This project allows the generation and calibration of cellml (and soon to be more) circulatory system models from an array of module/vessel names and connections. This array is written in a csv file such as test_vessel_array.csv where the entries are detailed as...

[vessel_name,

BC_type             ('vv', 'vp', 'pv', 'pp', pp_wCont, pp_wLocal, nn),

vessel_type         ('heart', 'arterial', 'arterial_simple', 'venous', 'terminal', 'split_junction', 'merge_junction', 2in2out_junction, gas_transport_simple, pulomonary_GE, baroreceptor, chemoreceptor),

inp_vessels         (name of the input vessels.)

out_vessels         (name of the output vessels)
]

IMPORTANT: The order of input and output vessels is important for the heart module. The order must be

inp_vessels: 1:inferior vena cava, 2:superior vena cava, 3:pulmonary vein

out_vessels: 1:aorta, 2:pulmonary artery.

If the pulmonary vessels aren't included, a simple 2 vessel pulmonary system will be used.

NOTE: currently the terminal vessels should only have a BC starting with 'pp'

Process for generating a model and running the parameter identification

Generating a model

First create a {file_prefix}_vessel_array.csv and {file_prefix}_parameters.csv file in resources/. file_prefix's of simple_physiological and physiological are good example files.

Then move to the user_run_files directory and ensure the user_inputs.yaml file and opencor_pythonshell_path.sh file are filled out correctly. For model generation the following must be set

file_prefix: {file_prefix} 
input_param_file: {file_prefix}_parameters.csv

opencor_pythonshell_path=path/to/opencor/pythonshell

then run the following to generate the model

./run_autogeneration.sh

If there were missing parameters in {file_prefix}_parameters.csv a new file named {file_prefix}_parameters_unfinished.csv will be created with the required parameters. This must be filled in and renamed to {file_prefix}_parameters.csv to generate a working model.

Parameter identification

To then run the parameter identification you must fill in the param_id parameters in user_inputs.sh, create a {file_prefix}_params_for_id.csv file and create a json file with the ground truth data. See resources/simple_physiological_params_for_id.csv and resources/simple_physiological_obs_data.json for an example.

./run_param_id.sh

Following a successful parameter id process the model with updated parameters can be generated with

./run_autogeneration_with_id_params.sh

The generated models will be saved in generated_models/

Creating your own modules.

This software is designed so the user can easily make their own modules and couple them with existing modules. The steps are as follows...

1: Either choose an existing '{module_category}_modules.cellml' file to write your module, or if it is a new category of module, create a '{module_category}_modules.cellml' file in 'src/generators/resources/'

2: Put you cellml model into the '{module_category}_modules.cellml' file.

3: create a corresponding module configuration entry into 'module_config.json'. These module declarations detail the variables that can be accessed, the constants that must be defined and the available ports of the module.

4: include your new module into the vessel array file. IMPORTANT: modules that are connected as eachothers inputs and outputs will be coupled together with any ports with corresponding name. i.eif VesselOne has an entrance 'vessel_port' and VesselTwo has in entrance 'vessel_port', they will be coupled with the variables declared in their corresponding 'vessel_port'. One must be careful, when making a new module, that the modules it couples to only has matching port types for the ones that are necessary for coupling.

CHANGES TO BE MADE

Currently "vessels" is used interchangeabley with "modules". This will be changes to use "modules" in all instances, as the project now allows all types of modules, not just vessels.

The connections between terminals and the venous system is hardcoded, as a terminal_venous_connection has to be made to sum up the flows and get an avergae concentration. This needs to be improved.

Tests!!

There is now a test for the autogeneration running. To run the test navigate to user_run_files and do ./run_test_autogeneration.sh

requirements

If the model being generated is a cellml model, OpenCOR must be downloaded and installed from opencor

To install required python packages for this opencors version of python you must do the following...

/path/to/opencor/dir/python/bin/python -m pip install packagename

Required packages for parameter identification

mpi4py scikit-optimize sympy

Required packages for mcmc bayesian identification

emcee corner schwimmbad tqdm statsmodels

Required packages for autogeneration

pandas pyyaml libcellml rdflib

Potential Errors:

IMPORTANT if you get an SSL error you must do the following before the pip install

export LD_LIBRARY_PATH=[OpenCOR]/lib

so that libcrypto.so.3 can be found to load the ssl module.

IMPORTANT intalling mpi4py requires mpi to be available. Therefore, the following lines may be required to install the mpi software on your computer

sudo apt install libopenmpi-dev
sudo apt install libffi7

Windows package instalment

Running on Windows is in development. It can work with some caveats...

To be able to importing mpi4py you may have to do the following:

download MS MPI, install both .mis and SDK.
set up environmental variables

control panel --> advanced system settings --> environmental variables --> add

C:\Program Files\Microsoft MPI\

C:\Program Files (x86)\Microsoft SDKs\MPI\

Running in Windows

For running scripts you cannot use the .sh files, So you have to run the scripts in src/scripts explicitly. e.g for running param id...

First move to C:\path\to\opencor\dir

Then run

.\pythonshell.bat C:\path\to\circulatory\autogen\src\scripts\param_id_run_script.py

License

circulatory_autogen is fully open source and distributed under the very permissive Apache License 2.0. See LICENSE for more information.

hsorby/circulatory_autogen