Personalized musculoskeletal models for assessment of dynamic patellofemoral function

膝盖中髌骨和股骨之间的关节是由于运动或职业活动导致的肌肉骨骼损伤和功能障碍的常见部位。该关节的功能和功能障碍仍然没有得到很好的理解，这是由于使用传统方法测量髌骨在皮肤下的运动的困难，例如使用皮肤安装的反射标记的光学运动捕获。计算模型为克服这些挑战提供了一个有吸引力的选择，因为它们能够估计在活体人类中难以或不可能获得的测量值，并提供了一种隔离个体因素对关节功能影响的方法。我的研究计划的长期目标是使用个性化的肌肉骨骼模型，以提高我们对动态髌股功能和功能障碍的个体因素的影响的理解。目前，生成个性化的肌肉骨骼模型是一个时间和成本密集型的过程，其使用仅限于实验室环境，从而限制了参与者的数量和可以研究的活动类型。因此，该提案的重点是开发和验证方法，以提高个性化肌肉骨骼模型的可访问性和适用性。这将通过两个具体目标来实现：1）开发和验证方法，使用具有稀疏成像和测量数据的机器学习来生成膝关节几何形状的个性化模型，以及2）通过将基于视频的无标记运动捕获和计算模拟框架内的外力估计集成在一起，实现肌肉骨骼模型的“野外”使用。人们越来越认识到，个性化的生物力学模型是更准确的，由于可变的性质的肌肉骨骼系统。这项研究将简化个性化肌肉骨骼模型的生成，并使其能够用于更广泛的应用。这将有助于研究更多的人进行各种活动，以更好地捕捉髌股解剖结构中存在的变异性，并将更好地代表那些来自不同群体的人，这些人目前没有被通用模型很好地捕捉到。这对于提高我们对髌股关节功能和功能障碍的理解至关重要，因为该关节的个体性质以及动态运动和职业活动中经常出现损伤的事实。生成大型数据集的能力也将允许使用机器学习和数据分析来揭示髌股功能的模式和表型。这些类型的进步也可以成为开发工具的基础，这些工具可以在诊所中实施，以帮助诊断和规划患者特定的干预措施。最后，参与这项研究将提供HQP的经验和技能，这将是有价值的，无论他们追求的职业生涯在工业或学术界。

The articulation between the patella and femur in the knee is a common site of musculoskeletal injury and dysfunction as a result of athletic or occupational activities. The function and dysfunction of this joint are still not well understood due to the difficulty of measuring the movement of the patella beneath the skin using traditional methods, such as optical motion capture which uses skin mounted reflective markers. Computational models provide an attractive option for overcoming these challenges as they enable estimation of measures that are difficult or impossible to obtain in living humans and provide a means to isolate the impact of individual factors on joint function. The long-term goal of my research program is to use personalized musculoskeletal models to improve our understanding of the influence of individual factors on dynamic patellofemoral function and dysfunction. Currently, generating personalized musculoskeletal models is a time- and cost-intensive process and their use is confined to the laboratory setting, thus limiting the number of participants and types of activities that can be studied. Therefore, this proposal is focused on the development and validation of methods to improve the accessibility and applicability of personalized musculoskeletal models. This will be achieved through two specific aims: 1) develop and validate methods to generate personalized models of knee geometry using machine learning with sparse imaging and measurement data, and 2) enable the use of musculoskeletal models `in the wild' by integrating video-based markerless motion capture and estimations of external forces within computational simulation frameworks. It is increasingly being recognized that personalized biomechanical models are more accurate due to the variable nature of the musculoskeletal system. This research will simplify the generation of personalized musculoskeletal models and enable their use for a wider range of applications. This would facilitate the study of much greater numbers of people performing a variety of activities to better capture the variability that exists in patellofemoral anatomy and would better represent those from diverse groups who are currently not well captured by generic models. This is vital for improving our understanding of patellofemoral function and dysfunction because of the individual nature of this joint and the fact that injuries often arise in dynamic athletic and occupational activities. The ability to generate large datasets would also allow for the use of machine learning and data analytics to reveal patterns and phenotypes in patellofemoral function. These types of advances could also form the basis for the development of tools that can be implemented in clinics to aid in diagnoses and the planning of patient-specific interventions. Finally, participation in this research will provide HQP with experiences and skills that will be valuable whether they pursue careers in industry or academia.