Sensible structure set resampling for optimization?

[wahln]

Question: When matRad resamples the cst to the dose / optimization grid, it uses nearest neighbour interpolation. When using constraints in optimization, that can be quite problematic for targets and OARs since we can not ensure that the structure is encompassed on the resampled grid. I see that this problem is not generally solvable since we lose information in downsampling, but we could prevent this if we consider the overlap in resampling (i.e. if we have to choose between a OAR and a non-labeled voxel we always choose the OAR?). Any other thoughts on this?

Related Files / Functions / Code: matRad_resizeCstToGrid(...)