Hi,
If I've understood correctly, you have figured out how to do 3 dof
translation-only registration using an appropriate schedule file
but you are unhappy with the results.
You can certainly try other cost functions, via the -cost option,
but I don't know if it will fix the problem. If it doesn't work then
make the data available on a website/ftp (don't email it) and I
can see if there are other ways of fixing it.
The second question - how to update the *.feat/reg directory
is slightly tricky. There are a number of commands in the
report.com that call flirt, convert_xfm and slicer which generate
the required transforms and images associated with them.
If you are familiar with how flirt works, you can pick out the
appropriate commands and re-run them. The only difficulty
is that you must not rerun the flirt *registration* calls - that is,
the ones that try to register example_func to highres and
highres to standard. These calls will overwrite the new
transformations that you have found - which is obviously
not what you want. So, as long as the flirt calls include
-applyxfm then they are safe to run. Also, the slicer calls
will only work if the appropriate images
(e.g. example_func2highres.hdr/img) have been generated
by the flirt registration call that you've done, or if not, by
calling flirt with -applyxfm and the saved matrix.
Hope this helps.
All the best,
Mark
Ruurd Schoonhoven wrote:
>Dear fMRIBbers,
>I'd appreciate your suggestions for the following issues:
>I have 13 3.5mm oblique functional slices (parallel to the lateral sulcus)
>that I want to register on a whole head anatomical scan. The information
>available appears to be a bit limited for straight-forward registration
>with 6 dof, and therefore I have first rotated the anatomical scans around
>the y axis over the appropriate angle (known from the experiment), so that
>the orientation of the new anatomical scan corresponds to that of the
>functional scans. Then I want to do the registration with only tranlational
>dofs. Here I get some problems:
>1) running flirt from the command line with -dof 3 gives an error message
>(erroneous dof 3, using 6 instead), and the result is a quite incorrectly
>rotated scan. Adding -nosearch does not change this (though it should
>according to the explanation of this option), nor does adding -searchrx -1
>1 etc to limit the rotation angle to 1 degree: the result remains a
>rotation about 30, 40, 50 or so degrees.
>Running flirt from the FEAT GUI with dof 3 makes that dof 6 is actually
>applied but a translation schedule is invoked as well, which leads to a
>pure translational registration, however with a mismatch in the result of
>sometimes 2-3 cm especially in the inferior/superior direction.
>My question is: how can I achieve an only translational registration which
>will also produce a good match? Other options? Other cost function? etc
>2)Once this is solved, probably by some command line run of flirt, what
>should I do next to make sure that the proper results are stored in the
>various feat subdirectories so that a) the result of the individual
>report.html is correctly updated for the registration and b) the proper
>files etc are substituted to allow for a higher level analysis which uses
>the feat (sub)directories as input? This I cannot exactly recover from the
>report.com/log files.
>Thanks in advance for your help,
>regards,
>Ruurd Schoonhoven
>VUmc Amsterdam
>
>
|