I’m curious if anyone has any suggestions on strategies for segmentation of some specimens we’re working on. These are micro-CT scanned insects (dried, pinned specimens) that have thin cuticular membranes internally that I want to segment out. The membranes are very thin and wind up having absorbances pretty close to the background air. Here’s a representative image as an example:
The red is my segmentation in progress, and you can see how the membrane is very thin and also very close to the background absorbance.
Thus far, my strategy has been to use sphere brush and symmetric scissors with a set intensity range as much as possible, periodically using island removal to “de-speckle”. This results in lots of holes in the reconstructed surface, however, that I wind up going back and cleaning up manually using a brush. This can be somewhat time consuming, and I’m looking for ways to speed up the process.
I’ve tried flood filling and other algorithms a couple of times without a lot of success. Just curious if anyone has run into any algorithms or anything that do any kind of “surface following”, for lack of a better term, or if there are any fancy ITK filters called “Automatic Insect Membrane Determination”. (OK, that last one is a bit of a stretch.)
Crop and resample your volume to make the structures that you want to segment at least 4-5 voxel wide. Otherwise all the region-growing based method will fail, as due to random noise a few voxels may be missed and your structure may break to parts.
Apply filtering, which enhances thin structures. We have found that unsharp masking helped.
You may paint the structure manually on a couple of (non-neighbor) slices and use Fill between slices effect to create a complete segmentation out of it.
These are great ideas, and thanks for the reference.
I’ll report back on the surface ideas (it is indeed a surface we want out of this, not necessarily a volume) to see if it makes sense to make a segment editor effect.
