Computational modeling have become a routine and powerful strategy for academic research and clinical care. Consequently, significant scientific discoveries were made, innovative products were launched, and individualized delivery of healthcare has become a possibility. The scientific and clinical domain of knee biomechanics is no exception. The knee is a major site of orthopaedic problems resulting in annual physician visits on the order of tens of millions. Modeling and simulation offers a cost-effective and prompt path to respond to the pressing medical needs for restoration of knee function. However, the reproducibility of simulation results, to inform scientific and clinical decision making, is questionable. Reproducibility is a pressing issue in scientific conduct. For modeling and simulation, there is added scrutiny particularly with the desire to repurpose and reuse virtual specimens for prospective solutions of diverse scientific and clinical problems. A significant portion of modeling and simulation workflow includes model development, evaluation, and simulation. This workflow, while based on objective scientific principles, commonly requires intuition during implementation; therefore relies on the knowledge and expertise of the modeler. This ?art of modeling? can be a fundamental source of diminished reproducibility. The goal of this study is to understand how modelers? choices to build models, even when using the same data, may influence predictions and therefore the reproducibility of simulation results. Five modeling and simulation teams will independently develop computational models of knees based on the same data sets and simulate the same scenarios, which are relevant to scientific and clinical understanding of knee biomechanics. Ideally, predicted joint and tissue mechanics will be the same. In practice, the skills and experiences of model developers will reflect upon their modeling choices; and as a result, discrepancies will exist. The proposed activity will document the magnitude and potential sources of such discrepancies through comparisons of model components and simulation results. This project will examine and critique the current state of model development and simulation reproducibility in joint and tissue mechanics. This will translate into reliable models of the knee joint for simulation-based discoveries and in silico design and evaluation of medical devices and interventions. The required exchange of data, model components, and simulation results among the teams and with the public will also impact developers and users of such resources. Specifications, to facilitate data and model exchange and to develop data and modeling standards, and guidance, to inform modeling and simulation workflows, will likely emerge as by-products of the research activity. The investigators are leaders in simulation-based explorations of knee biomechanics and include representatives of prominent research laboratories and clinical institutions worldwide. With this project, they will establish the necessary scientific rigor to recognize computational modeling and simulation as a dependable component of knee research and the joint's clinical care.
The potential of modeling and simulation to enable significant scientific discoveries and to improve clinical care is not limited to knee biomechanics. Yet, realization of simulation-based approaches as routine and dependable strategies for healthcare delivery requires establishing their reproducibility to be independent from the analyst?s preferences. By understanding the influence of modelers? approaches and decisions (essentially their art) throughout the lifecycle of modeling & simulation, this project will demonstrate the uncertainty of delivering consistent simulation predictions when the founding data to feed in to models remain the same.
| Mulugeta, Lealem; Drach, Andrew; Erdemir, Ahmet et al. (2018) Credibility, Replicability, and Reproducibility in Simulation for Biomedicine and Clinical Applications in Neuroscience. Front Neuroinform 12:18 |
| Erdemir, Ahmet; Hunter, Peter J; Holzapfel, Gerhard A et al. (2018) Perspectives on Sharing Models and Related Resources in Computational Biomechanics Research. J Biomech Eng 140: |
