Abstract

To better understand the influence joint mechanics may have on the development or progression of joint degenerative diseases such as knee osteoarthritis (OA), EMG-driven models have been developed to estimate in-vivo muscle forces that contribute to the internal joint mechanics. However, there is a limited understanding of the model's ability to estimate joint mechanics during stair ascent and stair descent, which induce larger joint loads that may contribute to the development or progression of knee OA. In addition, the model's ability to estimate the joint moment when only one trial is used during the tuning process is also unknown. Therefore, the ability to estimate the flexion/extension (F/E) knee joint moment during walking, stair ascent, and stair descent with an EMG-driven model was evaluated. In addition the influence of using a single trial to tune the model was evaluated.

Knee joint kinematics for five healthy subjects were measured using motion capture during walking, stair ascent, and stair descent. The muscle-tendon lengths and moment arms of 11 muscles crossing the knee were then determined from an anatomical model of the lower leg. The muscle-tendon lengths and respective EMG signals served as inputs to the EMG-driven model. The model consisted of a neural activation component to estimate the muscle activation and a muscle contraction dynamics component that was based on a modified Hill-type model. The tuning process adjusted specific muscle properties to minimize the difference between the F/E knee joint moment determined from inverse dynamics and the moment estimated by the model. Only one trial of a given task was used during the tuning process. The resulting muscle parameters were then used to predict the joint moment for trials of the same task that were not used in the tuning process.

The estimated F/E knee joint moment exhibited the smallest RMS error and largest coefficient of determination during stair ascent, followed by walking, and stair decent. Overall, the trial used during the tuning process did not appear to influence the model's ability to estimate the joint moment for walking, stair ascent, or stair descent.