Influence of the Experimental Protocol and the Optimization Method on the Noninvasive Estimation of Knee Ligaments Properties

Subject-specific ligament properties can be obtained by combining optimization procedures with experimental data and musculoskeletal modeling. These properties are computed by minimizing the difference between experimental laxity measurements and the respective simulations from subject-specific models. Different levels of agreement between experimental and simulated laxity have been reported due to the model complexity and the employed experimental protocol, and possibly the kind of the employed optimization method. However, how experimental and optimization choices affect the ligament properties estimations remains unknown, possibly due to lack of ground truth values. Therefore, this study investigates whether experimental protocols and optimization methods influence the properties estimations, using a computationally generated dataset free of experimental errors and inaccuracies. Initially, an experimental data set was generated by employing a subject-specific knee computational model, ground truth ligament properties and laxity tests simulations. Subsequently, ligament properties were estimated using the same computational model and combinations two laxity protocols and two optimization methods. The direct optimization method provided more accurate estimations of ligament parameters compared to simulated annealing while the addition of anterior posterior laxity tests to varus/valgus and internal/external rotation ones did not change the accuracy. Future research should provide the optimal combination of optimization method and laxity measurements that would enable noninvasive, in vivo estimations of knee ligament properties.