Date: December 18, 2018
Time: 4:00 PM US/Eastern
- Carl Imhauser (HSS)
- Shady Elmasry (HSS)
- Kevin Shelburne (DU)
- Peter Laz (DU)
- Donald Hume (DU)
- Thor Besier (ABI)
- Nynke Rooks (ABI)
- Ahmet Erdemir (CC)
This is a free form meeting primarily focusing on planning of the Model Calibration phase.
Immediate Action Items:
- Communicate with Jason to follow up with Jeff Bischoff and Nico Verdonschot on ORS workshop.
- E-mail modeling teams when Model Calibration wiki page is complete and earmarked data packages are completed.
- Follow up with BJ Fregly on ORS workshop.
- Announce ORS workshop in Biomch-L.
- Send Jason reviews of assigned papers for reproducibility in knee modeling manuscript.
- Ahmet opened up the discussions by providing a summary of decisions that needed to be made to pursue the Model Calibration phase. These decisions are related to which data sets to earmark for calibration and in what form, the flexibility to use all or some of the data earmarked for calibration, and what to do about the patellofemoral joint. The following discussions focused on Ahmet describing his intentions in each of these areas and feedback from the participating members.
- Ahmet went over disseminated data from Natural Knee Data and from Open Knee(s). He anticipates that calibration data for DU02 will include joint kinematics-kinetics data collected during passive flexion, anterior-posterior laxity, internal-external rotation laxity, and varus-valgus laxity (collected for flexion angles up to 120 degrees). Additional data are available for extension (intact and after ligament resection), which can be used for benchmarking. For Open Knee(s), oks003, data earmarked for calibration will include joint-kinematics-kinetics data during passive flexion, anterior-posterior laxity, internal-external rotation laxity, and varus-valgus laxity (collected for flexion angles up to 90 degrees). Additional data are available for combined loading, which can be used for benchmarking. The participants agreed to the extent of data for calibration.
- Both Natural Knee Data and Open Knee(s) have joints mechanics data that will be relevant to modeling and simulation of the patellofemoral joint. Only one model reuse case, sit-to-stand, requires modeling of the patellofemoral joint. Nonetheless, all teams are including the extensor mechanism in their models. Ahmet is not in favor of calibration and benchmarking activities specific to the patellofemoral joint as these may increase burden on the teams. He noted the possibility to return back to patellofemoral joint specific calibration and benchmarking as a follow up, when model reuse cases indicate potential issues. This strategy will be more reflective of the model lifecycle. None of the participants had raised a concern with this decision.
- Ahmet was wondering about processing of joint mechanics data sets before dissemination to facilitate their use. He was inclined to be faithful to original dissemination, while focusing on data in its raw (less processed) form to give the opportunity to modeling teams for their own interpretation of existing data sets. Mechanics data for DU02 are provided in a processed form in spreadsheets but also in a less processed form as CSV files, all describing six degrees of freedom joint kinematics-kinetics in an anatomical coordinate system. Ahmet intents to provide CSV files as these are more comprehensive. Kevin mentioned the possibility to provide marker data sets directly. Ahmet indicated that this may not be necessary and noted that these are not provided in the original dissemination anyways. Nonetheless, he raised the concern that documentation on the coordinate systems that are used for reporting in CSV files will likely be necessary. Mechanics data for oks002 are provided as TDMS files (can be read with Python or Matlab) with probing data provided in text files. TDMS files contain joint kinematics-kinetics in an anatomical coordinate system; tibia an femur marker data are also available in these files. Text files contain transformation matrices and probing data that can be used for registration. Documentation on file contents are available. The participants agreed that starting with disseminated data (as raw as possible) is a reasonable idea. It may be a lot of work but it is part of the challenge.
- Ahmet described his intentions to provide the opportunity for the teams to use all or some of the calibration data. The teams may want to use only certain portions of the data, as this decision may fit to their workflow. It may also be adequate for calibration. The risk is the subset of earmarked data used by each team may be different. The interpretation of the data may vary as well (see previous discussion). Nynke raised concerns that if model outcomes don't match, we may not know if this is because of the calibration process or the subset of data selected (and interpreted) for calibration. Ahmet noted that this is a valid concern but he is curious if this will be the case. If models and the processed calibration data are provided by each team, Nynke (or other interested parties) can take each team's models and calibrate with exactly the same data to see if data interpretation is the source of uncertainty. Carl mentioned that an approach providing the flexibility to use subsets of data (and the team's own interpretation) will provide the opportunity to identify error source but not isolate the specific errors, which can be left to a follow up project. Ahmet emphasized that while each team may use different subset of calibration data, e.g. passive flexion vs laxity, for fitting model parameters, all teams should simulate the loading scenarios of the full calibration data set (including those that are not used for fitting) to allow comparisons among groups.
Carl provided an update on ORS workshop efforts. The conference organizers asked for a three page handout. The panelists had two meetings so far and discussed developing this handout to formulate the talks. The modeling teams can help with the outreach to maximize participation. Carl will put an announcement in Biomch-L for broad distribution besides word of mouth. Ahmet reminded Carl to pay attention to potential restrictions on advertising the workshop due to panelists from industry and regulatory bodies. Carl will pass the announcements to Cheryl and Andrew before sending. The workshop focuses on perspectives from academy, industry and regulatory agencies. There will be an open Q&A session. Carl plans to invite people to be part of Q&A to establish and interactive audence. The team discussed a few names and agreed to follow up with Jeff Bischoff, Nico Verdonschot, and BJ Fregly. Ahmet and Jason will discuss to follow up with Jeff and Nico. Thor will talk to BJ. This group can form a flash opinion (5 min. talk) session before Q&A to liven up the discussions.
- Ahmet reminded the participants to send their review to Jason for the manuscript he has been working on. He also asked if the funding administration has been working well. The teams did not have any issues in this regard. Ahmet closed the meeting by informing the participants to look for an e-mail after Model Calibration wiki page is done.