Bug report: centerline generation fails with the pulmonary model 0082_H_PULM_H

[chlookaburra]

Description

@ktbolt Following up from our discussion - currently I have been using SimVascular 2023.05.31 to extract centerlines for a series of pulmonary models (work in progress - may or may not run into more issues later on).

For 0082_H_PULM_H, the GUI throws an error saying "Centerline consists of more than one region" when I try to create 0D simulation files from the ROMSimulations tab. I have attached a screenshot of what the centerlines looks like (not very close to the inlet face).

Reproduction

1. Download 0082_H_PULM_H model from the VMR.
2. Open the project in SimVascular.
3. Go to the ROMSimulations tab.
4. Select Inlet Face to be the "inflow" face.
5. Click Calculate Centerlines.
6. Select any BC for all the faces except the "inflow" one (for that, use prescribed velocities). I just set everything to be resistance (and a random number) except the "inflow" for which I just used the inflow_1d.flow file from flow-files in the project folder.
7. Choose any solver parameter for now (e.g., 100 time steps, 0.01 time step size).
8. Click on "Create Files for Simulation" and it will throw the error (see screenshot)

.

Expected behavior

I was hoping to extract a zero-dimensional file for this model, but the process failed for this particular pulmonary model. I am wondering why the GUI thinks that the centerline consists of more than one region.

(I have not tested this with the most recent release - I have been using the 2023.05.31 version because I sometimes need to open parasolid models.)

Additional context

No response

Code of Conduct

• I agree to follow this project's Code of Conduct and Contributing Guidelines

[ktbolt]

@chlookaburra Extracting centerlines I see that the Calculating CenterlineSections code silently fails with a message printed to the terminal

NumSources 1
NumTargets 82
Calculating CenterlineSections...
Input centerline consists of more than one piece. GenerateCleanCenterline failed!

This failure appears to be caused by an initial centerlines computation by VMTK produces a disjoint geometry, maybe from a poor model that is not manifold.

I tried to build a new model but that failed; probably some poor segmentations somewhere, I'll have a look.

[ktbolt]

The problem was with the LPA_21 segmentation

where the segmentation (red) osculates the lofted surface (yellow).

I reduced the LPA_21 segmentation a bit and now the model builds.

The centerline computation now completes but it is producing incorrect geometry

[ktbolt]

@chlookaburra I experimented a bit with centerline extraction; was failing for smaller models with only a few RPA vessels.

The centerline computation is failing to create correct geometry because of the poor surface triangulation at the caps

Using the Model Tool Remesh option set to 0.1 produces a righteous surface triangulation at the caps

It also reduces the size of the surface mesh which speeds up the centerline computation which was taking two hours with the original model surface mesh.

Using the model with the better surface the ROM Simulation Tool can now create pleasing centerline geometry and a 0D solver input file can be fearlessly written.

[chlookaburra]

@ktbolt Wow thank you!!! I will proceed to fearlessly write the 0D solver input file and report back here!

[ktbolt]

@chlookaburra I will need to update the VMR with the modified segmentation and model; will work with @natalia-rubio to figure out how to do that.