Recurring Meeting of Cleveland Clinic Core Team

Date: April 21, 2015

Time: 10:30 AM EST

Means: In person meeting


  1. Ahmet Erdemir
  2. Craig Bennetts
  3. Snehal Chokhandre
  4. Robb Colbrunn
  5. Tara Bonner


  1. Discuss immediate action items from the last meeting.
  2. Discuss large-scale analysis of simulation parameters for FE model convergence.
  3. Decide immediate action items for the next meeting.
  4. Other.

Immediate Action Items:

Ongoing Action Items:


  1. Discuss immediate action items from the last meeting.
    • Ahmet started listing ongoing action items in meeting minutes.
    • Snehal sent scheduling of specimen imaging to Tara for her to reserve the robotics facility. She needs to start the wiki page for cruciate ligament modeling and work on a modeling problem to evaluate probe-based tissue thickness measurements. She e-mailed Ahmet a draft document to send Biomomentum, Inc. about Mach-1 capabilities and feature requests. Ahmet will review it.
    • Team members provided status of ongoing action items:
      • Ahmet converted MR images to NIfTI format and started dissemination. He updated imaging specifications to reflect data conversion and dissemination.
      • Craig focused on Python scripts for large scale analysis of simulation parameters. He needs to add modification of many XML elements. Ahmet and Craig will meet to discuss post-processing of FEBio log file to extract convergence information.
      • Work on anatomical imaging manuscript has not started yet. Ahmet noted that him and Craig need to meet with Carl Winalski to include image based specimen evaluation as part of the manuscript. Ahmet will ask Carl's help in the upcoming Advisory Board meeting.
      • Robb has not started the work on the manuscript for patellofemoral joint testing.
      • Tara put a rough outline of tibiofemoral manuscript in the repository.
  2. Discuss large-scale analysis of simulation parameters for FE model convergence.
    • The group had an in-depth discussion on various simulation parameters where Ahmet provided descriptions of each parameter and their potential values.
      • Convergence tolerances are defined in relative terms (percentages). dtol is for displacement (default = 0.01). etol is for energy (default = 0.1). rtol is for residuals (force); (default = 0 meaning it is not checked); ideally, residuals should be minimized. lstol is for line search. Decreasing tolerances will likely increase solution time. Increasing them may result in solution inaccuracies.
      • Time stepper parameters include dtmax, maximum time step size; dtmin, minimum time step size; max_retries, maximum number of step cutbacks which is likely to be between 10 and 50; opt_iter; a parameter used to decide increases in step size; aggressiveness, for magnitude of step cutback (0 cuts by 10% or 1, which cuts time by half, it is set to 1). Newton or quasi-Newton parameters include max_refs and max_ups. When max_ups is equal to 0 a full Newton step is taken where the stiffness matrix is calculated rather than approximated (as in quasi-Newton).
      • Other parameters include cmax, min_residual (to set a minimum residual indicating that the system should be in equilibrium).
      • Analysis type can be set to dynamic, where inertial forces may help stabilize solution. Increasing mass will likely aid convergence but may result in inaccuracies. This is not an issue for static analysis.
    • The group also discussed model parameters that can be changed to facilitate convergence.
      • Ahmet noted that the parameters should be within physiological ranges. Their values not only affect convergence but will also influence potential interpretation of tissue and joint mechanics. In a sense, changing model parameters will also provide a sensitivity analysis.
      • It may be possible to add springs to stabilize rigid body motions. The values should be large enough to facilitate convergence but should be small to ensure these parameters do not affect joint kinematics/kinetics response. As a general rule, the loads carried by these springs should not exceed more than 5% of the dominant loading magnitude.
    • The team prioritized a set of parameters for the first convergence analysis.
      • Loading scenario will be passive flexion, with the focus on simulation parameters.
      • Convergence tolerances will be left at their default values.
      • max_retries and max_refs parameters will be changed to understand their combined role on convergence speed.
      • Craig will implement configuration files to run a set of simulations to vary max_retries and max_refs.
  3. Decide immediate action items for the next meeting.
    • See Immediate Action Items above.
  4. Other.
    • Ahmet started exploring joint mechanics testing files, including the use of nptdms, a Python library, to access the contents of tdms files. It appears that the "state" variables hold the information describing knee joint kinematics and kinetics. Robb will document the tdms file contents (group and channels) and provide the documentation in the wiki.
    • Ahmet also raised the issue that patellofemoral contact pressure data are in a closed format (.fsx). The data needs to be converted into a human readable form. Ahmet and Robb evaluated the data collection software and identified the possibility to export data in ASCII format (as .csv files). This provides the time history of pressure distribution. Robb will generate these files for readily collected contact pressure data and upload them to the in-house data management server.

RecurringMeetings/2015-04-21 (last edited 2016-05-04 22:09:52 by localhost)