Modeling and Simulation Credibility Assessments of Musculoskeletal Computational Models for Simulating Astronaut Injuries Due to a Poor Spacesuit Fit

The health of an astronaut's musculoskeletal (MS) system is crucial for the body to carry out physiological functions for NASA-centric extravehicular activity (EVA) tasks. Multiple injury scenarios affiliated with these EVA activities—on-Earth training and spaceflight operations—pose unique challenges to a suited astronaut that could lead to muscle strain, tendon, or ligament injuries. When an astronaut is confined within the space suit, especially a space suit that does not fit correctly, mobility of the extremities is limited, resulting in the astronaut's dependence on the excessive unnatural activation of the muscles to maintain a stable gait and maneuverability. This abnormal muscle activation and movement of the extremities could result in overuse and fatigue of multiple muscles, tendons, or ligaments [1]. Such overuse and fatigue of the musculature and connective tissue could decrease the astronauts performing tasks essential to space operations.

Whole-body MS computational models provide a viable path for assessing the biomechanics of astronaut injury mechanisms and associated risks. opensim [2] MS models have been validated and used for various terrestrial injury scenarios encountered in sports, ambulation, and physical activities associated with exercise and work. While these MS models have been verified and validated for injury modalities mentioned above [3,4] to the authors' best knowledge, these MS models' Modeling and Simulation (M&S) Verification, Validation, and Credibility (VV&C) assessments for NASA-centric applications have not been performed. The American Society of Mechanical Engineers, the Food and Drug Administration, and the National Institutes of Health recommend VV&C standards for evaluating a computational model's credibility for their respective domains of the application [5,6]. Within the NASA Human Research Program, VV&C assessments are performed using NASA Standard 7009A [7]. NASA-STD-7009 and NASA-STD-7009A have been developed after the Columbia accident investigation to improve the computational models' credibility for NASA space exploration-centric applications. NASA-STD-7009A provides a systematic and comprehensive approach to evaluating a computational model's credibility using eight credibility factors—data and input pedigrees, verification, validation, results uncertainty, results robustness, M&S or use history, and M&S management. In this study, data and input pedigree, verification (code and solution), validation (conceptual and referent), results uncertainty, and results robustness (sensitivity) credibility factors are used to assess opensim MS models for predicting poor suit fit related MS injuries. Figure 1 illustrates this credibility assessment process, assignment of subject matter experts informed sufficiency thresholds and ordinal scores for credibility factors and credibility improvement strategies for MS models' use in modeling astronauts' poor suit fit related MS injuries. The MS models quantified kinematic and kinetic muscle strain and ligament injury metrics at various anatomical locations, including the hip, lower back, knee, ankle, hand, and wrist. This paper seeks to elucidate the current M&S credibility of opensim computational models for simulating injury mechanisms arising from an astronaut's poor suit fit injury scenario.

The opensim computational models were developed to simulate walking and running motions. Since these activities primarily required the use of lower body muscles, the upper body muscles were excluded from the model since they contribute little to the biomechanical response and would add to the computational analysis time. opensim is an open-source software package with new models being created by users that include muscles throughout the entire body to allow for more complex motions and different analyses. The opensim MS models that are openly shared by the biomechanical modeling community are available from the online repository, SimTk.org (National Institutes of Health Grant No. R01GM124443 01A1). The opensim model simulates the MS system by representing individual muscles as musculotendon actuators with both active and passive components. opensim calculates the torque at each joint to compare to allowable values as a measure of determining muscle strain. Muscle group forces are also calculated to compare to allowable values to determine if a muscle is subject to excessive force. If the joint torque or muscle force from the analysis exceeds published values, the muscles or tendons could be strained. Muscle strain injuries result from tearing of the muscle fibers and attached tendons. Ligament injuries are sprains to the soft tissues connecting bones when the joint moves into an unnatural position. These mechanisms could arise during on-Earth training in simulated reduced gravity in the Neutral Buoyancy Lab (NBL) or the Active Response Gravity Offload System (ARGOS) in space under microgravity conditions or during Lunar or Martian surface activities under partial gravity. The models included in this credibility assessment of the poor suit fit injury scenario are illustrated in Fig. 2 with Table 1 providing details of the models. Descriptions of the full body and partial body MS models are given in the Supplemental Materials on the ASME Digital Collection.

The NASA Standard 7009A document was the reference for this credibility assessment of the poor suit fit injury scenario that included the following factors: data pedigree, input pedigree, code and solution verification, conceptual and referent validation, results uncertainty, and results robustness (sensitivity). The process for this assessment is illustrated in Fig. 1. The opensim models were evaluated based on published documentation regarding their capability to analyze a particular injury mechanism. A credibility score was assigned for each factor using the key aspects of credibility assessment levels from NASA Standard 7009A.

The opensim Full Body Model was used to assess a muscle strain injury occurring in the muscles of the pelvis, legs, and feet due to a poor space suit fit. The data pedigree score is 2 for a muscle strain based on the detailed Full Body Model available for analysis and the muscle data in the model. Musculotendon properties for the assessed opensim models were obtained from postmortem human subject (PMHS) data. The input pedigree score is 2 for all models since the same motion capture method is used to collect input data regardless of which model is used for analysis. The models have sufficient code and solution verification resulting in scores of 1 for code verification and 2 for solution verification, but the model authors recommend more comprehensive testing. The model documentation for the Full Body Model includes results validated to experimentally obtained electromyography (EMG) data, and thus a score of 2 is given for both conceptual and referent validation. Verification and validation procedures simulating the mass and joint restrictions of an EVA spacesuit will need to be performed once these future models are created. Figure 3 illustrates the credibility assessment scores for the muscle strain injury.

The two opensim models assessed for hand/glove muscle strain injuries are the Stanford Upper Limb Model and the Wrist Model. These are detailed models of the hand and arm but do not include substantial information regarding verification and validation procedures. The assessment score for code verification is 0 and 1 for the solution verification, conceptual validation, and referent validation scores for a muscle strain injury for both models. Verification and validation procedures will also need to be conducted on models modified to include simulated EVA gloves or an upgraded whole-body model incorporating an arm model. The hand muscle strain injury data pedigree score is 1 since musculotendon properties were obtained from PMHS data. Figure 4 illustrates the credibility assessment scores for the hand/wrist injury.

The Discrete Element Knee Model assessed for the ligament injury is a detailed six degree-of-freedom model of the right knee with 18 ligaments, bones, and knee cartilage. The assessment score for code verification is 0 and 1 for solution verification since there is no evidence of a whole-body model with ligaments which would be beneficial to analyze EVA related ligament injuries. The conceptual validation assessment score is 0 since studies need to be conducted representative of an EVA injury and 1 for referent validation since ligament parameters and cartilage forces calculated by the model were validated against various experimental and PMHS data sources per the model documentation. The Full Body Model does not include ligaments but an improved model with ligaments incorporated will need to have verification and validation procedures conducted. The ligament injury data pedigree score is 1 since the knee ligaments are modeled as a combination of ligament and muscle features with properties obtained from PMHS data. Figure 5 illustrates the credibility assessment scores for the ligament injury.

The Discrete Element Knee Model documentation presents results of both sensitivity and uncertainty analyses performed on the model. There is no evidence that results uncertainty or robustness (sensitivity) analyses were performed for the other models, and thus a score of 0 was given for both factors for all injury mechanisms. Results uncertainty and sensitivity analyses will also need to be performed on models with the spacesuit mass included or models modified to include additional features such as ligaments or hand and arm muscles.

The results from the M&S credibility assessment of the opensim models presented herein evaluate their effectiveness in analyzing NASA-specific biomechanics applications. Since the models were developed, verified, and validated primarily for on-Earth analyses of domestic and sports related injury scenarios, additional M&S credibility assessments are warranted when using these computational models for NASA-centric domains of application. These include simulating injury mechanisms arising from a poor suit fit scenario during on-Earth suited astronaut training in the NBL or ARGOS and EVAs in the altered gravity (microgravity or lunar) environments. An actual in-space EVA or EVA preparatory training requires the astronaut to wear an extravehicular mobility unit or exploration extravehicular mobility unit spacesuit, resulting in a different level of muscle and joint activities. Current MS modeling of suited astronauts uses simplified versions of these spacesuits and might have different musculature responses compared to a fully suited astronaut. Of the assessed models, one model alone does not have all the anatomical MS features required to analyze the muscle strain and ligament injuries an astronaut may encounter on the body. Partial models can be used for injury risk analyses of isolated body regions, but accurately incorporating the body interface and boundary conditions—to abstract missing body anatomy—is imperative to avoid unrealistic results. The opensim models included in this credibility assessment of the poor suit fit injury scenario are illustrated in Fig. 2 with Table 1 providing details of the models.

The input pedigree assessment of the models is based on the input required by opensim which consist of files containing three-dimensional motion capture marker coordinates and ground reaction forces. This input data is generated using a motion capture system, and the output is processed into files used as input to the model. The motion created by this data is simulated within the code, and the model virtual markers are observed within the graphical user interface to match the experimental markers from the test data. This coinciding of markers is evidence that the model is correctly simulating the motion of the test.

The data pedigree assessment is based on information obtained from published literature. Muscle specimens are dissected from cadavers. Their properties are determined from biomechanical testing and then publicly disseminated through peer reviewed publications for others to use when creating biomechanical models. Muscle properties included in the Full Body Model were obtained from cadaver studies by Ward et al. [8] and magnetic resonance imaging muscle volume data measured by Handsfield et al. [9]. The muscle properties were measured from cadaver specimens which is an inherent limitation since the activities analyzed with the models are performed by living human beings. There may be a difference between the cadaver muscle properties and those from live muscle tissue, but this is acceptable since certain properties can only be obtained from cadaver research. Although the existing muscle definitions in the models include biomechanical properties evaluated and calibrated using published data, new ligament features added to a model will require such biomechanical properties determined using the mechanical behavior of PMHS ligament specimens or from other surrogate subjects.

The opensim software was created to analyze human body motion calculating joint kinematics, joint torques, and muscle forces. The human body model used as input to the software does not discern between movements that result from on-Earth activities such as walking, lifting, a squat exercise, or the motion of an astronaut performing an EVA during training on Earth or the reduced gravity environment of space. The model has an option to modify the applied magnitude and direction of the gravitational acceleration vector acting on the model for spaceflight. The base model is representative of the size and weight of the human body with no additional outerwear or objects attached and will need to be tailored for specific applications. A test subject simulated by the model will typically perform a task with no restrictions on joint movement within the default range. If the test subject is performing the task with outerwear that may restrict joint movement, such as a spacesuit, then the joint range should be modified to remain within the restrictions of the spacesuit. The model should also include the additional weight of the spacesuit so that when the muscles are activated to perform a task, they will overcome the resistance of the spacesuit.

The current level of verification and validation of the assessed models varies. The Full Body Model documentation provides the details of the model verification studies performed and is summarized here. Muscle forces and joint moments for two test subjects, one walking and one running, were calculated independently using the computed muscle control and inverse dynamics algorithms in opensim and compared to one another. Validation was performed by comparing the calculated muscle activations from the model to measured EMG data for both test subjects for the major lower body muscles. The results were also compared to published data documented in Ref. [10]. The Discrete Element Knee Model documentation includes a summary of the validation studies performed on the model. The flexion knee angle varied passively. The kinematic displacement and contact force results were compared to experimental data from Refs. [11] and [12]. Two additional validation cases were performed by separately applying a force and a torque to the tibia and calculating the knee joint displacements and rotations. The Stanford Upper Limb Model and the Wrist Model are similar models of the arm and hand. Verification was performed on the Stanford Upper Limb Model by calculating the muscle moment arms and isometric forces within a range of 70 degrees of extension and flexion and comparing the results of the arm and wrist muscles to experimental data from Refs. [13] and [14]. The Stanford Upper Limb Model was validated by using experimental EMG muscular response data for ten test volunteers. These cases were reproduced with the model, and the resulting maximum flexion and extension moments per wrist flexion angle were compared to experimental results. The model documentation provides evidence of the verification and validation procedures that can be applied to an EVA or training activities since the models are capable of analyzing body motions, although the addition of the spacesuit will require model modifications to limit motion. The Full Body Model evidence is limited since this model does not include ligaments or upper torso and extremity muscles. The partial body model procedures are limited since they include only a part of the body (arm or knee) and are not affected by whole-body reactions. Additional analysis will be required to establish confidence in the model results for NASA EVA activities.

The current credibility assessment results provide awareness of the credibility shortfalls of these models for NASA-centric applications and enlist strategies for the elevation of the M&S credibility factors for the intended domain of EVA-related applications. The models assessed for the poor suit fit injury scenario will require credibility improvement strategies to be most effective in predicting EVA injuries. These improvements will benefit the researchers evaluating EVA injury scenarios during on-Earth training in the NBL and ARGOS and reduce gravity by having a model available that has been tailored for EVA-specific applications. A poor suit fit could produce a muscle strain injury, which the Full Body Model is currently capable of analyzing since muscle modeling is the primary feature but is not prepared for evaluation of ligament injuries since the whole-body models lack ligament features. Analysis of EVA injuries may require modifying an existing whole-body model to include these additional features to take advantage of the astronaut's body motion throughout the model. An updated model used for future analysis will also benefit from a new credibility assessment that includes verification and validation procedures and a sensitivity analysis and uncertainty quantification targeted specifically at EVA-related studies. The Discrete Element Knee Model documentation includes information summarizing both uncertainty and sensitivity analyses for this model with ligament features. The documentation for the other models assessed for the muscle strain injury mechanism does not contain evidence of model results' uncertainty quantification or sensitivity analysis. Both of these analyses are an integral part of the model's credibility and a part of the strategies to elevate the scores. Table 2 includes a detailed account of the elevation strategies for each credibility factor for the assessed opensim models, and Table 3 provides in detail a description of each score.

The authors would like to thank the Cross-cutting Computational Modeling Project team at the NASA Glenn Research Center led by Project Manager Courtney Schkurko and the NASA Human Research Program for supporting this work. Additional thanks to Nathaniel Newby and Jeffrey Somers at the NASA Johnson Space Center for the opportunity to perform this study.

• The NASA Human Research Program managed at the NASA Johnson Space Center (Funder ID: 10.13039/100006203).

The authors attest that all data for this study are included in the paper.