Complementing femur model validation with a variability-focused approach

Abstract

Objective

This short communication presents an approach as an objective means to validate that population variability is potentially incorporated into human body models in an accurate way, complementing existing validation techniques based on individual experiment-simulation comparison. This shall provide a further option for the assessment of the quality of large-number statistical simulations with human body models regarding their biofidelic behavior.

Methods

This population-based approach uses mathematical clustering methods to group similar curves of a combined population of numerical simulation results and experimental curves together. The resulting clusters can be used to assess the biofidelic behavior of numerical simulations, also with characteristics substantially differing from the experimental objects. This developed population-based approach was tested on a reference load case, the dynamic 3-point bending of the femur (Forman et al. 2012).

Results

The clustering approach rendered a distinction into 4 groups of response curves. For this small number, the grouping can be manually assessed as plausible. All experimental, and most numerical responses were grouped into one cluster. Three result curves constitute a cluster of their own, with their meta-data ranking on the margins of the population in at least one of the crucial biomechanical parameters. Such a result can be considered in accordance with the included experimental and anthropometric data.

Conclusions

The feasibility of using such a cluster analysis without individual comparisons is demonstrated on a small set of results. It is used to judge whether a finite element model including aspects of the variation in a population is in agreement with experimental and anthropometric data. For experiments as the femur bending addressed here, it is of high importance to firstly ensure a gross match of curve shapes between experiments and simulation, i.e., capturing the relevant biomechanical aspects.

Keywords:

• Stochastic modeling
• human body model
• validation approaches
• FE-modeling