Computer methods and programs in biomedicine · 2026
A computer vision-based modeling pipeline to facilitate efficient video-to-strain analysis of human tibia during dynamic gait.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Xu Xin-Sheng, Zhang Jia-Xing, Guo Xuan, Lee Jung-Min, Ma Xin, Chen Wen-Ming
Аннотация
BACKGROUND AND OBJECTIVE: Tibial stress and strain are crucial indicators of bone health and injury risk, yet their estimation with conventional marker-based motion capture systems is time-consuming and constrained to multi-camera laboratory setups, limiting broader applicability.
METHODS: We propose a novel video-to-strain analysis (V2S) method, enabling subject-specific musculoskeletal modeling and tibial stress/strain computation from single-view smartphone recordings of treadmill walking. Markerless motion was reconstructed using SMPL-based fitting with HuMoR, a deep generative human motion prior, together with PlaneRCNN ground-plane estimation and contact and bilateral symmetry constraints to yield more stable 3D pose and gait reconstruction from single-view camera. Feasibility was evaluated by comparing kinematics, muscle forces, and finite element-computed tibial stress/strain against a marker-based reference workflow.
RESULTS: Compared to the marker-based method, the proposed markerless method yielded joint kinematics with R2 = 0.86 and muscle forces with R2 = 0.81. Relative to the baseline HuMoR model, our approach achieved lower kinematic errors (hip, knee, and ankle RMSE of 3.6°, 4.8°, and 4.0°, compared with 9.1°, 11.6°, and 12.3° for HuMoR). The model-computed results with peak von Mises stress varying by 3.8% and the maximum principal strain differing by 11.9%, with the largest discrepancies occurring in stance-phase.
CONCLUSIONS: Single-view markerless approach offers a practical alternative for efficient motion and bone strain analysis, where marker-based motion capture is not feasible.
METHODS: We propose a novel video-to-strain analysis (V2S) method, enabling subject-specific musculoskeletal modeling and tibial stress/strain computation from single-view smartphone recordings of treadmill walking. Markerless motion was reconstructed using SMPL-based fitting with HuMoR, a deep generative human motion prior, together with PlaneRCNN ground-plane estimation and contact and bilateral symmetry constraints to yield more stable 3D pose and gait reconstruction from single-view camera. Feasibility was evaluated by comparing kinematics, muscle forces, and finite element-computed tibial stress/strain against a marker-based reference workflow.
RESULTS: Compared to the marker-based method, the proposed markerless method yielded joint kinematics with R2 = 0.86 and muscle forces with R2 = 0.81. Relative to the baseline HuMoR model, our approach achieved lower kinematic errors (hip, knee, and ankle RMSE of 3.6°, 4.8°, and 4.0°, compared with 9.1°, 11.6°, and 12.3° for HuMoR). The model-computed results with peak von Mises stress varying by 3.8% and the maximum principal strain differing by 11.9%, with the largest discrepancies occurring in stance-phase.
CONCLUSIONS: Single-view markerless approach offers a practical alternative for efficient motion and bone strain analysis, where marker-based motion capture is not feasible.