Surgical Simulations Based on Limited Quantitative Data: Understanding How Musculoskeletal Models Can Be Used to Predict Moment Arms and Guide Experimental Design

Explainable AI Elucidates Musculoskeletal Biomechanics: A Case Study Using Wrist Surgeries

As datasets increase in size and complexity, biomechanists have turned to artificial intelligence (AI) to aid their analyses. This paper explores how explainable AI (XAI) can enhance the interpretability of biomechanics data derived from musculoskeletal simulations. We use machine learning to classify the simulated lateral pinch data as belonging to models with healthy or one of two types of surgically altered wrists. This simulation-based classification task is analogous to using biomechanical movement and force data to clinically diagnose a pathological state. The XAI describes which musculoskeletal features best explain the classifications and, in turn, the pathological states, at both the local (individual decision) level and global (entire algorithm) level. We demonstrate that these descriptions agree with assessments in the literature and additionally identify the blind spots that can be missed with traditional statistical techniques.

In vivo human gracilis whole-muscle passive stress-sarcomere strain relationship

We measured the passive mechanical properties of intact, living human gracilis muscles (n=11 individuals, 10 male and 1 female, age: 33±12 years, mass: 89±23 kg, height: 177±8 cm). Measurements were performed in patients undergoing surgery for free-functioning myocutaneous tissue transfer of the gracilis muscle to restore elbow flexion after brachial plexus injury. Whole-muscle force of the gracilis tendon was measured in four joint configurations (JC1-JC4) with a buckle force transducer placed at the distal tendon. Sarcomere length was also measured by biopsy from the proximal gracilis muscle. After the muscle was removed, a three-dimensional volumetric reconstruction of the muscle was created via photogrammetry. Muscle length from JC1 to JC4 increased by 3.3±1.0, 7.7±1.2, 10.5±1.3 and 13.4±1.2 cm, respectively, corresponding to 15%, 34%, 46% and 59% muscle fiber strain, respectively. Muscle volume and an average optimal fiber length of 23.1±0.7 cm yielded an average muscle physiological cross-sectional area of 6.8±0.7 cm2 which is approximately 3 times that measured previously from cadaveric specimens. Absolute passive tension increased from 0.90±0.21 N in JC1 to 16.50±2.64 N in JC4. As expected, sarcomere length also increased from 3.24±0.08 µm at JC1 to 3.63±0.07 µm at JC4, which are on the descending limb of the human sarcomere length-tension curve. Peak passive muscle stress was 27.8±5.5 kPa in JC4 and muscle modulus ranged from 44.8 MPa in JC1 to 125.7 MPa in JC4. Comparison with other mammalian species indicates that human muscle passive mechanical properties are more similar to rodent muscle than to rabbit muscle. These data provide direct measurements of whole-human muscle passive mechanical properties that can be used in modeling studies and for understanding comparative passive mechanical properties among mammalian muscles.

Interlimb differences in coordination of rapid wrist/forearm movements

We have previously proposed a model of motor lateralization that attributes specialization for predictive control of intersegmental coordination to the dominant hemisphere/limb system, and control of limb impedance to the non-dominant system. This hypothesis was developed based on visually targeted discrete reaching movement made predominantly with the shoulder and elbow joints. The purpose of this experiment was to determine whether dominant arm advantages for multi-degree of freedom coordination also occur during continuous distal movements of the wrist that do not involve visual guidance. In other words, are the advantages of the dominant arm restricted to controlling intersegmental coordination during discrete visually targeted reaching movements, or are they more generally related to coordination of multiple degrees of freedom at other joints, regardless of whether the movements are discrete or invoke visual guidance? Eight right-handed participants were instructed to perform alternating wrist ulnar/radial deviation movements at two instructed speeds, slow and fast, with the dominant or the non-dominant arm, and were instructed not to rotate the forearm (pronation/supination) or move the wrist up and down (flexion/extension). This was explained by slowly and passively moving the wrist in each plane during the instructions. Because all the muscles that cross the wrist have moment arms with respect to more than one axis of rotation, intermuscular coordination is required to prevent motion about non-instructed axes of rotation. We included two conditions, a very slow condition, as a control condition, to demonstrate understanding of the task, and an as-fast-as-possible condition to challenge predictive aspect of control, which we hypothesize are specialized to the dominant controller. Our results indicated that during as-fast-as-possible conditions the non-dominant arm incorporated significantly more non-instructed motion, which resulted in greater circumduction at the non-dominant than the dominant wrist. These findings extend the dynamic dominance hypothesis, indicating that the dominant hemisphere-arm system is specialized for predictive control of multiple degrees of freedom, even in movements of the distal arm and made in the absence of visual guidance.

Connecting the wrist to the hand: A simulation study exploring changes in thumb-tip endpoint force following wrist surgery

The wrist is essential for hand function. Yet, due to the complexity of the wrist and hand, studies often examine their biomechanical features in isolation. This approach is insufficient for understanding links between orthopaedic surgery at the wrist and concomitant functional impairments at the hand. We hypothesize that clinical reports of reduced force production by the hand following wrist surgeries can be explained by the surgically-induced, biomechanical changes to the system, even when those changes are isolated to the wrist. This study develops dynamic simulations of lateral pinch force following two common surgeries for wrist osteoarthritis: scaphoid-excision four-corner fusion (SE4CF) and proximal row carpectomy (PRC). Simulations of lateral pinch force production in the nonimpaired, SE4CF, and PRC conditions were developed by adapting published models of the nonimpaired wrist and thumb. Our simulations and biomechanical analyses demonstrate how the increased torque-generating requirements at the wrist imposed by the orthopaedic surgeries influence force production to such an extent that changes in motor control strategy are required to generate well-directed thumb-tip end-point forces. The novel implications of our work include identifying the need for surgeries that optimize the configuration of wrist axes of rotation, rehabilitation strategies that improve post-operative wrist strength, and scientific evaluation of motor control strategies following surgery. Our simulations of SE4CF and PRC replicate surgically-imposed decreases in pinch strength, and also identify the wrist's torque-generating capacity and the adaptability of muscle coordination patterns as key research areas to improve post-operative hand function.