A normative database of hip and knee joint biomechanics during dynamic tasks using anatomical regression prediction methods

Abstract

Many methodologies exist to predict the hip joint center (HJC), of which regression based on anatomical landmarks appear most common. Despite the fact that predicted HJC locations vary depending upon chosen method, inter-study comparisons and inferences about populations are commonly made. The purpose of this study was to create a normative database of hip and knee biomechanics during walking, running, and single leg landings based on five commonly utilized HJC methods to serve as a reference for inter-study comparisons. Secondarily, we devised to provide comparisons of peak knee angles and hip angles, moments, and powers from the five HJC methods. Thirty healthy young adults performed walking, running, and single leg landing tasks at self-selected speeds (walking/running) and at 90% of their maximum jump height (landing). Three-dimensional motion capture and ground reaction forces were collected during all tasks. Five different HJC prediction methods: Bell, Davis, Hara, Harrington, and Greater Trochanter were implemented separately in a 6 degree of freedom model. Predicted HJC locations, direct kinematics, and inverse dynamics were computed for all tasks. Predicted HJC mediolateral, anteroposterior, and superior-inferior locations differed between methods by an average of 1.3, 2.9, and 1.4 cm, respectively. A database was created using the mean of all subjects for all five methods. In addition, one-way ANOVAs were used to compare triplanar peak angles, moments, and powers between the methods. The database of hip and knee biomechanics illustrates (1) variability between methods increases with more dynamic tasks (running/landing vs. walking) and (2) frontal and transverse plane hip and knee biomechanics are more variable between methods. Comparisons between methods found 38 and 16 main effect differences in hip and knee biomechanics, respectively. The Greater Trochanter method provided the most differences compared with other methods, while the Davis method provided the least differences. The database constructed provides an important reference for inter-study comparisons and details the impact of anatomical regression methods for predicting the HJC.

Introduction

Three-dimensional (3D) motion analysis is an integral component of biomechanics research, and depends heavily on accurate methodology to estimate the location of the joint centers (Fiorentino et al., 2016b, Stagni et al., 2000). The hip joint center (HJC) is impossible to identify using palpation and marker placement due to its location deep within soft tissue (Bell et al., 1989). Motion capture utilizes various methods to predict HJC locations, which are utilized within calculations of joint kinematics and kinetics during activities of daily living and sports maneuvers (Assi et al., 2016, Kainz et al., 2017a, Sangeux et al., 2014). For the femur segment, the HJCs are utilized in deriving the superior-inferior axis and frontal plane, which then create the anteroposterior and mediolateral axes, and are typically set as the femur segment origin. Therefore, the accuracy and reliability of HJC prediction methods are of upmost importance in assessments of hip biomechanics. Currently, there is a lack of thorough analyses of HJC prediction methodologies’ resulting kinetic and kinematic data. In addition to hip biomechanics, the HJC can influence knee kinematics through modifications of femur reference frames.

Existing assessments of HJC prediction methods utilize the gold standard of medical imaging technology to evaluate validity (Assi et al., 2016, Kainz et al., 2017a, Sangeux et al., 2014). Although much can be gleaned from such assessments, current findings are conflicting as to a unanimous best available method, and little information is available regarding the differences in dynamics between methodologies. Additionally, a multitude of past and current hip joint based research studies do not disclose what HJC method was used in their analyses (Aizawa et al., 2016, Astephen et al., 2008, Decker et al., 2003, Donohue et al., 2015, Kulas et al., 2008, McLean et al., 2004, Myer et al., 2015, Nordin and Dufek, 2016, Yeow et al., 2011), despite receiving continued citations and inter-study comparisons. This lack of HJC method specificity and uniformity within the literature base makes it difficult to compare kinetic and kinematic results from multiple studies. Therefore, standards must be determined in regards to the accuracy and reliability of the various motion capture-based HJC methods that are commonly used in biomechanics research, as well as an established, all-encompassing comparison among the various methods. Specifically, a comparison of method based hip and knee biomechanics during a multitude of dynamic movements is necessary.

Two primary options exist for HJC prediction methods: (1) calculations based on a range of functional movements and (2) regressions utilizing anatomical landmarks. Functional calibration methods determine the center of rotation at the hip joint based on the movement of the thigh relative to the pelvis (Ehrig et al., 2006, Leardini et al., 1999, Piazza et al., 2004, Schwartz and Rozumalski, 2005). Reliability of many functional methods has been investigated, confirming several available methods are reliable (Besier et al., 2003, Weinhandl and O'Connor, 2010). However, the accuracy and reliability of functional methods relies on adequate hip range of motion (Kainz et al., 2017b, Sangeux et al., 2014) and the movement task (Camomilla et al., 2006). For this reason, functional methods are not always possible for certain populations that may have limited or impaired range of motion. Over thirteen methods are included in the category of functional methods (Camomilla et al., 2006, Ehrig et al., 2006, Kainz et al., 2015, Leardini et al., 1999, Siston and Delp, 2006); all of which are performed using various movement tasks, require additional data reduction time and knowledge of programming, and result in a wide range of HJC outcomes. Therefore, many research labs utilize regression predictions based on palpable anatomical landmarks.

Similar to functional methods, there are multiple regression methods available, which employ predictions based on anatomical landmarks from cadaver (Seidel et al., 1995) or medical imaging data (Bell et al., 1989, Bell et al., 1990, Davis et al., 1991, Hara et al., 2016, Harrington et al., 2007), to estimate HJC locations. Because regression methods do not require additional movement tasks, these methods are often recommended for populations with insufficient hip range of motion (Wu et al., 2002). Furthermore, no additional computing or programming time or knowledge is required for regression methods. Although functional HJC methods may provide more accurate predictions of static HJC locations compared to regression methods (Fiorentino et al., 2017, Hicks and Richards, 2005, Leardini et al., 1999, Sangeux et al., 2011, Sangeux et al., 2014), many researchers may choose to utilize regression models due to their ease of implementation. This paper will include hip and knee biomechanics comparisons of the following predictive methods: Bell et al., 1989, Bell et al., 1990, Davis et al., 1991, Harrington et al., 2007, Hara et al., 2016 and Greater Trochanter (GT) (Bennett et al., 2016, Weinhandl and O'Connor, 2010) during several movements.

Few studies have investigated the effects of anatomical regression or functional movement HJC methods on kinematics and kinetics during activities of daily living (such as walking) and planar movements (Kirkwood et al., 1999, Sinclair and Bottoms, 2013, Stagni et al., 2000). Although each study compares varying combinations of available methods, the conclusion is the same: HJC kinematics/kinetics are affected by HJC prediction method. For level walking, joint kinetics are affected by prediction method, as hip moment differences can range from −0.10 to +0.25 Nm/kg (Kirkwood et al., 1999). Additionally, implementing 3 cm location errors can impact sagittal and frontal plane hip moments by 22 and 25%, respectively (Stagni et al., 2000). Prediction methods can also produce kinematic differences, reaching 5–6°in children with cerebral palsy (Kainz et al., 2017a) and adults during walking (Żuk et al., 2014) and fencing lunges (Sinclair and Bottoms, 2013). Recent research has also found significant variations in predicted HJC locations during tasks such as treadmill walking and planar hip motions compared with imaging results (Fiorentino et al., 2016a, Fiorentino et al., 2016b), which certainly impacts hip biomechanics. As the influence of HJC prediction method is apparent, development of a database containing method-based hip and knee biomechanics during commonly investigated dynamic tasks (walking, running, jumping/landing) is integral.

Therefore, the purpose of this study was to create a normative database of hip and knee biomechanics during walking, running, and single leg landings based on five commonly utilized HJC methods. Secondarily, we devised to provide comparisons of peak knee angles and hip angles, moments, and powers from the five HJC methods.