zasexton / synthetic-vascular-toolkit Goto Github PK

filter4 array for initiating connections is not of an integer or boolean type which throws errors during computation

radius information is lost in some trees during growth of competing trees in a non-convex domain

When projecting root points to an (LV) surface, two different root points can be projected to the same mesh element causing an infinite loop. Implement a flag to assign one of the two root points to a random adjacent mesh element as a quick fix to resolve this.

will involve careful testing of input meshes

To allow for more effective boolean operations outside of SimVascular, a NURBS implimentation should loft vessels along b-spline centerlines to create implicit models.

create github action for building python packages and testing on windows, linux, macOS

During vessel formation of connections for close-loop vascular network generation, connecting vessels sometimes extend out of the perfusion domain. Potential errors for fabrication and perfusion simulation.

Problem: Handling of non-convex boundary domains requires at least quadratic assignment

• Attempt to simplify the assignment problem to a linear-assignment regime

check the competitive addition function for network_id argument

In left ventricle example, when assigning 24 seed points for 24 different arterial trees, the forest object was unable to finish root placement (most likely due to collision conflicts that could not resolve)

Sister vessels are currently ignored during collision detection for connections. Also connecting vessels do not check with collisions against themselves. These should be added to the cost function formulation.

need to constrain vessel networks meant for connection so that vasculature is self supported. This is a requirement for stereolithography techniques so that additional supports can be avoided and uncured resin can easily be evacuated from the system.

during volume tetdrahedralization. surfaces obtained from marching cubes may be at an insufficient voxel resolution to guarantee that the surface mesh is well-posed for tetrahedralizing.

remove pyvista dependency

need standard benchmarks for speed, memory consumption, performance scaling for core features of svcco routines

look at fast_geodesic.py benchmark file for reference