Muscle activity of upper extremity during the is tennis forehand overhead smash: Experimental VS musculoskeletal modeling

Abstract

One of the main parts of body that play key role in tennis matches is shoulder complex [1,2]. There are many joints and muscles caused shoulder to be complex [2–5]. Evaluation of the muscle activities is necessary to improve safety and performance [5]. The fundamental challenge for evaluation of muscle activity is measuring by EMG due to limitation of equipment, expensiveness, and inaccessibility to deep muscles [6–8]. Therefore, it is important to use musculoskeletal modeling to evaluate muscle activation [9–12]. On the other hand, there have been different musculoskeletal models with different joint definitions and the DOF [13,14]. Thus, the goal of this study was to validate the muscle activation output from different model by EMG data for the TFOS. How does muscle activity from experimental and modeling valuations change during the tennis forehand overhead smash (TFOS)? Twenty-five professional tennis athletes (Mass: 69.3±7.5 kg, Heights: 178±9.3 cm, Age: 29.5±7.5 years). The kinematics of markers were recorded by a 12 high-speed motion captures (Vicon, Oxford, UK, 100 Hz). The shoulder model of Holzbaur et al. [15–17] selected as base model and three version of models extracted based on the DOF: (5 DOF) a model with only three rotational DOF between humerus and trunk Glenohumeral joint, (11 DOF) a model with three rotational DOF for Scapulothoracic joint, Acromioclavicular joint, and Glenohumeral joint, (Stanford) a model with coupled motions for scapula, clavicle, and humerus. All models include two DOF for radio-ulna and elbow joints. After scaling the models, the inverse kinematics, inverse dynamics, and static optimization tools were applied to compute kinematics, kinetics, and muscle activity variables. The EMG activity in selective muscles was measured by the Myon wireless EMG system with a sampling frequency of 1000 Hz [18]. The average RMS of differences between each model and EMG (RMSE) over the muscles were 0.27±0.10, 0.29±0.12, and 0.22±0.10 for 5DOF, Stanford, and 11DOF models, respectively. Furthermore, the average Pearson's correlation coefficient over the muscles were 0.89±0.08, 0.88±0.09, and 0.93±0.60 for 5DOF, Stanford, and 11DOF models, respectively. The minimum RMS error (0.22±0.10) and maximum Pearson's correlation coefficient (0.93±0.60) were observed for 11 DOF model. Table 1: Muscle activity comparison between musculoskeletal simulation outputs (from three different models) and experimental data (EMG) including the RMSE, and Pearson's correlation coefficient for the TFOS movement.Download : Download high-res image (181KB)Download : Download full-size image According to the results, the 11 DOF model are more similar to the experimental (EMG) based on both RMSE and Pearson's correlation coefficient. Although the simulation results of some muscles were significantly different from the experimental results. Therefore, the alternative method to quantify muscle activation is musculoskeletal modeling. Moreover, the best model to reconstruct the muscle activation is 11DOF model.