Handling Missing Slice Data in RT Structures: How Does SlicerRT Interpolate Missing RT Data?

Support
1

Hello Slicer Community,

I’m currently working with RT structure sets (RTSS), and I’ve encountered an interesting issue when using Plastimatch’s convert command to extract RT structures as label images. The issue is that Plastimatch outputs label images where certain slices are missing, particularly when there’s an unequal slice spacing. Here’s the warning I received from Plastimatch:

Warning: Slice spacing of RTSS may be unequal
-2.012 - -17.012 = 15 vs. 2.5

This suggests that there’s a 15mm gap between certain slices, where Plastimatch reports the usual 2.5mm spacing between slices. Matlab confirmed this, as shown below:

Z-coordinates and slice spacing (in mm):
From -37.012 mm to -34.512 mm: Spacing = 2.500 mm
From -34.512 mm to -32.012 mm: Spacing = 2.500 mm
From -32.012 mm to -29.512 mm: Spacing = 2.500 mm
...
From -17.012 mm to -2.012 mm: Spacing = 15.000 mm

This gap (15mm) between -17.012 mm and -2.012 mm results in missing RT structure data for some slices. However, when I load the same RT structure set in Slicer using SlicerRT, the visualization displays the full RT structure data with no missing slices or gaps, which is quite puzzling.

My Questions:

  1. How does SlicerRT handle this missing slice data or unequal slice spacing when visualizing RT structure sets?
  • Does SlicerRT apply any interpolation to the RT structure data in order to fill the gaps caused by unequal slice spacing?
  • If so, what part of the SlicerRT code is responsible for this interpolation or handling of unequal slice spacing? Is there a specific function or algorithm that addresses this issue?
  1. Is there a specific process in Slicer that allows it to visualize the missing slices properly, even when Plastimatch reports unequal spacing?
  2. Could you point me to any documentation or relevant code sections in SlicerRT that explain how the system handles or interpolates RT structure sets with unequal slice spacing or missing slices?
2

RTSS contains a series of 2D contours. These contours are triangulated to a closed surface, and then that surface is converted into a labelmap. Here is the paper about the actual triangulation:

If contours are missing, it simply means that the adjacent contours to triangulate will be farther from each other there than in the rest of the structure.

About the rest of the conversions and the conversion mechanism itself please see these links
https://slicer.readthedocs.io/en/latest/user_guide/image_segmentation.html

3

Thank you so much for the detailed explanation and the helpful links! I’ll take a look at the resources you’ve shared and dive deeper into the code. Appreciate your help,