Segmentation with higher resolution

Development
1

Hi everyone,

I would like to make a segmentation of the eye, with a resolution of 0.1 mm.
For this purpose, I would like to start my segmentation from building spheres from python code.

I tried to start from this code

import SampleData
# Load master volume
sampleDataLogic = SampleData.SampleDataLogic()
masterVolumeNode = sampleDataLogic.downloadMRBrainTumor1()
# Create segmentation
segmentationNode = slicer.vtkMRMLSegmentationNode()
slicer.mrmlScene.AddNode(segmentationNode)
segmentationNode.CreateDefaultDisplayNodes() # only needed for display
segmentationNode.SetReferenceImageGeometryParameterFromVolumeNode(masterVolumeNode)
# Create segment sclera
eye = vtk.vtkSphereSource()
eye.SetCenter(30, 68, -31)
eye.SetRadius(12)
eye.Update()
segmentationNode.AddSegmentFromClosedSurfaceRepresentation(eye.GetOutput(), "1_Sclera", [1.0,1.0,0.0])

That works, but with a resolution of nearly 1mm.

image
image766×385 67.4 KB

Then I tried to crop the volume, increasing the resolution (scaling), with success, following this instructions. However, now I don’t know how to link the sphere code to the segmentation with this higher resolution.

Unfortunately, I am not good at programming with python, I tried to find help with the vtk library and tutorials, without success.
Thanks in advance for your help

2

You did it all right, the only point that was missed that you used the input image resolution as segmentation resolution. The simplest would be to crop&resample the input volume before running your script.

You could create a new volume with 0.1mm spacing and use that as input of segmentationNode.SetReferenceImageGeometryParameterFromVolumeNode to avoid additional step of manual cropping&resampling, but without cropping you would get quite large segmentation volumes (2000-3000 voxels along each axis if you have an image of the entire head) and you may run out of memory and/or segmentation may become slow.

3

It worked :slight_smile: thanks a lot!

1 Like