0116_H_PULM_ALGS has a model creation error

[chlookaburra]

Description

Moving on to the next Alagille model 0116_H_PULM_ALGS, we are faced with a "[GetSingleLoop] An error occurred processing cell 2" warning while trying to create the model from the segmentations.

Tracing back to SimVascular/SimVascular#1154 I discovered the following: "The PolyDataException exception object returns the triangle where the computation goes into an infinite loop. This is then displayed in red. The user can then loft the segmentations used in the union to get an idea what the problem is and possibly modify the segmentations a bit to fix it."

Posting this issue to start my investigation and logging of findings.

Reproduction

Launch SimVascular 2025.12.21 from Ubuntu 24 computer and try to build a model from the segmentations.

Expected behavior

Create a nice model from the segmentations.

Additional context

No response

Code of Conduct

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

[chlookaburra]

Providing more details here. The terminal says:

[WARNING:GetSingleLoop] Infinite loop detected
[CreateModel] ERROR: The Boolan union has failed.
[CreateModel] ERROR: [GetSingleLoop] An error occured processing cell 2

Consistent with the issue !

[chlookaburra]

Going through this branch by branch, I see that the first error occurs for LPA_13. Note to self to continue from here.

[alisonmarsden]

Just commenting here that these models may have been created with Parasolid originally? Also, we should be careful about changing them since we're also referencing the papers they came from in the associated DOI. Perhaps @wgyang has comments since he was a co-author on that work? Whatever changes we make, we need to make sure the models remain consistent with the underlying image data.

[chlookaburra]

@alisonmarsden Thank you for your comment! Currently, for your and @wgyang 's reference - I am facing the following issues:

1. Each SV project has too many Parasolid files; it is very difficult to pinpoint which model was used in the paper.
2. I thought of using the existing mesh / centerline data in the Meshes folder, but the naming conventions are different and inconsistent with the current 0D centerline extracted information (e.g., BifurcationId, etc). To mitigate this, we need polydata .vtp files (and yes they can be converted from the parasolid models but which one? I also tried to download the simulation results and compare those meshes but that led me down another rabbit hole...)
3. Another issue is with SimVascular versions - I have been using the 2023.05.31 version for parasolid since there is no plugin in the recent 2025.12.21 release for Ubuntu 24. However Dave told me that many modeling bugs have been fixed in the latest release.

Dave suggested that the best way to go about this is to go through the modeling process using the preexisting segmentations because at least that way we know that all the data is correct, so that is what I have been up to recently.

[alisonmarsden]

This makes sense to me, except that it's possible that not all segmentations that were saved in the project were actually used for final model creation. Sometimes a user will create many segmentations, but then only use some of them for final model building. I'm not sure how we should tell which ones were used. Perhaps @gemacabero or @wgyang could be helpful here.

[chlookaburra]

@alisonmarsden I will take this into account! I hope to hear from @gemacabero or @wgyang soon.

For everyone's reference - I am trying to streamline this process a bit, so I created a folder called "VMR_cleanup" in the lab drive, which contains an Excel sheet with pulmonary model information. Dave (thank you!) already started contributing to it by adding his corrected segmentations to the "fixed_models" subfolder.

I started posting issues here to share what I am doing in case anyone wants to jump in, so thank you for responding!

[chlookaburra]

Ok I just looked into this more and we can see a list of segments that were used by clicking on one of the Parasolid models, which opens the "SV Modeling" tab in the SimVascular GUI. Note I'm using an older version of SimVascular 2023.05.31 to look at these models.

Having gone through this process, I understand the SV project folder for 0116_H_PULM_ALGS a bit more now.

I think this is what happened:

1. The user first created SU0238_mm.mdl using something called Legacy Segmentations, which include segments such as RPA55 that do not exist in the later versions.
2. Then the user created SU0238_mm_stent.mdl using the segments stored in Legacy Segmentations, and these include the "_stented" segments
3. Then a new set of segmentations were created and saved as SU0238_v2_mm.mdl
4. Then, the user realized the units were in mm, so they changed to cgs (very likely using Dave's scaling script that he showed me). This is now saved in SU0238_v2_cgs.xmt_txt
5. Then the user decided to do some blending which is saved in SU0238_v2_cgs_blend_pt03cm.xmt_txt which also has SU0238_v2_cgs_blend_pt03cm.mss and a suspiciously named (typo?) SU0238_o03cm.blends
6. This is consistent with the mesh-surfaces contained in the Meshes folder.

So this means that likely the "newest" model is SU0238_v2_mm.mdl and we need to decide if we want to use the blended model or not?

Does this sound about right, @wgyang ?

Also, is there a nice way to cross-check the existing segmentations vs the ones used in the model itself? I have been doing it manually.

[alisonmarsden]

I believe the blends file contains the radius based blending values that are used to blend the model - the GUI used to require a value for every intersection of two vessels so for a model with many vessels this could be a long list of values. They define the degree of blending between each pair of two vessel surfaces - this is a user input during model building and it's saved to a file / can be scripted and read in.

[ktbolt]

@chlookaburra The .mdl file contains a list of the segmentations used for that model.

You can modify those files for PolyData by replacing Parasolid with PolyData in the .mdl file; SV will then create a Model Tool instance with the segmentations for that .mdl file.

[chlookaburra]

@ktbolt welcome back!

Update - Claude, ChatGPT, @jeffbli and I created a script that converts a mesh.exterior.vtp to an .mdl file. Centerlines generate successfully for 0114_H_PULM_ALGS, 0115_H_PULM_ALGS, 0116_H_PULM_ALGS, and 0117_H_PULM_ALGS ! I think these four models should be ok. I put them in the lab drive in case you would like to take a look!

[chlookaburra]

Since 0090_H_PULM_ALGS does not have a publication associated with it, and the centerlines fail to generate for preexisting mesh (mesh.exterior.vtp), I will use your modified file for that one - still need to fix the mesh though.

0089_H_PULM_ALGS is fine as is.

That should be all 6 Alagille models in the VMR (pre-op).

[ktbolt]

@chlookaburra Good progress with the help of your artificial friends !

Please put this wondrous script that converts a mesh.exterior.vtp to an .mdl file someplace where others can find it; maybe in https://github.com/SimVascular/Tools.

[chlookaburra]

@ktbolt PR ready for you !