Synthetic CT from MRI - editing MRI segments to Hounsfield Units

Support
1

Hi all,

I’m trying to generate a synthetic CT from an MRI for use in an MRI-only radiotherapy treatment planning approach (with SlicerRT).

I’ve segmented out bone, air, tissue, and a mock tumour from the MRI - is there a way to convert their voxel values to Hounsfield Units? This is an exploratory approach so I only need those 4 basic features in the sCT; ideally I would like to set the HU for each segment (similar to assigning label values in the ‘Model to LabelMap’ module).

Thanks for your time.

2

You can use the segmentation to change voxels of a volume of the same geometry using numpy array operations, something like this:

volumeNode = getNode('MRHead')
segmentationNode = getNode('Segmentation')

segmentNames = ['bone', 'air', 'tissue', 'tumor']
voxelIntensities = [2000, -1000, 60, 150]

volumeArray = slicer.util.arrayFromVolume(volumeNode)

for segmentName, voxelIntensity in zip(segmentNames, voxelIntensities):
    # Get segment as numpy array
    segmentId = segmentationNode.GetSegmentation().GetSegmentIdBySegmentName(segmentName)
    segmentArray = slicer.util.arrayFromSegmentBinaryLabelmap(segmentationNode, segmentId, volumeNode)
    # Update voxel of the volume
    volumeArray[ segmentArray > 0 ] = voxelIntensity

slicer.util.arrayFromVolumeModified(volumeNode)
1 Like
3

Thanks for the reply, Andras. I’m getting the following error:

Traceback (most recent call last):
File "<console>", line 3, in <module>
TypeError: arrayFromSegmentBinaryLabelmap() takes 2 positional arguments but 3 were given

I’ve tried limiting it to two (of those three) arguments but it errs every time. Any ideas?

4

There are several AI-based algorithms for converting MRI into sCT.

However, to check quality conversion, you can use this module:
https://www.slicer.org/wiki/Documentation/Nightly/Extensions/ImageCompare

Best,
Paolo

5

The code snippet works in recent Slicer Preview Releases. The latest Slicer Stable Release users slightly different syntax.

6

Thanks for the help. I’ve downloaded the most recent preview release but I’m now getting a different error - I’ve attached a screenshot in case the problem is my structuring of the nodes/segments.

image
image1920×1208 149 KB

The code I’m using is:

volumeNode = getNode('101')
segmentationNode = getNode('Segmentation')

segmentNames = ['bone', 'gas', 'tissue', 'mass']
voxelIntensities = [2000, -1000, 60, 150]

for segmentName, voxelIntensity in zip(segmentNames, voxelIntensities):
    segmentId = segmentationNode.GetSegmentation().GetSegmentIdBySegmentName(segmentName)
    segmentArray = slicer.util.arrayFromSegmentBinaryLabelmap(segmentationNode, segmentId, volumeNode)
    volumeArray[ segmentArray > 0 ] = voxelIntensity

slicer.util.arrayFromVolumeModified(volumeNode)
7

It seems that I’ve missed a line. Call volumeArray = slicer.util.arrayFromVolume(volumeNode) after setting volumeNode.

1 Like