General questions about Shocker

[edorachlee]

As the first post on this discussion thread, I thought it would be appropriate to ask some general questions about Shocker. Hopefully other people who want to use Shocker will also jump on this thread to create a productive conversation. Questions to follow in comments below!

[edorachlee]

Here are some of my questions so far:

• What is the difference between merge_network and concatenate_network?
• Why are there two different outputs minimum_network.vtk and minimum_network_monoalg.vtk? Will there be a functional difference when attempting to run MonoAlg3D simulations?
• How does Shocker calculate the reference LAT when evaluating the LAT error during optimization?

Also:

• What are some tips and tricks to speed up mesh generation? So far, with my tissue meshes, one Shocker mesh can take up to 32 hours per ventricle.
  • I needed to increase the LAT error tolerance to 100ms and lower l_d to 5000 to get generation time to a couple of hours. However this results in unnaturally straight segments so I am not sure this is the best way to proceed.

Thank you!

Best,
Lucas

[bergolho]

Hi @edorachlee,

Thank for your questions and I am really sorry for the delay in answering your questions. These last few weeks are being very busy for me and I will try to concentrate more time in the Shocker Github on the week of 09/12/24. Although, here are the answers and a few suggestions and tips that might help you generate the Purkinje networks using Shocker.

1. The main difference between the "merge_network" and "concatenate_network" is that the first one is used to generate the fully connected Purkinje network with the His-Bundle+LV+RV.
   The "concatenate_network" script is used to join two different Purkinje networks represented as VTK into one. This is done by adding the points and lines from the first Purkinje network to the second tree.

2. The "minimum_network.vtk" file is the output from the Purkinje network generated by Shocker with the parameters from its data structure inside the CELL_DATA . This file is useful for you to debug the parameters you use to generate the network.
   The "minimum_network_monoalg.vtk" is a file generated with the Purkinje conductivity defined in the CELL_DATA. This value is used by MonoAlg3D to run the Purkinje simulations. If you delete this section from the VTK, the Purkinje conductivity will be initialised with the value defined in the "sigma_purkinje" variable located inside the [assembly_matrix] section from the MonoAlg3D configuration file.

3. The reference LAT is calculated by Shocker using the cable equation with parameters calibrated to Purkinje cells, as can be seen in detail in the paper (https://doi.org/10.1038/s41598-023-38653-1).
   For this procedure you will need to calibrate the conduction velocity of the Purkinje network using the "start_diameter" parameter from the Shocker configuration file. To help with this, I included a new script called "diameter_calibration.py" inside the "/scripts" folder so that you can use to tune the cable diameter based on the target conduction velocity you want for the Purkinje to have inside Shocker. Normally, this value should be in range of 2-4 m/s. Normally, I use 3.5m/s to debug my Purkinje networks first. Maybe by increasing the conduction velocity of the Purkinje network with the diameter parameter you might be able to reach the PMJs that are being connected with straight segments.

4. Regarding your last question about performance, there are few things you can try. First you can use a coarse endocardium surface, instead of a refined one. The complexity of the Shocker method is very attached to the number of points and elements that the endocardium surface has, if you are able to decrease this number by keeping the topology of the surface you can safe execution time.
   Another suggestion is to calibrate the characteristic length parameter (l_d) using a simulation without the PMJs first. This can be achieved by removing or commenting the [pmj] section from your Shocker configuration file. The result will be a generic network covering the endocardium surface, but you can already see how the "l_d" parameter will affect both the execution time and the surface coverage by the Purkinje network.

Please let me know if that help you decrease the computation time.

Best regards,
Lucas