Clinical biomechanics (Bristol, Avon) · 2025
Wearable monitoring for rehabilitation: Deep learning-driven vertical ground reaction force estimation for anterior cruciate ligament reconstruction.
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.
Chen Tianxiao, Xu Datao, Wang Meizi, Zhou Zhifeng, Jie Tianle, Zhou Huiyu, Yuan Yi, Baker Julien S, Gao Zixiang, Gu Yaodong
Аннотация
BACKGROUND: Anterior cruciate ligament reconstruction (ACLR) can restore knee stability, yet many patients fail to regain pre-injury function or develop secondary injuries. Vertical ground reaction force (vGRF) reflects joint loading and recovery but is typically measured via lab-based force plates, limiting real-world use. Wearable sensors and deep learning could enable portable monitoring, but current studies lack accuracy in complex movements and patient-specific adaptations.
METHODS: Lower-limb kinematics and vGRF data from 25 ACLR patients during three daily activities (walking, running, descending stairs) was collected by wearable sensors and Vicon system. Three deep learning models were developed and optimized for the prediction tasks. The collected data was used to train the three developed models and the performance of each model was evaluated.
FINDINGS: Among the three deep learning models, CNN-BiGRU-Attention model demonstrated the best predictive performance across all three movement tasks (R2walking = 0.953 ± 0.006, R2running = 0.971 ± 0.005, R2descending stairs = 0.979 ± 0.003). Additionally, for the three selected daily activities, all models showed superior vGRF prediction performance in running and stair descending tasks compared to walking.
INTERPRETATION: By integrating data from wearable sensors with a hybrid deep learning framework, the proposed CNN-BiGRU-Attention model successfully achieved accurate estimation of vGRFs of ACLR patients in various movements. This provides a key technical reference for optimizing personalized rehabilitation strategies and improving patient outcomes, demonstrating significant clinical application value and social benefits.
METHODS: Lower-limb kinematics and vGRF data from 25 ACLR patients during three daily activities (walking, running, descending stairs) was collected by wearable sensors and Vicon system. Three deep learning models were developed and optimized for the prediction tasks. The collected data was used to train the three developed models and the performance of each model was evaluated.
FINDINGS: Among the three deep learning models, CNN-BiGRU-Attention model demonstrated the best predictive performance across all three movement tasks (R2walking = 0.953 ± 0.006, R2running = 0.971 ± 0.005, R2descending stairs = 0.979 ± 0.003). Additionally, for the three selected daily activities, all models showed superior vGRF prediction performance in running and stair descending tasks compared to walking.
INTERPRETATION: By integrating data from wearable sensors with a hybrid deep learning framework, the proposed CNN-BiGRU-Attention model successfully achieved accurate estimation of vGRFs of ACLR patients in various movements. This provides a key technical reference for optimizing personalized rehabilitation strategies and improving patient outcomes, demonstrating significant clinical application value and social benefits.