edu/pub/radhika/subject_brain2standard.img and
> > subject_brain.img. I used to latter to register the images to the avg152
> > image.
> >
> > b) Is there any field in the analyze header that flags the byte swapping. We
> > have custom built routines to view analyze images and they dont seem to work
> > on FSL image files.
> > c) I was not able to understand why avg152T1.img had only 902629 bytes when
> > its image dimension (91x109x91) should indicate twice as many bytes. Are
> > these short integer?
> >
> > Thanks in advance for your help.
> > radhika
> >
> >
> > pride.270> time ./RUN &
> > [1] 24167
> > pride.271>
> > FSL Evaluation and Example Data Suite v1.3
> >
> > start time = Tue Jun  4 18:50:22 PDT 2002
> > hostname = pride.ucsf.edu
> > os = SunOS pride.ucsf.edu 5.8 Generic_108528-14 sun4u sparc SUNW,Ultra-5_10
> >
> > Starting BET at Tue Jun  4 18:50:22 PDT 2002
> > /home/rsriniva/fsl/bin/bet /data/cashew/perfusion/feeds/data/structural
> > /data/cashew/perfusion/feeds/results/structural_brain
> > % error = 0.0
> >
> > Starting SUSAN at Tue Jun  4 18:52:31 PDT 2002
> > /home/rsriniva/fsl/bin/susan_smooth
> > /data/cashew/perfusion/feeds/data/structural 2000
> > /data/cashew/perfusion/feeds/results/structural_susan 2 3 1 0
> > % error = 0.0
> >
> > Starting FAST at Tue Jun  4 19:33:29 PDT 2002
> > /home/rsriniva/fsl/bin/fast -od
> > /data/cashew/perfusion/feeds/results/structural_brain -ob -or -e -ov
> > /data/cashew/perfusion/feeds/data/structural_brain
> > checking error on single-image binary segmentation:
> > No output image created
> > Warning - test failed!
> > checking error on bias field:
> > No output image created
> > Warning - test failed!
> > checking error on restored image:
> > No output image created
> > Warning - test failed!
> > checking error on partial volume images:
> > No output image created
> > Warning - test failed!
> > No output image created
> > Warning - test failed!
> > No output image created
> > Warning - test failed!
> >
> > Starting FLIRT at Tue Jun  4 20:14:13 PDT 2002
> > /home/rsriniva/fsl/bin/flirt -ref
> > /home/rsriniva/fsl/etc/standard/avg152T1_brain -in
> > /data/cashew/perfusion/feeds/data/structural_brain -out
> > /data/cashew/perfusion/feeds/results/structural_brain2standard -omat
> > /data/cashew/perfusion/feeds/results/structural_brain2standard.mat -interp
> > sinc
> > % error = 0.0
> >
> > Starting PRELUDE & FUGUE at Tue Jun  4 20:34:57 PDT 2002
> > /home/rsriniva/fsl/bin/prelude -c /data/cashew/perfusion/feeds/data/fieldmap
> > -u /data/cashew/perfusion/feeds/results/unwrapped_phase
> > % error = 0.0
> > /home/rsriniva/fsl/bin/fugue -i /data/cashew/perfusion/feeds/data/epi -p
> > /data/cashew/perfusion/feeds/results/unwrapped_phase -d 0.295 -u
> > /data/cashew/perfusion/feeds/results/unwarped_epi
> > % error = 0.0
> >
> > Starting SIENAX at Tue Jun  4 20:35:13 PDT 2002
> > /home/rsriniva/fsl/bin/sienax
> > /data/cashew/perfusion/feeds/results/structural -d
> > syntax error in expression "100.0 * ( save:args
> > 164 is unimplemented
> > 141 is unimplemente"
> >     while executing
> > "expr 100.0 * ( $a - $b ) / $denom "
> >     (procedure "simpleperror" line 5)
> >     invoked from within
> > "simpleperror $A $B 1600000"
> >     (file "./RUN" line 196)
> >
> > [1]    Exit 1                        ./RUN
> >
>
>  Stephen M. Smith
>  Head of Image Analysis, FMRIB
>
>  Oxford University Centre for Functional MRI of the Brain
>  John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
>  +44 (0) 1865 222726  (fax 222717)
>
>  [log in to unmask]  http://www.fmrib.ox.ac.uk/~steve