Hammond Claire V, Henninger Heath B, Fregly Benjamin J, Gustafson Jonathan A
Department of Mechanical Engineering, Rice University, Houston, TX, United States.
Departments of Orthopaedics and Biomedical Engineering, The University of Utah, Salt Lake City, UT, United States.
Front Bioeng Biotechnol. 2025 Aug 11;13:1547373. doi: 10.3389/fbioe.2025.1547373. eCollection 2025.
The shoulder joint complex is prone to musculoskeletal issues, such as rotator cuff-related pain, which affect two-thirds of adults and often result in suboptimal treatment outcomes. Current musculoskeletal models used to understand shoulder biomechanics are limited by challenges in personalization, inaccuracies in predicting joint and muscle loads, and an inability to simulate anatomically accurate motions. To address these deficiencies, we developed a novel, personalized modeling framework capable of calibrating subject-specific joint centers and functional axes for the shoulder complex.
We developed a novel personalized modeling framework utilizing the Joint Model Personalization (JMP) Tool from the Neuromusculoskeletal Modeling Pipeline, incorporating in vivo biplane fluoroscopy data of the glenohumeral and scapulothoracic joints. Initially, open-chain scapula-only models with 3, 4, and 5 degrees of freedom (DOFs) were created and optimized using synthetic marker data derived from subject-specific geometry. Subsequently, closed-chain shoulder models including scapula, clavicle, and humerus were constructed and optimized through a two-stage personalization approach. Model accuracy and generalizability were assessed using marker distance errors and leave-one-out cross-validation across multiple shoulder motions.
Increasing the number of scapula DOFs in open-chain models improved kinematic accuracy, with the 5 DOF scapula model yielding the lowest marker distance errors (average: 0.8 mm; maximum: 5.2 mm). The closed-chain shoulder model demonstrated high accuracy (average: 0.9 mm; maximum: 5.7 mm) and consistency across subject in cross-validation tests (average marker distance errors = 1.0-1.4 mm). Models personalized with synthetic noise representative of skin-based marker data resulted in slightly increased, yet acceptable marker errors (average: 3.4 mm).
Our personalized, closed-chain shoulder modeling framework significantly improves the accuracy and anatomical fidelity of shoulder kinematic simulations compared to existing approaches. This framework minimizes errors in joint kinematics and provides a foundation for future models incorporating personalized musculature and advanced simulations.
肩关节复合体容易出现肌肉骨骼问题，如肩袖相关疼痛，影响三分之二的成年人，且常常导致治疗效果欠佳。当前用于理解肩部生物力学的肌肉骨骼模型受到个性化挑战、预测关节和肌肉负荷不准确以及无法模拟解剖学精确运动的限制。为解决这些不足，我们开发了一种新颖的个性化建模框架，能够为肩部复合体校准特定受试者的关节中心和功能轴。
我们利用神经肌肉骨骼建模管道中的关节模型个性化（JMP）工具开发了一种新颖的个性化建模框架，纳入了肩肱关节和肩胛胸壁关节的体内双平面荧光透视数据。最初，使用从特定受试者几何形状导出的合成标记数据创建并优化了具有3、4和5个自由度（DOF）的仅肩胛骨开链模型。随后，通过两阶段个性化方法构建并优化了包括肩胛骨、锁骨和肱骨的闭链肩部模型。使用标记距离误差和对多个肩部运动进行留一法交叉验证来评估模型的准确性和通用性。
在开链模型中增加肩胛骨自由度提高了运动学准确性，5自由度肩胛骨模型产生的标记距离误差最低（平均：0.8毫米；最大：5.2毫米）。闭链肩部模型在交叉验证测试中表现出高精度（平均：0.9毫米；最大：5.7毫米）和受试者间的一致性（平均标记距离误差 = 1.0 - 1.4毫米）。用代表基于皮肤的标记数据的合成噪声进行个性化的模型导致标记误差略有增加，但仍可接受（平均：3.4毫米）。
与现有方法相比，我们的个性化闭链肩部建模框架显著提高了肩部运动学模拟的准确性和解剖逼真度。该框架将关节运动学误差降至最低，并为未来纳入个性化肌肉组织和高级模拟的模型奠定了基础。
