Estimation of Muscle Forces of Lower Limbs Based on CNN-LSTM Neural Network and Wearable Sensor System

AI-Reinforced Wearable Sensors and Intelligent Point-of-Care Tests

Artificial intelligence (AI) techniques offer great potential to advance point-of-care testing (POCT) and wearable sensors for personalized medicine applications. This review explores the recent advances and the transformative potential of the use of AI in improving wearables and POCT. The integration of AI significantly contributes to empowering these tools and enables continuous monitoring, real-time analysis, and rapid diagnostics, thus enhancing patient outcomes and healthcare efficiency. Wearable sensors powered by AI models offer tremendous opportunities for precise and non-invasive tracking of physiological conditions that are essential for early disease detection and personalized treatments. AI-empowered POCT facilitates rapid, accurate diagnostics, making these medical testing kits accessible and available even in resource-limited settings. This review discusses the key advances in AI applications for data processing, sensor fusion, and multivariate analytics, highlighting case examples that exhibit their impact in different medical scenarios. In addition, the challenges associated with data privacy, regulatory approvals, and technology integrations into the existing healthcare system have been overviewed. The outlook emphasizes the urgent need for continued innovation in AI-driven health technologies to overcome these challenges and to fully achieve the potential of these techniques to revolutionize personalized medicine.

Enhancing postural balance assessment through neural network-based lower-limb muscle strength evaluation with reduced markers

Aiming to simplify the data acquisition process for balance diagnosis and focused on muscle, a direct factor affecting balance, to assess and judge postural stability. Utilizing a publicly available kinematic dataset, the research retained 3D coordinates and mechanical data for 8 markers on the lower limbs. By integrating this data with the musculoskeletal model in OpenSim, inverse kinematic calculations were performed to derive muscle forces. These forces, alongside the coordinates, were split into an 8:2 training and test set ratio. A neural network was then developed to predict muscle forces using normalized coordinate data from the training set as input, with corresponding muscle force data as training labels. The model's accuracy was confirmed on the test set, achieving coefficients of determination (R 2 ) above 0.99 for 276 muscle forces. Furthermore, the Force Maximum Percentage Difference (FMPD) was introduced as a novel criterion to evaluate and visualize lower limb balance, revealing significant discrepancies between the patient and control groups. This study successfully demonstrates that the neural network model can precisely predict lower limb muscle forces using reduced markers and introduces FMPD as an effective tool for assessing limb balance, providing a robust framework for future diagnostic and rehabilitative applications.

Optimization of Torque-Control Model for Quasi-Direct-Drive Knee Exoskeleton Robots Based on Regression Forecasting

The choice of torque curve in lower-limb enhanced exoskeleton robots is a key problem in the control of lower-limb exoskeleton robots. As a human-machine coupled system, mapping from sensor data to joint torque is complex and non-linear, making it difficult to accurately model using mathematical tools. In this research study, the knee torque data of an exoskeleton robot climbing up stairs were obtained using an optical motion-capture system and three-dimensional force-measuring tables, and the inertial measurement unit (IMU) data of the lower limbs of the exoskeleton robot were simultaneously collected. Nonlinear approximations can be learned using machine learning methods. In this research study, a multivariate network model combining CNN and LSTM was used for nonlinear regression forecasting, and a knee joint torque-control model was obtained. Due to delays in mechanical transmission, communication, and the bottom controller, the actual torque curve will lag behind the theoretical curve. In order to compensate for these delays, different time shifts of the torque curve were carried out in the model-training stage to produce different control models. The above model was applied to a lightweight knee exoskeleton robot. The performance of the exoskeleton robot was evaluated using surface electromyography (sEMG) experiments, and the effects of different time-shifting parameters on the performance were compared. During testing, the sEMG activity of the rectus femoris (RF) decreased by 20.87%, while the sEMG activity of the vastus medialis (VM) increased by 17.45%. The experimental results verify the effectiveness of this control model in assisting knee joints in climbing up stairs.