Influence of the Initial Guess on the Estimation of Knee Ligament Parameters via Optimization Procedures

Optimization procedures provide ligament parameters by minimizing the difference between experimental measurements and computational simulations. Literature values are used as initial guesses of ligament parameters for these optimization procedures. However, it remains unknown how these values affect the estimation of ligament parameters. This study evaluates the effects of the initial guess on estimations of ligament parameters. A synthetic data set was generated using a subject-specific knee computational model, reference ligament parameters and simulated laxity tests. Subsequently, ligament parameters were estimated using an optimization routine and four different initial guesses. The distance of these initial guesses from their true values ranged from 0 to 3.5 kN and from 0 to 3.6% for the stiffness and reference strains, respectively. The optimized ligament parameters had an average absolute mean error ranging from 0.15 (0.09) kN and 0.08 (0.04)% to 3.67 (2.46) kN and 1.25 (0.76)%, while the kinematic error remained below 1 mm and 1.2° for all conditions. Our results showed that the estimations of the ligament parameters worsened as the initial guesses moved farther away from their true values. Moreover, the optimization procedure resulted in suboptimal ligament parameters that provided similar behavior to the true laxity behavior, which is an alarming finding that should be further investigated.