I have a few questions regarding the use of a statistical shape model (SSM) of the tibia within the AnyBody Modeling System, particularly with respect to exporting finite element (FE) model inputs using the ConvertToAbq pipeline. Specifically, I am interested in understanding how variations in tibial geometry might influence tibial strain. My plan is to replace the subject-specific (or scaled) tibia with both an average SSM tibia and perturbed models (up to ±2 standard deviations). I would be obliged of you kindly share your feedback on this:
How the AnyBody model environment accommodates such geometric changes, and whether the system can provide reasonably adjusted joint loads, muscle forces, and insertion points required for FE input generation. I will run the inverse dynamics with same kinematics and GRF data, and the model will be same except the perturbed tibia geometry.
While I understand that considering the same load (predicted with subject-specific tibia) allows for shape-only effect analysis, I am particularly interested in whether the inverse dynamics can realistically redistribute loads when a perturbed geometry is used—thus creating a more physiologically relevant scenario.
If the model successfully runs with a 2 SD perturbed tibia, how can I assess the reliability or validity of the recalculated forces and points of application (e.g., muscle and ligament insertions)? Are there recommended approaches or metrics to evaluate the realism and quality of these outputs?
I would greatly appreciate your insights on the feasibility of this approach, and whether using perturbed geometries in this way can yield realistic and reliable inputs for downstream FE analysis.
It is, of course, possible to use SSM bones to morph human body parts automatically and do what you plan. You can certainly relate parameters of SSM to "individualized" loads in the human body.
When using the same GRF and kinematics - please be aware, that the shape of the tibia might become inconsistent with the movement. Just imagine the tibia becoming longer and the previous joint locations are now inside the bone, not where they should be. This aspect needs to be thought through. You may want to reoptimize the movement to the existing markers after deforming the bones. Secondly, if you do pure inverse dynamics the articular surfaces will not be fully taken into account, as they may give slightly different joint configurations. A more sophisticated analysis might be needed if you plan to process complex motion, where the joint behaviour is shape dependent. But it all very much depends on the details of what you want to do. In certain situation it might be okay.
It will redistribute the loads as realistic as how realistic you will construct your "patient-specific" model. The muscle insertion sites will follow the SSM morphing. The joint definition might need to become surface dependent. For example, in our repository we have a knee axis behavior that follows epicondyle landmarks, which should follow morphing quite well (if morphing is correct).
The best way to assess validity is to look at the corresponding experimentally or clinically measured metrics. But quite often this is not feasible and researchers look at available literature to assess the credibility of prediction. Or they follow methodology of V&V40 to establish credibility of the model. This involves a series of verification tests and/or validation studies. For example, you can make analytical estimations of certain configurations and compare with the predicted results to make sure that the model performs as expected. You can check here how different people approached this question.
Please check our publication list and recorded webinars - you may find similar studies and learn how your questions were addressed in those studies.
Thank you very much for your detailed response—I truly appreciate your insights.
I have a few follow-up questions to better understand the process and clarify some of the points you raised:
I fully understand the need for re-optimization to accommodate the perturbed SSM tibia. However, I would appreciate some elaboration on your second point:
My overall goal is to extract the joint reaction forces and muscle forces associated with the tibia when using an SSM-based geometry. Additionally, I plan to export the related inputs for FEA using the ConvertToAbq pipeline to estimate tibial strain, with the broader aim of investigating stress fracture risk. When you mention “sophisticated analysis,” are you referring to a more detailed modeling of the articular contact surfaces at the knee and ankle joints? The movements I will simulate—primarily treadmill running, and walking with load carriage—are relatively standard. My primary interest lies in predicting peak tibial strain, rather than performing an in-depth contact analysis of joint surfaces. Given this context, would you still recommend incorporating a more complex joint model, or is the standard inverse dynamics analysis sufficient for estimating realistic joint and muscle forces (and relevant FEA inputs) in this case?
Could you also elaborate on the following point?
Do you have any specific recommendations or resources on how to define such surface-dependent joints in AnyBody, particularly for applications like mine?
My initial aim was to assess the feasibility of this modeling approach from a computational perspective, particularly in light of the assumptions and constraints inherent in the AnyBody framework. I’ll certainly review the literature you recommended to explore validation strategies in more detail.
Thank you again for your time and support. I’d be very grateful for any additional feedback or guidance you can provide on the above points.
When you mention “sophisticated analysis,” are you referring to a more detailed modeling of the articular contact surfaces at the knee and ankle joints?
3. Could you also elaborate on the following point?
Yes, in a regular gait pattern the knee joint does behave more like a hinge joint and this assumption is acceptable because it provides realistic kinematics. But for movements with higher RoM like squat - it may need a more sophisticated modelling mechanism for the articulation (i.e., a surface-to-surface contact analysis or multiaxial knee joint definition). For surface-dependent joint modeling, we have a section on Force-Dependent Kinematics in our tutorials and some associated template models, please have a look. But it does not sound like this should be your first choice.
I believe what you plan to do is feasible : 1) morphing is possible and can be automated; 2) the MoCap framework will help you process existing recording. There is only a slight concern on what happens when your tibia is inconsistent (say, gets too large) with the MoCap recording. But this may be an important finding in itself.