[ITK-users] Whole body Segmentation-Identification

[Bumpy RG]

Dear All,

I have a dicom series having whole body image sequences, starting from head
up to legs.

My aim is to write an algorithm to automatically identify the body parts
(lungs, spines, head etc.) from the dicom sequence.

Although I am very new to the ITK world, I am following these steps in ITK (*I
would appreciate if you comment whether I am following correct idea or not*
):

   1. Read dicom series using 'DicomSeriesReadImageWrite2.cxx ' to create a
   3D volume and save the result in mhd format.
   2. Use 'ImageReadRegionOfInterestWrite.cxx' for separating various parts
   (like left lung, right lung, spines, head etc.) and save them in respective
   3D volumes in mhd format.
   3. Pick the various body parts prepared above, one by one and segment
   them using robust algorithm (say
   ThresholdSegmentationLevelSetImageFilter.cxx).
   4. Use Mathematical Morphology to get better results.
   5. Using the various results from above, form an algorithm for
   identifying the various body parts.

Currently I am stuck in the second step because it is so hard to choose the
co-ordinates for selecting proper region of interest.

For example,  I chose these points for ROI of right lung (I checked the
co-ordinates using 3D slicer visualizer) and it does not give me good
results :
start[0] = 255;
start[1] =165;
start[2] = 0;

*However I do not understand how to choose the points properly in ITK.*


I would be really glad to have your response.

Thanking you in advance.
Kind Regards,
Bumpy.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/insight-users/attachments/20140801/c80b0e48/attachment.html>