Development of an inverse musculoskeletal model of the wrist

Abstract

The wrist is a complex mechanical system that plays a crucial role in many activities of daily living. Some pathologies that affect the wrist are mechanically instigated or propagated, like osteoarthritis, and can have significant effects on quality of life. The small size of the joint complex precludes some of the investigative techniques that are employed in investigating lower-limb pathology. One way to gain understanding of the biomechanics of the system is to create a computational model to perform investigations that cannot be carried out in vivo. The ambition is to apply an inverse musculoskeletal model of the wrist, previously developed at Imperial College London and implemented using a novel anatomical data set, to answer clinical questions by using biomechanical research to inform intervention.

As a key input to the model is joint kinematics, the formation of the joint coordinate system (JCS) used to collect upper limb kinematics was a primary focus of this thesis. The recommendations for building the coordinate system commonly used, published by the International Society of Biomechanics (ISB), are difficult to implement in vivo as they depend on observations only feasible with cadavers. Likewise, the model uses the natural anatomical axes identified by calculating screw displacement axes of passive motions of a cadaveric wrist and thus the axes may differ from axes defined in vivo.

Inconsistencies in the relative position and orientation of these axes in the literature raised the question of whether their in vitro definition would match the in vivo definition. A study was conducted to investigate the relative position and orientation of the natural axes of the wrist and to create an alternate joint coordinate system for the wrist using readily palpable anatomical landmarks of the hand and forearm. Participants performed flexion-extension (FE) and radial-ulnar deviation (RUD) motions with their dominant limb, both unrestricted and with a single-plane constraint, as well as pronation- supination (PS) and dart throwing motion. The muscle activities for the flexor digitorum superficialis, flexor carpi ulnaris, flexor carpi radialis, pronator teres, extensor digitorum communis, extensor carpi ulnaris, and extensor carpi radialis were recorded using surface electromyography (EMG).

It was determined that defining the axes of the wrist with a prescribed motion pathway produces different results to unconstrained in vivo motion. The mean distance between the unconstrained FE and RUD axes, in the direction of the long axis of the forearm, was 2.5 ± 3.9mm and this was statistically different (p < 0.03) from that of the constrained axes (1.6 ± 4.0mm). The mean angular distance in the plane perpendicular to the long axis of the forearm was 53.2 ± 10.8◦. Again, this was statistically different (p < 0.001) from the constrained axes where the angular difference was 107.8 ± 17.7◦. The distance and angular difference between the constrained FE axis with the unconstrained RUD axis were similar to those documented in the literature. This suggests that the reason for the inconsistencies is that the motions were performed in different ways, rather than that they resulted from anatomical differences.

Proposed alternate joint coordinate systems were compared to the ISB recommended system. Landmark palpation repeatability, axes direction repeatability, and amount of secondary rotation (e.g. rotation in RUD and PS axes during FE) were the metrics used

to compare the systems. No difference was found between the ISB recommended JCS and those created as part of the study in any of the three metrics. This means that, for the given metrics, the proposed JCSs performed as well as the ISB recommended system and thus could be used instead, making the quantification of kinematics more feasible in a clinical setting. As a result, I recommend that an alternate JCS that uses the medial and lateral epicondyles, radial and ulnar styloids, the base of the third metacarpal, and the heads of the second and fifth metacarpal is used for in vivo clinical and research use.

EMG signals were normalised by activity during maximal voluntary contraction (MVC) of the observed muscles. Nine tasks, selected from the literature, were performed and the task most likely to elicit MVC in each muscle was noted. The non-dominant limb was also investigated to determine whether dominance had an effect on the task most likely to elicit MVC. Dominance had limited effect with statistical differences being found only in the finger flexors and extensors (p < 0.031). Tasks most likely to elicit MVC were identified for each muscle. These results can be used to produce MVC protocols tailored to the muscles being investigated, can help check for crosstalk during electrode placement, and show that limb dominance needs to be considered when recording EMG for the finger muscles.

The collected MVCs were used to normalise the EMG data that are presented in the thesis. It was found that the EMG pattern for each participant was statistically different from the others (p< 0.001) meaning that each individual employs a unique neuromuscular control algorithm for motions of the wrist. The primary differences were levels of co-contraction. This was consistent within the participants’ trials which suggests that there may be an anatomical reason for the level of co-contraction as this would be unique to each participant. The EMG data were also used to validate a

musculoskeletal model of the wrist, previously developed at Imperial College, for in vivo applications. The kinematics for each participant were input into the model and the muscle forces were calculated. Simulated muscle activity was then calculated by normalising the muscle force by the maximum muscle force for each muscle. Five simulated muscle activities could be compared with the EMG data. The simulated muscle activity patterns matched the recorded EMG patterns both qualitatively and quantitatively, using statistical parameter mapping. No statistical difference was found between the recorded and simulated muscle activity. Thus the model is considered to be valid for predicting muscle activity during in vivo motion of the wrist. Though there was poor correlation between the model results and the EMG (r< |0.18|), this was due to the lack of co-contraction in the model. A magnitude-phase-comprehensive measure was also used to compare the model muscle activity to the EMG. The two were in good phase agreement (P< 0.3) but differed in magnitude (M> |0.65|), with the model producing the pattern with the smaller magnitude. It is hypothesised that this is again due to the lack of co-contraction, as agonist muscles would need to be more active to counter the forces generated by antagonists.

Thus a JCS for the wrist that is employable in a clinical setting and performs as well as the ISB JCS has been identified; muscle activation patterns for the wrist have been identified; and the Imperial College London wrist model has been validated for in vivo use.