This code was written during the course of the thesis "On differentiable simulations of haemodynamic systems" and the corresponding document can be found here. It offers the ability to parse models, run models, and infer parameters.
The parsing scripts can be found in the parsing folder. Models can be parsed from two different formats. The first format being models provided by vascularmodel.com. In order to parse these kinds of models the files representing a 3D-model from vascularmodel.com needs to be placed in the parsing folder. Furthermore an encoding of the network structure done by centre-line extraction needs to be placed in the model folder as well. This network structure can be generated using the Slicer 3D software. In order to parse a vascularmodel network, once the files have been placed appropriately, run
python parse_vm.py <model_name>
The second format that can be parsed is the format used in the openBF-hub repository. In order to parse these models the repository need to be cloned in to the parsing folder. Then the models can be parsed by running
python parse_bf.py <model_name> <model_sub_dir>
where depending on the model the model_sub_dir argument might not be relevant. The parsed models are all stored in the test folder.
The parsed models can be run by executing the following command in the base directory of the jaxFlowSim repository
python run_model.py <model_name>
The results of a successfully simulated model can be found in the folder results/<model_name>.
Parameter inference can be run by executing
python run_inference.py <model_name>
in the base directory as well. We note that the inferring of a parameter in this manner is very experimental and requires that the user hand tunes the file src/inference.py. However by running
python run_inference.py
without an argument the toy problem from the last section of the results chapter in the thesis can be computed.