Author:

Abstract:

This thesis contributes to subject-specific modeling in biomechanical an alysis by (i) designing an experimental setup to obtain a more accurate subject-specific angle-moment relationship of the knee joint (chapters 4 and 5), (ii) developing an algorithm for the estimation of the muscle-t endon parameters of the actuators of the knee joint in a simulation envi ronment and comparing its performance to the performance of the algorith m of Garner and Pandy (2003), and (iii) validating the outcomes of the a lgorithm using forward and inverse simulation techniques for different t ypes of movement based on two case studies. Chapter 2: Muscle-tendon modeling and parameter estimation This chapter explains the general concepts of muscle-tendon modeling and parameter estimation, in order to provide the required background for t he remainder of this thesis. Furthermore, the relation between the remai ning chapters is explained. Chapter 3: Sensitivity of dynamic simulations of gait and dynamometer experiments to hill muscle model parameters of knee flexors and extenso rs This chapter presents the results of a sensitivity analysis which relies on moment-angle relationships obtained by dynamometry. Dynamometry is a useful experimental setup, because muscle group specific angle-moment r elationships can be obtained in a controlled way per individual. The stu dy revealed that dynamometer experiments contain information on muscle-t endon (MT-) parameters. Also, an hierarchy is found in the MT-parameters . As the hierarchy was equivalent for dynamometry and dynamic simulation s of gait, dynamometry can be used to obtain experimental data in order to identify the most sensitive parameters to enhance the accuracy of the simulations(further explained in chapters 6 and 7).This chapter is based on my master thesis [29], and hence it is not a contribution of th is thesis. However, it is the foundation for the work presented in the n ext chapters. Chapter 4: Functional knee axis based on isokinetic dynamometry data: Comparison of two methods, MRI validation, and effect on knee joint kin ematics This chapter presents the first step towards obtaining more accurate dyn amometer data. Experimental dynamometry typically faces problems when it comes to data accuracy. These problems result from the assumption that the moment generated around the joint axis of rotation corresponds to th e measured moment around the axis of rotation of the dynamometer. This a ssumption does not hold for following two reasons: (i) the fixation betw een the dynamometer device and the body segments is not rigid, and (ii) the pose of the joint axis of rotation is not known. The consequences of the former are that the joint axis of rotation moves relative to the dy namometer axis of rotation, and the pose of the joint axis of rotation i s not known in time. The overall consequence is that the registered mome nt is not a good representation of the joint moment generated by the mus cles around the joint axis of rotation and hence, the strength of the ac tuators. In reality, the knee joint axis of rotation is an instantaneous axis i.e. its pose depends on the segmental kinematics and on the load (amount of force produced by the muscles and the external load). It is h owever not possible to estimate the joint axis of rotation instanteneous ly using conventional measurement techniques: kinematic data of the segm ents (being the postion of the tibia relative to the femur) are typicall y obtained by skin mounted markers which implies that the data are noisy due to soft tissue artefacts. Alternatively, the joint axis of rotation can be determined based on bone geometry or for a certain range of moti on. The contribution in this thesis is that the validity of geometry-based a xes and functional axes (which are motion-based axes) is veri ed. Many a lgorithms have already been presented for the estimation of functional k nee axes of rotation, none of them have been validated on real data. The refore, the best performing sphere fitting algorithmand axis transformat ion algorithm according to simulation studies are applied on real data. The validation relies on the comparison with the pose of equivalent axes which are calculated based on magnetic resonance images (MRI) of the kn ee in different knee flexion angles and hence, directly reflect the posi tion of the bones. Chapter 5: An extended dynamometer set-up to improve the accuracy of knee joint moment assessment This chapter describes a new dynamometer setup which allows us to perfor m a full three dimensional (3D) inverse dynamic analysis resulting in an improvement of the accuracy of the experimental data for knee joint dyn amics. The contributions in this thesis are twofold. First, the introduction of the combination of 3D motion tracking and the 3D external forces and mo ments registration allows us to perform the inverse dynamic analysis. Se cond, the limb model contains a knee joint axis of rotation de ned as a functional axis of rotation instead of a geometry-based axis of rotation . Chapter 6: A new method for estimation of subject-specific muscle-ten don parameters of the knee joint actuators: a simulation study This chapter describes the identifiability of the muscle-tendon paramete rs and a new estimation procedure to obtain two subject-specific muscle- tendon parameters per muscle being the optimal muscle fiber length and t he tendon slack length. Here, the focus is on the subject-specific defin ition of the features of the actuators of the knee joint in a simulation environment. By optimal experimental design, a trade-off between determ inistically chosen experimental sets is made. To this end, the experimen tal cost has to be evaluated in light of the information on the actuator s contained in the experiments. In addition, the use of different transf ormations of the muscle-tendon parameters as variables to be estimated i s evaluated. The estimation procedure is two-phased. In phase I, the fea sible set and the initial guess for the non-linear optimisation problem in phase II is de ned. In phase II, a constrained non-linearproblem is s olved by fitting simulated moments and synthetically generated joint mom ents (simulation of dynamometer experiments). An important feature of th e estimation procedure is that the operating range of the muscles is pre served in combination with the use of subject-specific strength-informat ion. The operating ranges are obtained from literature. The strength inf ormation is obtained via (simulated) dynamometry experiments. The influe nce of the initial guess and measurement noise is quantified. The perfor mance of the new algorithm and the algorithm presented by Garner and Pan dy [42] are evaluated and compared. Following contributions are made: (i) a new transformation of muscle-ten don parameters is proposed which enhances the numerical properties of th e parameter estimation, and hence allows us to estimate the most crucial muscle-tendon parameters from a minimum set of experiments, (ii) a new estimation algorithm is proposed which performance is evaluated in a sim ulation environment, and (iii) the performance of the previously present ed algorithm of Garner and Pandy is evaluated in a simulation environmen t to allow a comparison between both methods. Chapter 7: The added value of the estimation of subject-specific musc le-tendon parameters in musculoskeletal modeling of the knee joint actua tors: two case studies This chapter describes the validation of the parameter estimation proced ure as described in chapter 6 on experimental data. To this end, the opt imal muscle fiber lengths and tendon slack lengths of the most sensitive knee joint actuators according to De Groote et al. are estimated based on five isometric dynamometer experiments. The input to the optimisation is: the muscle activations which are registered via surface electromyog raphy (EMG), the total length of the actuators and the moment arms, and the remaining MT-parameters (maximum isometric muscle force, pennation a ngle, and maximum contraction velocity) which are adopted from the gener ic model. Four musculoskeletal (MS-) models are evaluated for two subjec ts. The first model includes linearly scaled geometry and linearly scale d MT-parameter values. The second model includes image-based, hence subj ec-specific, geometry and linearly scaled MT-parameter values. The third model includes linearly scaled geometry and estimated MT-paramete rs for the knee joint actuators. The fourth model includes image-based g eometry and estimated MT-parameter values for the knee joint actuators. The performances of the MS-models are studied for three conditions being isokinetic dynamometry at 30 deg/s, treadmill walking at 4km/h, and cou ntermovement jumping. A forward dynamic analysis is performed (only isok inetic dynamometry) resulting in a predicted knee joint moment which is compared to the experimentally obtained moment, and an inverse dynamic a nalysis is performed resulting in muscle activations which are compared to experimentally obtained activations.The main contributions of thi s study are the evaluation of MS-models including subject-specifically e stimated MT-parameters (of knee joint actuators), and the comparison of the performance of MS-models which are subject-specific to a smaller or a greater extent. It is the rst time that MS-models including functional ly scaled MT-parameters in combination with image-based geometry have be en evaluated.