Abstract
Telemetric knee implants have provided invaluable insight into the forces occurring in the knee during various activities. However, due to the high amount of cost involved only a few of them have been developed. Mathematical modeling of the knee provides an alternative that can be easily applied to study high number of patients. However, in order to ensure accuracy these models need to be validated with in vivo force data. Previously, mathematical models have been developed and validated to study only specific activities. Therefore, the objective of this study was compare the knee force predictions from the same model with that obtained using telemetry for multiple activities.
Kinematics of a telemetric patient was collected using fluoroscopy and 2D to 3D image registration for gait, deep knee bend (DKB), chair rise, step up and step down activities. Along with telemetric forces obtained from the implant, synchronized ground reaction forces (GRF) were also collected from a force plate. The relevant kinematics and the GRF were input into an inverse dynamic model of the human leg starting from the foot and ending at the pelvis (Figure 1). All major ligaments and muscles affecting the knee joint were included in the model. The pelvis and the foot were incorporated into the system so as to provide realistic boundary conditions at the hip and the ankle and also to provide reference geometry for the attachment sites of relevant muscles. The muscle redundancy problem was solved using the pseudo-inverse technique which has been shown to automatically optimize muscle forces based on the Crowninshield-Brand cost function. The same model, without any additional changes, was applied for all activities and the predicted knee force results were compared with the data obtained from telemetry.
Comparison of the model predictions for the tibiofemoral contact forces with the telemetric implant data revealed a high degree of correlation both in the nature of variation of forces and the magnitudes of the forces obtained. Interestingly, the model predicted forces with a high level of accuracy for activities in which the flexion of the knee do not vary monotonically (increases and decreases or vice-versa) with the activity cycle (gait, step up and step down). During these activities, the difference between the model predictions with the telemetric data was less than 5% (Figure 2). For activities where flexion varies monotonically (either increases or decreases) with activity (DKB and chair rise) the difference between the forces was less than 10% (Figure 3).
The results from this study show that inverse dynamic computational models of the knee can be robust enough to predict forces occurring at the knee with a high amount of accuracy for multiple activities. While this study was conducted only on one patient with a telemetric implant, the required inputs to the model are generic enough so that it is applicable for any TKA patient with the mobility to conduct the desired activity. This allows kinetic data to be provided for the improvement of implant design and surgical techniques accessibly and relatively inexpensively.
