EXPERIMENTALLY VALIDATED FINITE ELEMENT MODEL OF A HUMAN TIBIA WITH A UNICOMPARTMENTAL KNEE REPLACEMENT

Abstract

Finite element (FE) analysis is widely used to calculate stresses and strains within human bone in order to improve implant designs. Although validated FE models of the human femur have been created (Lengsfeld et al., 1998), no equivalent yet exists for the tibia. The aim of this study was to create such an FE model, both with and without the tibial component of a knee replacement, and to validate it against experimental Results: A set of reference axes was marked on a cleaned, fresh frozen cadaveric human tibia. Seventeen triaxial stacked strain rosettes were attached along the bone, which was then subjected to nine axial loading conditions, two four-point bending loading conditions, and a torsional loading condition using a materials testing machine (MTS 858). Deflections and strain readings were recorded. Axial loading was repeated after implantation of a knee replacement (medial tibial component, Biomet Oxford Unicompartmental Phase 3). The intact tibia was CT scanned (GE HiSpeed CT/i) and the images used to create a 3D FE mesh. The CT data was also used to map 600 transversely isotropic material properties (Rho, 1996) to individual elements. All experiments were simulated on the FE model. Measured principal strains and displacements were compared to their corresponding FE values using regression analysis.

Experimental results were repeatable (mean coefficients of variation for intact and implanted tibia, 5.3% and 3.9%). They correlated well with those of the FE analysis (R squared = 0.98, 0.97, 0.97, and 0.99 for axial (intact), axial (implanted), bending, torsional loading). For each of the load cases the intersects of the regression lines were small in comparison to the maximum measured strains (< 1.5%). While the model was more rigid than the bone under torsional loading (slope =0.92), the opposite was true for axial (slope = 1.14 (intact) 1.24 (implanted)) and bending (slope = 1.06) loads. This is probably due to a discrepancy in the material properties of the model.