Neuromusculoskeletal (NMS) modeling is a valuable tool in orthopaedic biomechanics and motor control research. In this study, three NMS models of the upper extremity were developed, which described three different testing conditions:​ (1) maximum isometric voluntary flexion/extension (MIVF/E), (2) voluntary elbow flexion/extension (VF/E), and (3) constant angular velocity elbow extension (CAVE) respectively.

The MIVF/E model estimated the important musculotendon parameters namely, muscle optimal lengths (lmo), tendon slack lengths (lt0), and maximum muscle stresses (a m) of seven prime elbow muscles by fitting the predicted resultant joint torque profile with the experimentally measured one across the range of 0° – 120°. Results indicated that crm of both flexors and extensors were apparently not significantly different from each other and the normal group seemed to have larger a m as compared to the hemiparetic group. It was also noted that both lmo and lto found by the MIVFE model were comparable to other cadaver studies reported in the literature. We believed that subject-specific musculotendon parameters could be properly predicted in vivo using the present technique.

To evaluate whether it is feasible to incorporate EMG signals with NMS modeling to estimate individual muscle force during dynamic movement, an EMG driven model (i.e. the VF/E model) was developed. It computed the elbow joint trajectories using two different EMG-to-activation processing schemes without involving any trajectory fitting procedures. It seemed that both schemes interpreted the EMG somewhat consistently but their prediction accuracy varied among testing protocols. In general, the VF/E model succeeded in predicting the elbow flexion trajectory in the moderate loading condition but over-drove the flexion trajectory under unloaded condition. The predicted trajectories of the elbow extension in normal subjects were noted to be continuous but the general shape did not fit very well with the measured one. It appears that EMG driven musculoskeletal modeling is a promising approach for noninvasive prediction of individual muscle forces.

The CAVE model extended past modeling efforts in the investigation of spasticity by incorporating explicit musculotendon, muscle spindle, and motoneuron pool models to analyze the effects of changes in motoneuron pool and muscle spindle properties as well as muscle mechanical properties on the biomechanical behavior of the elbow joint observed during a constant angular velocity extension perturbation of the elbow joint. We succeeded in getting a set of motoneuronal related parameters (i.e. the motoneuron threshold (μⁱ₀) and gain (Gⁱ₀) for the biceps, brachialis, and brachioradialis) that resulted in a good fit between the measured and predicted reflex torque. These motoneuron thresholds and gains were found to be substantially different between muscles. Based on results of sensitivity analysis, it appeared that spindle static gain and motoneuron threshold were the most sensitive parameters that could augment reflex torque response, followed by motoneuron gain, and spindle dynamic gain.

Finally, a detailed analysis on the elbow torque and the associated EMG of five prime elbow muscles generated during maximum isometric voluntary flexion/extension (MIVF/MIVE) at eight different elbow positions was conducted. In patients with hemiparesis, the mean MIVE torque among the eight tested positions was relatively more affected as compared with the mean MIVF torque. A non-uniform distribution of weakness at different joint positions was noted for both MIVF and MIVE. More specifically, during MIVE, extension torques at the more extended positions are relatively more reduced than in normal subjects, whereas during MIVF, flexion torque generation in the most flexed position seemed to be more compromised.