Frontiers in physiology · 2026
IMU-based rotator cuff injury recognition with varying configurations and combinations.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
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.
Xiong Wanwan, Zeng Xianwu, Sun Jianning, Zhang Yuexin, Pei Chun, Chan Yi-Hao, Chen Peng, Ma Yunru
Аннотация
INTRODUCTION: Assessing functional movements is important for evaluating shoulder impairments, as these movements directly reflect patients' capacity to perform daily activities. Intelligent rotator cuff injury (RCI) recognition is beneficial for clinical management.
METHODS: This study aimed to develop accurate and cost-effective RCI recognition models by integrating machine learning (ML)/deep learning (DL) algorithms with upper limb kinematic data collected in functional movement, and to explore optimal motion task combinations and model configurations for clinical practice. A total of 102 participants were prospectively enrolled, comprising 51 patients diagnosed with RCI, 25 patients diagnosed with adhesive capsulitis and 26 healthy volunteers. Patients with adhesive capsulitis were included to assess model's ability to distinguish RCI from other shoulder disorders that present with similar function limitations. Each participant performed six functional movements simulating daily activities and two shoulder range of motion (ROM) tests. Upper limb kinematic data were collected by inertial measurement units (IMUs) and analyzed via eight classification models, including the proposed DL based RCI recognition model (RCIRNet) and seven conventional ML models (e.g., KNN, SVM, DT, RF, NB, AdaBoost, and XGBoost). Given the relatively small sample size, data augmentation was applied to mitigate overfitting and improve model generalization. Model performance was evaluated across single and combined motion tasks via Accuracy, Precision, Recall, F1-score, and AUC with its 95% confidence interval (CI).
RESULTS: Results demonstrated that shoulder frontal ROM test alone achieved the best performance with RCIRNet, yielding an Accuracy of 0.89, F1-score of 0.89, and AUC of 0.93 (95%CI, 0.88-0.98). Combining four tasks increased Recall to 1.0 with high level Accuracy, F1-score, and AUC. However, adding more movements (5-8 tasks) did not improve model performance. Among different model configurations, RCIRNet performed best, SVM demonstrated overall stability, and KNN excelled in Recall.
DISCUSSION: It is concluded that selecting discriminant movements is more critical for effective RCI recognition than simply increasing the quantity of movement tasks. The shoulder frontal ROM test combined with RCIRNet provides an optimal approach for accurate and efficient clinical screening and rehabilitation monitoring.
METHODS: This study aimed to develop accurate and cost-effective RCI recognition models by integrating machine learning (ML)/deep learning (DL) algorithms with upper limb kinematic data collected in functional movement, and to explore optimal motion task combinations and model configurations for clinical practice. A total of 102 participants were prospectively enrolled, comprising 51 patients diagnosed with RCI, 25 patients diagnosed with adhesive capsulitis and 26 healthy volunteers. Patients with adhesive capsulitis were included to assess model's ability to distinguish RCI from other shoulder disorders that present with similar function limitations. Each participant performed six functional movements simulating daily activities and two shoulder range of motion (ROM) tests. Upper limb kinematic data were collected by inertial measurement units (IMUs) and analyzed via eight classification models, including the proposed DL based RCI recognition model (RCIRNet) and seven conventional ML models (e.g., KNN, SVM, DT, RF, NB, AdaBoost, and XGBoost). Given the relatively small sample size, data augmentation was applied to mitigate overfitting and improve model generalization. Model performance was evaluated across single and combined motion tasks via Accuracy, Precision, Recall, F1-score, and AUC with its 95% confidence interval (CI).
RESULTS: Results demonstrated that shoulder frontal ROM test alone achieved the best performance with RCIRNet, yielding an Accuracy of 0.89, F1-score of 0.89, and AUC of 0.93 (95%CI, 0.88-0.98). Combining four tasks increased Recall to 1.0 with high level Accuracy, F1-score, and AUC. However, adding more movements (5-8 tasks) did not improve model performance. Among different model configurations, RCIRNet performed best, SVM demonstrated overall stability, and KNN excelled in Recall.
DISCUSSION: It is concluded that selecting discriminant movements is more critical for effective RCI recognition than simply increasing the quantity of movement tasks. The shoulder frontal ROM test combined with RCIRNet provides an optimal approach for accurate and efficient clinical screening and rehabilitation monitoring.