[Insight-users] itk neighborhood operator image filter error
Luis Ibanez
luis.ibanez at kitware.com
Tue, 20 Apr 2004 14:34:15 -0400
Hi raghu,
The optimizer scales do not need to be different among
X, Y, and Z. Unless the anisotropy ratio as big as 1:10.
The goal of the parameter scaling is to deal with large
dynamic range differences such as degrees versus
millimeters where the ratio is in the range of 1:100.
If your metric plots shows the Metric behaving smoothly,
you can take the risk of *increasing* (not decreasing)
the step length of the optimizer, and probably reducing
the number of iterations.
Note however that if your registration is using all the
allotted iterations, it means that it has not converged,
and that you still have homework to do in parameter
tunning.
Regards,
Luis
------------------------
raghu venkatram wrote:
> Hello Luis,
>
> Thank you for the tip, it was ofcourse my CT volume, I was only reading
> 2 slices. I am not sure why this is happening as I used MRICro to
> convert both my CT and MR volumes into mayo clinic analyze format and
> everything works fine. using ITK's dicom series reader reads my MR
> volume and am able to wrtie it out, but it dies on the third slice of my
> CT volume.
>
> Also regarding registration itself, i am using viola-wells mutual
> information, with gradient descent and the translation transform. in
> this set up should my my optimizer scales for x and y be equal to 1
> while my optimizer scale for z slightly greater than 1? my z in the
> volume is much smaller than the inplane dimensions.
>
> with a step size of 18 and 500 epochs, i am getting good results, and a
> final metric value of 0.2-0.4, a plot of the iteration vs metric shows
> that the metric is fairly stable with this, should i decrease the step
> size and increase the iterations further?
>
>
>
> thanks
>
> raghu
>
> >From: Luis Ibanez <luis.ibanez at kitware.com>
> >To: raghu venkatram <raghu_420 at hotmail.com>
> >CC: insight-users at itk.org
> >Subject: Re: [Insight-users] itk neighborhood operator image filter error
> >Date: Mon, 19 Apr 2004 19:43:23 -0400
> >
> >
> >Hi Raghu,
> >
> >It is likely that your process for reading the
> >DICOM series is failing. The fact that it shows
> >up in the Neighborhood iterator may simply indicate
> >that you are reading an image of a single slice,
> >for example.
> >
> >Please connect a ImageFileWriter directly to the
> >output of the DicomSeries reader and save the
> >image in another format. MetaImage will be the
> >best choice since then you can look at the header
> >and figure out what information from the DICOM
> >file is getting lost on the way.
> >
> >Chances are that you only have a Input/Output
> >problem and not a processing problem.
> >
> >
> >I will also suggest you to try the example
> >
> > Insight/Examples/IO/
> > DicomSeriesReadImageWrite.cxx
> >
> >for reading your Dicom images, and to make sure
> >that you are using the series reader correctly.
> >
> >For example, note that you should specify the
> >string identifying the series to be read.
> >
> >
> >Please let us know what you find.
> >
> >
> >
> > Thanks,
> >
> >
> > Luis
> >
> >
> >-----------------------
> >raghu venkatram wrote:
> >>Hello Luis,
> >> I am trying to register a MR and CT image using ImagetoImage
> >>mutual information measure, I have no problems whatsoever using the
> >>analyze image format.
> >> when i switch to dicom, and use the series reader, i can read
> >>both the series, but my program terminates with an error window
> >>asking me to notify microsoft, debug option takes me to the
> >>itkneighborhoodoperator filter.
> >> is there something obvious that i am doing wrong? my image types
> >>are unsigned short 3 dimensions.
> >> thanks,
> >>raghu
> >>
> >
> >
> >_______________________________________________
> >Insight-users mailing list
> >Insight-users at itk.org
> >http://www.itk.org/mailman/listinfo/insight-users
>
> ------------------------------------------------------------------------
> Check out MSN PC Safety & Security to help ensure your PC is protected
> and safe. <http://g.msn.com/8HMAENUS/2749??PS=>
> _______________________________________________ Insight-users mailing
> list Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users