README

Installation

You will need openCARP installed.

Firstly, compile the stan programs in the directory "stanmodels" by calling python <filename>.py. Then move the resulting .pkl files to the mmserp/data directory.

Then install with python setup.py install from the top-level folder.

Dependancies should reveal themselves when attempting to run code.

Using this code

For convenience, this project is organized via scripts, which should be accesible from the Python environment after installation. Most scripts can be called as mmserp_scriptName --help to see the available input arguments. The rest of this README explains how to use these scripts.

Mesh to data storage

HDF5 files are use to keep everything organized.

• mmserp_meshToHDF5 - save a mesh into a new HDF5 file
• mmserp_browseHDF5 - graphically browse the HDF5 file
• mmserp_duplicateHDF5 - create a new HDF5 file, containing only mesh related data, from an existing HDF5 file (requires that the eigenproblem has been solved)

Solving the Laplacian eigenproblem

The calibration utilizes Gaussian Process Manifold Interpolation (GPMI)

• mmserp_viewMesh - plot the mesh colored by Universal Atrial Coordinates (UACs)
• mmserp_decimateMesh - create a lower resolution mesh, and solve Laplacian eigenproblem
  • use --loc, --num, and --runs to determine vertices of lower resolution mesh
  • use --tri to triangulate these vertices
  • use --holes, --layers, --eigs, --numeigs to solve eigenproblem
• mmserp_viewEigs - plot the mesh colored by a specified eigenvector

Running atrial simulations

• mmserp_generateFields - generate and store random spatially-correlated parameter fields
• mmserp_createStimulus - define a set of vertices to use for stimulus in simulations, and save them with a specific name
• mmserp_createCARPfiles - create a directory of files for running the CARP simulations
• runCARP.sh - run simulations from the directory created above
• mmserp_getSimResults - get simulation results from the directory created above

Calibrating parameter fields

• mmserp_inference - perform inference using a virtual experimental design
  • use --params to specify the name of parameters as specified in the HDF5 file
  • use --obs and --num to create and save a design of observation locations
  • use --deltaS2 to define the resolution of the S1S2(S3) pacing
  • use --inference and --basis to do MCMC inference of the parameter fields
  • use --plotGroundTruth, --plotAPD (with --pacesite), and --plotInferred for plotting