Evaluation Physics Based Shape Matching (PBSM) approach

This repository contains data, source, and results from using the physics-based shape matching or PBSM algorithm on three different experiments.

Data

Suwelack S, Röhl S, Bodenstedt S, Reichard D, Dillmann R, Santos T, Maier‐Hein L, Wagner M, Wünscher J, Kenngott H, Müller BP. Physics‐based shape matching for intraoperative image guidance. Medical physics. 2014 Nov 1;41(11). Data from the Open-CAS: Validating and Benchmarking Computer Assisted Surgery.

PBSM approach

The employed PBSM approach is a SOFA based implementation of the PBSM approach. The implementation uses the Medical Simulation Markup Language (MSML) and is available as a docker container.

Install docker container

The PBSM approach used for the evaluation is available as a Docker container.

1. Install the Docker framework for your Linux distribution. Further details can be found on the Docker homepage. For Ubuntu Linux there is the possibility to use either pre-configured packages or run a manual installation
2. To avoid the use of sudo in front of the docker commands, you may grant access to a user using: $ sudo usermod −aG docker username (Log out for the changes to take effect)
3. Pull the SOFA container from the Docker Hub using: $ docker pull ssuwelack/msml_sofa or clone it from Github using $ git clone https://github.com/ssuwelack/msml−docker−runtime.git.

Run container

There are two different options to run a graphical application inside the simulation container. Either start an SSH server inside the container and connect to it, or link the host X-server into the container. Both start-up procedures are available as shell scripts from the aforementioned Github repository, e.g. $ ./ run_msml_sofa.sh. When prompted for a user and password, use msml for both. For linux clients, you may also run the container as a server and connect to it with an instance using $ ssh −XC msml@servername −p portnumber. The recommended approach is communicating over SSH. It is described in four steps:

1. $ docker run −d −p 127.0.0.1:22000:22 −−name msml_sofa ssuwelack/msml_sofa/root/start_ssh.sh
2. $ ./ start_ssh_msml_sofa.sh
3. $ ssh −XC msml@localhost −p 22000 (alternative usage using the user/pwd msml)
4. The docker name is msml_sofa and is responsive to docker start/stop/rm commands.

PBSM using MSML

The interfacing is Python and XML based. In the docker, the python script example can be found in /examples/PythonExamples/LiverShapeMatching/Shape- MatchingRunner.py In order to run this example from the command line, the MSML source should be in the Python path: $ export PYTHONPATH="$PYTHONPATH:/opt/msml/src. Then move inside the folder that contains the script $ cd /opt/msml/examples/PythonExamples/LiverShapeMatching and run $ python ShapeMatchingRunner.py. Example .py and .xml files are provided for the Delaunay triangulation and coupled optimization mesh in the src folder.

Evaluation

Once the volume is registered to the target surface. The surface mesh of the deformed volume is extracted (VTK/Paraview). It is then voxelized using Binvox creating a grid size of 1024^3. Example run: ./binvox -d 1024 -t vtk path-to-stl-file

The voxelized volumes are compared against the full surface registration using Visceral. Example run: ./EvaluateGrids path-to-deformed-ground-truth-voxelized-vtk-file path-to-deformed-test-voxelized-vtk-file -use qll -unit millimiter/voxel > output-path.txt Both executables are reported in the source folder src. All results are reported in the res folder in each setting (either in silico or phantom experiment).

Paper

If you adapt, remix, transform, or build upon the material, please cite the published paper: Hattab, G., Riediger, C., Weitz, J. et al. A case study: impact of target surface mesh size and mesh quality on volume-to-surface registration performance in hepatic soft tissue navigation. Int J CARS (2020). doi.org/10.1007/s11548-020-02123-0