We currently use the marching_cubes method to get more accurate surface area values. It got me thinking that throat perimeter could/should be done more accurately too.
Find the 3D image containing the two neighboring regions.
Identify the throat voxels shared by the two regions (this is already done in getnet)
Find the perimeter voxels with dt < 2
Perform fast marching on this from a random starting point
The perimeter will be 2x the maximum value since the march goes in both direction and will meet in the middle on the other side
We currently use the marching_cubes method to get more accurate surface area values. It got me thinking that throat perimeter could/should be done more accurately too.