Application of an Entropy-Assisted Optimization Model in the Prediction of Agonist and Antagonist Muscle Forces
No Thumbnail Available
Files
Date
2007-12-18
Authors
Journal Title
Series/Report No.
Journal ISSN
Volume Title
Publisher
Abstract
Work-related musculoskeletal disorders are a significant problem in industry. A considerable amount of research has been conducted to try to reduce the incidence and severity of these injuries/illnesses. Biomechanical modeling techniques are often employed to estimate the muscle forces and joint loading that occur during physical exertions. One class of these models utilizes optimization techniques to arrive at predictions. Unfortunately, most existing optimization-based biomechanical models fail to predict antagonist muscle activity, which results in an underestimation of overall muscle force requirements and an underestimation of joint reaction forces. Some optimization models predict such a co-contraction, but they either lack a compelling physiological basis or are computationally formidable. The current study takes advantage of the flexible definition of entropy as a scientific measure, and utilizes it in the objective function of an optimization formulation to construct a new optimization model for predicting both agonist and antagonist muscle forces. Recent neurophysiological studies suggested that there exist two separate central nervous systems for generation of two motor patterns: agonist only contraction and agonist-antagonist co-contraction. These have provided a sound physiological basis for a novel optimization-based approach.
In the proposed model (Entropy-ASsisted Optimization Model, or EASOM), the objective function consists of a weighted sum of two components: the sum of cubed muscle stresses, which promotes agonist muscle only exertion, and an entropy related term which encourages the agonist and antagonist co-contraction. The weight factor assigned to the entropy term (W) is set between 0.1 and 1, and it represents the level of co-contraction associated with EASOM muscle force predictions. The constraints include a moment equilibrium constraint and a muscle stress limit constraint. This study focused on a single-plane model in the elbow where the elbow flexion motion in the sagittal plane was investigated. The task variables (or input to the optimization model) included load magnitude, angular velocity, and elbow flexion angle. EMG data were collected and used to obtain the relationship between W and task variables. The muscle forces generated from these empirical EMG data were considered the gold standard and used during model validation.
The results showed that the value of W increases with higher load magnitude, higher angular velocity, and greater elbow extension, which confirms the physiological basis of EASOM, thereby providing support for the theoretical foundations of the approach. The performance of EASOM was compared with three of the most often used optimization-based approaches that exist in the literature (linear programming (LP), double linear programming (DLP) and non-linear programming (NLP) formulations). For predicted muscle forces, EASOM yielded consistently reduced average absolute errors in the prediction of the individual muscle forces for both agonist and antagonist muscle groups. When these muscle forces were used to calculate the joint reaction forces, EASOM showed superiority where it produced average absolute percent errors from 21.13% to 25.23%. The other three approaches generated average absolute percent errors ranging from 77.99% to 107.79% for LP, 21.70% to 58.87% for DLP, and 23.71% to 55.91% for NLP. This evidence supports the EASOM modeling approach in more accurate prediction of muscle forces and joint loading than LP, DLP and NLP.
EASOM employs neurophysiology to construct a computationally economical optimization model that predicts muscle co-contraction, which leads to more accurate prediction of joint loading. It also helps quantify level of muscle co-contraction relative to different task variables. The current study established a sound protocol and solid foundation that will facilitate the expansion of EASOM to other body segments (e.g. the lumbar region of the torso). Data generated from such models can be used as more accurate guidelines in designing safer work postures that can lower the risks of musculoskeletal disorders.
Description
Keywords
electromyography, operations research, kinesiology, soft computing, ergonomics, upper extremity biomechanics, EMG-assisted model, muscle antagonism
Citation
Degree
PhD
Discipline
Industrial Engineering
