[Insight-users] Re: Brain Tumor + Level Set Filter

Luis Ibanez luis.ibanez at kitware.com
Wed Dec 22 22:45:13 EST 2004


Hi Neha,

Thanks for posting the .zip file with your image.

The problem is that you specified the endianess
incorrectly in the MetaImage header.


You have now

        BinaryDataByteOrderMSB = True


but you should have

        BinaryDataByteOrderMSB = False


Simply displaying the image exposes this error, because
you can see how the pixels on high intensity get saturated
and wrapped.


    Regards,


       Luis



-------------
neha k wrote:

> Hello Luis,
>  
> I am using ITK 1.8.  I did not change pixel type. (They are read as - 
> float in the program )
> Also,  my original image is RAW - 256 * 256,  Signed-Short, Big Endian.
> I read the image as .mhd format.   Please find original image attached.
>  
> Thanks,
> Neha
> 
> */Luis Ibanez <luis.ibanez at kitware.com>/* wrote:
> 
> 
>     Hi Neha,
> 
>     What version of ITK are you using ?
> 
>     You are probably using the ITK version that used to read the PNG
>     images inverting the Y access, and therefore the position of the
>     pixel (30,63) is no longer the center of the tumor.
> 
>     What are you using to read the Raw image into ITK ?
> 
>     Did you changes the pixel types in the code in order to match the
>     input image ?
> 
>     The parameters of a segmentation method are specific for the input
>     pixel type, the image modality and the anatomical structures that
>     you plan to segment.
> 
> 
>     Please make available the original raw image.
> 
> 
>     Thanks
> 
> 
>     Luis
> 
> 
>     ----------------
>     neha k wrote:
> 
>      > Hello Luis,
>      > Thanks for the reply. I have chanegd interataions to 100. But with
>      > same png image, I get different result (I have used ctr.
>     as(79,93). The
>     &! gt; center that you have specified (30,63) seems to be out of
>     brain area.
>      > Also, if i use original raw image, output is blank. I am not sure
>     what
>      > is going wrong. (I saved original raw image as .png image by
>     making it
>      > 8-bits per channel from 16-bit per channel in photoshop , and
>     then sent
>      > to you)
>      > Neha.
>      >
>      > */Luis Ibanez /* wrote:
>      >
>      >
>      > Hi Neha,
>      >
>      > Thanks for posting your image.
>      >
>      > The ThresholdSegmentationLevelSet worked fine in this image,
>      >
>      > I used the example code in:
>      >
>      >
>      > Insight/Examples/Segmentation/
>      > ThresholdSegmentationLevelSetImageFilter.cxx
>      >
>      >
>      > Just replaced the maximum number of iteration in line 200.
>      >
>      > The current value is 1200 iteration,
>      > you must use here 100 iterations.
>      >
>      >
>      > Then in the command line use
>      >
>      >
>      > ThresholdSegmentationLevelSetImageFilter
>      > tumor57.png segmented.png 30 63 5 70 207
>      >
>      >
>      > (in a single command line)
>      >
>      > The numbers: 30 63 are respectively the X,Y approximate coordinates
>      > of the tumor center.
>      >
>      > The number: 5 is a radius to be used for creating a initial level set
>      > as a circle of radius 5 pixels centered at (30,63).
>      >
>      > The numbers: 70 and 207 and the lower and upper thresholds that
>     define
>      > a borde! r for the tumor.
>      >
>      >
>      >
>      > In the future you may find useful to use the interactive application
>      >
>      >
>      > InsightApplications/
>      > ThresholdSegmentationLevelSetFltkGui
>      >
>      >
>      > That allows you to set all this parameters interactively and see the
>      > evolution of the level set iteration by iteration.
>      >
>      > Note that it is *very* important to monitor/supervise the level set
>      > because if you run ! it for too few iterations the tumor will be
>     under
>      > segmented, an if you run it for too many, the level set will leak and
>      > start invading other brain structures.
>      >
>      > Looking at the level set as it is evolving is the best way to figure
>      > out how many iterations are enough for stopping the level set.
>      >
>      > By fine tunning the values of the lower/upper thresholds, the
>     curvature
>      > scaling and the number of iterations you will be able to refine the
>      > borders of the segmented region.
>      >
>      > You could also post-process the segmented image with a filter such as
>      > the BinaryMedianImage! Filter or the new itkVotingBinaryImageFilter.
>      >
>      >
>      >
>      > Please find attached the PNG image with the tumor segmented from your
>      > image.
>      >
>      >
>      >
>      > Regards,
>      >
>      >
>      > Luis
>      >
>      >
>      >
>      >
>     ---------------------------------------------------------------------------
>      > neha k wrote:
>      >
>      > > Hello Luis,
>      > >
>      > > Thanks for your reply. While using this filter, with my Brain tumor
>      > > Image attached, I am not getting any luck with application, i
>      > have tried
>      > > different parameter values.
>      > > I am not sure what Level Set filters would be good to delineate
>      > tumor in
>      > > my image attached. Smoothing did not seem to help for using
>      > > ThresholdLevelSet and ShapeDetection Level Set. My original
>     Image is
>      > > RAW with .mhd header file.
>      > > Any suggestions? How can I tune pramaters to separate tumor? What
>      > type
>      > > of preprocessing and other filters might help?
>      > >
>      > > Thanks,
>      > > Neha
>      > >
>      > > */Luis Ibanez /* wrote:
>      > >
>      > >
>      > > Hi Neah,
>      > >
>      > > Thanks a l! ot for posting such a detailed message.
>      > >
>      > >
>      > > We attempted to reproduce the result that you reported,
>      > > but the program worked fine for us.
>      > >
>      > >
>      > >
>      > > Here are the command line parameters that we used,
>      > > and the output from the program:
>      > >
>      > >
>      > > A) Command line:
>      > >
>      > > ShapeDetectionLevelSetFilter BrainT1Slice.png T1.png
>      > > 81 113 5 1 -0.5 3 0.05 1
>      > >
>      > > Output:
>      > >
>      > > Max. no. iterations: 800
>      > > Max. RMS error: 0.02
>      > >
>      > > No. elpased iterations: 108
>      > > RMS change: 0.0199696
>      > >
>      > >
>      > >
>      > > B) Command line:
>      > >
>      > > ShapeDetectionLevelSetFilter BrainProtonDensitySlice.png PD.png
>      > > 81 113 5 1 -0.5 3 0.05 1
>      > >
>      > > Output:
>      > >
>      > > Max. no. iterations: 800!
>      > > Max. RMS error: 0.02
>      > >
>      > > No! . elpased iterations: 125
>      > > RMS change: 0.0189581
>      > >
>      > >
>      > > The output images (T1.png and PD.png) are attached.
>      > >
>      > >
>      > >
>      > >
>      > > Did you modified th! e code of this Example
>      > > in any way before running it ?
>      > >
>      > >
>      > >
>      > > Please let us know,
>      > >
>      > >
>      > > Thanks
>      > >
>      > >
>      > >
>      > > Luis
>      > >
>      > >
>      > > --------------------------
>      > > neha k wrote:
>      > >
>      > > > Hello,
>      > > >
>      > > > I am trying to check with
>      > > ITK1.8\Example\ShapeDetectionLevelSetFilter,
>      > > > and do not get desired output. I tried this compiled example with
>      > > > following input images and parameters -
>      > > >
>      > > > *1. BrainT1Slice.png(in! put image from \ITK\Examples\Data) -
>     with *
>      > > >
>      > > > Seed Point - (81 113) , Distance - 5.0,
>      > > >
>      > > > Sigma SigmoidAlpha SigmoidBeta are -- 1 .0 -0.5 3.0
>      > > >
>      > > > Curv.! Scaling -- 0.05 Propogation Scaling -- 1
>      > > >
>      > > > *Output File Name Attached -
>      > ShapeDetectionLevelSetFilterOutput4.png*
>      > > >
>      > > > *2. BrainProtonDensitySlice.png with Seed Point - (81,114) and
>      > > rest of
>      > > > the parameters *have same values as above.
>      > > >
>      > > > Problem - Result of SigmoidFilter is Blank Image while output
>     ! of
>      > > > ShapeDetectionLevelSet Filter for both images is something weird.
>      > > > (attached files)
>      > > >
>      > > > Any input on this appreciated.
>      > > >
>      > > > Thanks,
>      > > >
>      > > > Neha
>      > > >
>      > > >*> > >
>      > >
>      > >
>      > > > ATTACHMENT part 2 image/png name=T1.png
>      > >
>      > >
>      > > > ATTACHMENT part 3 image/png name=PD.png
>      > >
>      > > __________________________________________________
>      > > Do You Yahoo!?
>      > > Tired of spam? Yahoo! Mail has the best spam protection around
>      > > http://mail.yahoo.com*>
>      > >
>      >
>     ------------------------------------------------------------------------
>      > >
>      >
>      >
>      >
>      > > ATTACHMENT part 2 image/png name=tumor57Segmented.png
>      > *
>      >
>      > * *
>      >
>      >
>     ------------------------------------------------------------------------
>      > *Do you Yahoo!?
>      > Yahoo! Mail
>      >
>      > - You care about security. So do we. *
>      >
>      > *
>      > *
>      >
>     ------------------------------------------------------------------------
>      > *
>      > _______________________________________________
>      > Insight-users mailing list
>      > Insight-users at itk.org
>      > http://www.itk.org/mailman/listinfo/insight-users
>      > *
> 
> 
>     *
> 
> * *
> 
> ------------------------------------------------------------------------
> *Do you Yahoo!?
> Read only the mail you want - Yahoo! Mail SpamGuard 
> <http://us.rd.yahoo.com/mail_us/taglines/spamguard/*http://promotions.yahoo.com/new_mail/static/protection.html>. 
> *






More information about the Insight-users mailing list