A practical solution to reduce soft tissue artifact error at the knee using adaptive kinematic constraints

Abstract

Musculoskeletal modeling and simulations have vast potential in clinical and research fields, but face various challenges in representing the complexities of the human body. Soft tissue artifact from skin-mounted markers may lead to non-physiological representation of joint motions being used as inputs to models in simulations. To address this, we have developed adaptive joint constraints on five of the six degree of freedom of the knee joint based on in vivo tibiofemoral joint motions recorded during walking, hopping and cutting motions from subjects instrumented with intra-cortical pins inserted into their tibia and femur. The constraint boundaries vary as a function of knee flexion angle and were tested on four whole-body models including four to six knee degrees of freedom. A musculoskeletal model developed in OpenSim simulation software was constrained to these in vivo boundaries during level gait and inverse kinematics and dynamics were then resolved. Statistical parametric mapping indicated significant differences (p < 0.05) in kinematics between bone pin constrained and unconstrained model conditions, notably in knee translations, while hip and ankle flexion/extension angles were also affected, indicating the error at the knee propagates to surrounding joints. These changes to hip, knee, and ankle kinematics led to measurable changes in hip and knee transverse plane moments, and knee frontal plane moments and forces. Since knee flexion angle can be validly represented using skin mounted markers, our tool uses this reliable measure to guide the five other degrees of freedom at the knee and provide a more valid representation of the kinematics for these degrees of freedom.

Introduction

The knee joint is a complex load bearing joint and moves in all six degrees of freedom (DoF. Musculoskeletal modeling and computer simulations are very useful clinical and research tools to objectively address musculoskeletal disorders and their costly effect on the health care system (Delp et al., 1990, Fregly et al., 2011, Kinney et al., 2013, Komistek et al., 2005, Reinbolt et al., 2005). Yet, these models often simplify the knee with a 1-DoF hinge joint (Anderson and Pandy, 2001, Arnold et al., 2010, Delp et al., 1990, Modenese et al., 2011, Sasaki and Neptune, 2010, Shourijeh and McPhee, 2014, Xu et al., 2014), limiting its physiological realism. The remaining degrees of freedom are typically neglected or defined as functions (prescribed motions) of the flexion/extension angle of the knee based on cadaveric studies and geometric modeling of the femur, tibia and patella (Coughlin et al., 2003, Delp et al., 1990, Yamaguchi and Zajac, 1989). A musculoskeletal model developed by Xu et al. (2014) defines the knee more flexibly with four un-prescribed DoFs, with the remaining two (anterior/posterior and distal/proximal translations) prescribed as functions of the knee flexion angle from 0° to 120°. This model dramatically improves the physiological representation of the knee; however, while valid kinematics in flexion can be determined using skin mounted markers, valid motions of internal/external rotation, abd-adduction and medial/lateral translation are also required to correctly define model motion. Furthermore, using prescribed motions limits important in vivo inter-participant variability (Benoit et al., 2007). Tibiofemoral motions are also sensitive to loading conditions such that for a given knee flexion angle, more than one secondary motion is possible (Andriacchi and Dyrby, 2005, Benoit et al., 2007).

Typically, reflective skin mounted markers tracked by motion capture systems used to track experimental body kinematics are applied to musculoskeletal models. Soft tissue artifact (STA), the movement of markers caused by skin deformation and not actual bone motion, has a large effect on determining body kinematics (Benoit et al., 2006, Gao and Zheng, 2008, Gasparutto et al., 2015, Li et al., 2012, Reinschmidt et al., 1997). Benoit et al. (2006) compared knee joint skeletal segment movement in healthy adult males using skin mounted markers as well as rigidly linked marker clusters directly implanted into the femur and tibia with bone pins. Skin markers were found to be heavily affected by STA in all but the larger motion of flexion/extension. Errors of up to 13.1° in rotations and up to 16.1 mm in translations were identified. This error is a function of marker cluster translation and rotation, rather than cluster deformation (inter-marker motions; Benoit et al., 2015) and while modeling the error is possible (Andersen et al., 2012), it is not yet feasible without subject specific data obtained in vivo to generate the model. Different methods have been used to introduce anatomical, kinematic or geometrical constraints in musculoskeletal models to limit STA (Andersen et al., 2010, Duprey et al., 2010, Gasparutto et al., 2015, Leardini et al., 2005); however, these constraints may in fact increase the error due to over-simplification of the complex six DoF human knee joint (Andersen et al., 2010).

Despite STA, it is generally accepted that relative errors in the sagittal plane are small enough to conclude that knee joint flexion/extension can be validly represented using skin-mounted marker clusters. In order to improve the biofidelity of musculoskeletal models of the lower limb, we have developed an adaptive model of knee joint motion using the valid information from knee flexion angles to guide the motion boundary ranges for each of the remaining five DoF based on in vivo tibiofemoral motion data. For these boundaries to be valid, they must represent the potential ranges of motion across a wide range of activities, loading conditions, and inter-participant anatomical differences, allowing for variable secondary motions for a given knee flexion angle.

The purpose of this study is to present and evaluate a tool designed to reduce the error caused by STA at the knee. We combined data from gait, hopping and cutting (Benoit et al., 2006, Benoit et al., 2007) knee joint motions to establish the maximum physiological boundary constraints of the healthy knee (i.e. the maximum potential range of motion of the DoFs of the knee for given flexion angles).

We evaluated the tool during gait by comparing the lower extremity kinematics and kinetics of our adaptive constrained knee joint model on four variations of a whole body musculoskeletal model (Xu et al., 2014) available in the OpenSim repository. Despite the fact that the boundary ranges for knee joint motions were developed from more challenging activities, including the side-cut, it is hypothesized that implementing our adaptive constraints will yield kinematic and kinetic results more reflective of in vivo bone motions during gait when compared to the unconstrained models.