Interpolation of 2 areas far away is failing

[finetjul]

In Slicer, if I paint 2 areas in axial within a few slices of differences and the regions are not overlapping, the interpolation is producing odd result.

It would be good to:

• detect that the regions do not overlap
• if so, give expose an option to decide of the behavior: interpolate (but correctly) vs do not interpolate

[finetjul]

Segmentation-label.zip

[dzenanz]

In this case, both contours are interpreted as being end contours of a "branch", so they are smoothly interpolated to nothing, just in the opposite direction.

I guess an option could be added to detect this case (two non-overlapping contours in consecutive slices). But what to do if one slice has one contour, and the other has more then one and there is no overlap? Interpret it as a branching topology change? Or choose the closest one by center of area, or closest one by distance from edge to edge (which requires a distance field computation)? Those are options which immediately come to mind. Possible to do, but requires some effort, and exposing the options to the user.

[finetjul]

I understand the difficulty.

I think the main issue might simply be the "interpolation to nothing", maybe have an option to disable it or not even expose an option and never interpolate to nothing. I can not see when that could be useful.

[dzenanz]

Take a look at the regression test data, and you should find some instances where it is quite logical.

[dzenanz]

Closing as we discussed this and concluded that this module is not a good solution to the problem @finetjul was having.