I managed to segment an aorta and a coronary tree with VMTK and Slicer. However for my case I would also like to extract calcifications present in the coronaries. I managed to threshold the calcifications, but from what I got from clinicians the calcifications are not completely surrounded by lumen. So I would need a way to make the calcifications stick to parts of the wall. Is there any way of achieving this? I attached a screenshot of my current segmentation.
This is an interesting project.
Cant´you just “hollow” the vessel segmentation, then “add” the calcification segmentation via “logical operators” ?
thanks for your answer. I tried this but since the hollow algorithm removes symmetrically it will shrink the lumen also at the side where there’s no calcification present or?
Do you have a test dataset available?
Hi, sorry for the late response, I was on vacation.
under this link you will find the dataset I’m currently working on.
I was able to download the data.
Isn´t that quite good already for planning an intervention?
Made the “COMBINED” segment hollow with a thickness of 0.3 mm and enabled the “CALC” segment.
yes I’m also quite happy with the segmentation, but the cardiologists I showed this dataset to told me that the calcifications are too much floating in the coronaries, they need to be stuck closest to the lumen wall, that’s why I started to redo some things and asked myself if I can get around without to much manual work.
First of all, I think you need to rework your calcification (CALC) segmentation a bit.
produces this segmentation:
which may just be a little overdone in size and smoothed too much.
From the literature, the thickness of the coronary artery wall is around 0.75 mm.
Use this as the thickness in the “Hollow” tool With the hollow tool settings of
thickness: 0.75 and
Use current segment as: medial surface
I get this result
with your segmentation which leaves the calcifications sticking or very close to the artery walls.
If you rework the CALC segmentation maybe you can switch to
Use current segment as: inside surface
which should actually be the case here.
… you may also try a combination of volume rendering as shown here:
and adding the calcifications as separate segmentations
May I ask how you segmented the coronary arteries and what threshold you are using to identify the calcifications? Thanks in advance!
I could segment the LAD with this tool (with calcifications outside the lumen). The key part is placing control points as close as possible to its center for such small arteries. That’s tedious still.
Used parameters :
27 control points
Intensity tolerance : 60
Neighbourhood size : 2.0
Tube diameter : 4 mm
The calcifications were segmented with usual tools of the ‘Segment editor’:
Paint : with sphere brush in 3D.
Centerline extraction failed, I suppose because of a very tight stenosis.
N.B : This tool is still experimental and not conventional inside. And it’s one tube at a time.