Publications by authors named "Sheng-Quan Xie"
Knee joint disorders pose a significant and growing challenge to global healthcare systems. Recent advancements in robotics, sensing technologies, and artificial intelligence have driven the development of robot-assisted therapies, reducing the physical burden on therapists and improving rehabilitation outcomes. This study presents a novel knee exoskeleton designed for safe and adaptive rehabilitation, specifically targeting bed-bound stroke patients to enable early intervention.
View Article and Find Full Text PDFThe widespread application of exoskeletons driven by soft actuators in motion assistance and medical rehabilitation has proven effective for patients who struggle with precise object grasping and suffer from insufficient hand strength due to strokes or other conditions. Repetitive passive flexion/extension exercises and active grasp training are known to aid in the restoration of motor nerve function. However, conventional pneumatic artificial muscles (PAMs) used for hand rehabilitation typically allow for bending in only one direction, thereby limiting multi-degree-of-freedom movements.
View Article and Find Full Text PDFArticle Synopsis
- Post-stroke upper limb dysfunction significantly reduces patients' quality of life, and using sEMG signals to predict motion intentions can improve rehabilitation outcomes by adjusting robotic assistance levels.
- This paper introduces four deep learning models focusing on muscle synergy (MS) feature extraction, with a notable model utilizing 3D Convolutional Neural Networks (3DCNN) that processes sEMG data from an anatomical perspective for enhanced accuracy.
- Experimental results show the 3DCNN model outperforms other models in predicting wrist motion, achieving impressive accuracy and efficiency metrics, and demonstrates superiority over traditional musculoskeletal and deep learning models.
Article Synopsis
- The study developed a portable system that uses red, green, and infrared lights to assess rehabilitation in stroke patients, capturing detailed physiological data through photoplethysmography (PPG).
- An advanced processing model called Multi-Channel Convolutional Neural Network-Long Short-Term Memory-Attention (MCNN-LSTM-Attention) was created to enhance data analysis by dynamically adjusting the significance of different input features.
- Clinical validation showed strong results with high accuracy and precision metrics from tests on both stroke patients and healthy volunteers, indicating the system's potential for improving stroke diagnosis and rehabilitation efforts.
Effectiveness of Intelligent Control Strategies in Robot-Assisted Rehabilitation-A Systematic Review.
IEEE Trans Neural Syst Rehabil Eng
May 2024
Article Synopsis
- - This review analyzes recent research (2015-2022) on intelligent control systems in robotics for physical rehabilitation, focusing on trends and effectiveness in improving rehabilitation outcomes.
- - Key findings show that intelligent algorithms enhance traditional control strategies, improve sensor mapping, and demonstrate better performance in accuracy and adaptability compared to non-intelligent controllers.
- - However, challenges include limited studies on impaired participants and a lack of universal evaluation criteria, making it hard to assess the overall effectiveness and compare different control systems.
Continuous Motion Intention Prediction Using sEMG for Upper-Limb Rehabilitation: A Systematic Review of Model-Based and Model-Free Approaches.
IEEE Trans Neural Syst Rehabil Eng
August 2024
Article Synopsis
- Upper limb impairments after a stroke greatly diminish patients' quality of life, highlighting the need for tailored robotic assistance during rehabilitation.
- This paper reviews 186 studies on predicting motion intentions of arm joints using Model-Based (MB) and Model-Free (MF) approaches, uncovering ongoing challenges related to subject diversity, algorithm reliability, and practical application.
- It recommends combining MB and MF strategies with advanced technologies like deep learning and muscle synergy features to enhance prediction accuracy and facilitate faster adaptation of algorithms for individual patients.
Simultaneous Hip Implant Segmentation and Gruen Landmarks Detection.
IEEE J Biomed Health Inform
November 2023
Article Synopsis
- The assessment of Total Hip Replacement (THR) complications currently relies on manual analysis of X-ray images, which is time-consuming and error-prone.
- The research introduces a new method that combines clinical knowledge with a Convolutional Neural Network (CNN) to enhance the segmentation of implant features in X-ray images.
- This integrated approach, utilizing a Statistical Shape Model (SSM) and multitask CNN, has improved the accuracy of implant shape estimation and facilitates automatic detection of clinical zones, increasing the dice score from 74% to 80%.
Article Synopsis
- Digital twins (DTs) in healthcare are increasingly gaining attention and offer innovative solutions for disease treatment and health enhancement as explored in this review.
- The study utilized Structure Topic Modeling (STM) to analyze 94 high-quality papers from 2018 to 2022, identifying key research focuses on technology development and the impact of current events like COVID-19.
- Key findings highlight a strong emphasis on real-time capabilities and the influence of emerging technologies, suggesting that future developments in DT will rely heavily on accurate algorithms and data transmission methods.
SSVEP-Based Brain-Computer Interface Controlled Robotic Platform With Velocity Modulation.
IEEE Trans Neural Syst Rehabil Eng
September 2023
Article Synopsis
- * Current SSVEP BCIs often use fixed movement speeds and directions, overlooking the user's need to adjust speed during operation.
- * This study introduces a method for controlling robotic arm speed based on the brightness of visual stimuli, validated through experiments showing improved efficiency and accuracy in reaching targets.
LSTM-AE for Domain Shift Quantification in Cross-Day Upper-Limb Motion Estimation Using Surface Electromyography.
IEEE Trans Neural Syst Rehabil Eng
June 2023
Article Synopsis
- Deep learning techniques have shown promise in controlling upper-limb myoelectric devices, but they struggle with consistency across different days due to variability in sEMG signals.
- A new method using a combination of CNN and LSTM networks, specifically LSTM-AE, is introduced to quantify the impact of domain shifts on model performance.
- Experimental results demonstrate a strong correlation between reconstruction errors from LSTM-AE and the accuracy of CNN-LSTM classification and regression tasks, highlighting the importance of monitoring these errors to improve system robustness over time.
Cross-Subject Transfer Learning for Boosting Recognition Performance in SSVEP-based BCIs.
IEEE Trans Neural Syst Rehabil Eng
March 2023
Article Synopsis
- Steady-state visual evoked potential (SSVEP) based brain-computer interfaces (BCIs) offer fast communication and high signal clarity, leading to increased research focus on their performance enhancement.
- This study introduces a new inter-subject transfer learning approach that uses templates and spatial filters to improve SSVEP recognition across different subjects by effectively utilizing auxiliary data.
- The method's effectiveness was tested using publicly available and self-collected datasets, confirming its potential for better SSVEP detection accuracy.
Physics-informed Deep Learning for Musculoskeletal Modelling: Predicting Muscle Forces and Joint Kinematics from Surface EMG.
IEEE Trans Neural Syst Rehabil Eng
December 2022
Article Synopsis
- Musculoskeletal models are essential for biomechanical analysis, helping to estimate movement variables like muscle forces and joint moments that are hard to measure directly.
- Traditional physics-based models are often slow, limiting their use in real-time applications, while emerging data-driven methods prioritize speed but fail to capture important neuromechanical processes.
- This paper presents a physics-informed deep learning framework that incorporates physics-based knowledge as soft constraints into a data-driven model, utilizing convolutional neural networks (CNNs) to effectively predict muscle forces and joint kinematics from surface electromyogram (sEMG) data, with successful experimental validation on multiple datasets.
Article Synopsis
- Robotic exoskeletons aim to replicate the torque and movement of healthy individuals, focusing on lightweight, portable designs to assist elderly users in daily activities by reducing power and mass requirements.
- The study introduces a multi-factor optimization technique for designing elastic elements in exoskeletons, aiming to match the torque-angle characteristics of healthy humans while providing adequate support to elderly users.
- Results showed that using optimized spring stiffness in parallel elastic actuators can significantly lower power and torque needs by up to 90%, resulting in a more efficient, lightweight exoskeleton that enhances the portability and usability for older adults.
Article Synopsis
- Spinal cord injury (SCI) and acquired brain injury (ABI) significantly impact individuals, causing disabilities and decreased quality of life, especially in low and middle-income countries like Nepal, where rehabilitative care is limited.
- The study implemented a telerehabilitation program connecting discharged patients with a multidisciplinary team via video conferencing, assessing its effectiveness and participant satisfaction.
- Results showed improved functional independence, reduced depression and anxiety, and enhanced quality of life for participants, revealing that telerehabilitation is a feasible and well-received approach in Nepal's healthcare setting.
Article Synopsis
- The article discusses the use of machine learning (ML) and deep learning (DL) techniques to help disabled individuals regain upper-limb functions by interpreting surface electromyography (sEMG) signals.
- It highlights the limitations of current ML/DL systems due to the complexity of upper-limb movements and the instability of sEMG signals, stressing the need for improved model robustness and reliability.
- The review categorizes recent advancements into multi-modal sensing fusion, transfer learning methods, and post-processing approaches, while also addressing challenges and opportunities in hardware, resources, and decoding strategies for future improvements.
Automatic detection and classification of peri-prosthetic femur fracture.
Int J Comput Assist Radiol Surg
April 2022
Article Synopsis
- - This study focuses on using deep learning (DL) models to classify and locate fractures near joint implants, a crucial aspect of computer-aided diagnosis (CAD).
- - Researchers analyzed a dataset of 1,272 annotated X-ray images of peri-prosthetic femur fractures (PFF), employing various classification and object detection models to assess their effectiveness.
- - The results indicated that the Resnet50 model performed the best in both binary and multi-class fracture classification, suggesting that this DL-based approach could be a promising tool for diagnosing such fractures.
A review of gait disorders in the elderly and neurological patients for robot-assisted training.
Disabil Rehabil Assist Technol
April 2020
Article Synopsis
- Ambulation is crucial for those with abnormal gaits, and exoskeleton robots can help them perform daily activities effectively.
- This article reviews the support needed by elderly individuals and those with neurological gait disorders, analyzing data from multiple research databases.
- The systematic review and meta-analysis indicate that while healthy elderly individuals show some gait changes with age, neurological patients exhibit significant deviations, highlighting the design needs for future robotic assistive devices.
Article Synopsis
- The study focuses on enhancing the detection of steady-state visual evoked potentials (SSVEP) in EEG signals for brain-computer interface (BCI) applications, highlighting the challenges posed by noise and artifacts in the data.
- An improved method called MEMD-CCA, which combines multivariate empirical mode decomposition and canonical correlation analysis, was tested on EEG signals from nine healthy volunteers.
- The findings showed that MEMD-CCA significantly outperformed traditional methods like CCA and TMSI, leading to increased accuracy in SSVEP recognition and demonstrating its potential for better BCI performance.
Article Synopsis
- - The study introduces a detailed musculoskeletal model of the ankle complex to improve understanding of how the ankle moves, diagnose disorders, and evaluate treatments, acknowledging the complexity due to the interactions between the ankle and subtalar joints.
- - The researchers developed this model focusing on the biaxial structure of the ankle, analyzing forces acting on ligaments and muscle-tendon units, and created a state space model for simulating the ankle dynamics, along with a graphical interface for visualizing anatomical information.
- - Validation of the model was conducted by comparing its results with existing literature and experimental data from a rehabilitation robot, showing a good match in terms of displacements and moments for both passive and active ankle motions.
Article Synopsis
- - The research highlights the importance of brain-computer interfaces (BCIs), especially for helping disabled individuals communicate, with a focus on steady state visual evoked potential (SSVEP) due to its effectiveness.
- - A new method based on multiple signal classification for extracting multi-dimensional SSVEP features was developed, achieving high accuracy in detecting idle states and reaching recognition rates of up to 100% in some conditions.
- - The experimental results demonstrated better frequency resolution and recognition accuracy compared to traditional methods, successfully allowing users to control a virtual keyboard in a real-world, unshielded setting.
A real-time computational model for estimating kinematics of ankle ligaments.
Comput Methods Biomech Biomed Engin
October 2016
Article Synopsis
- - The study emphasizes the importance of accurately assessing ankle ligament movement to better understand injuries and improve treatment, especially when integrated with robotic therapy.
- - A new computational model has been developed to evaluate ankle ligament kinematics in real-time, featuring three bone segments and twelve ligaments, using sensor inputs to track foot movement.
- - Validation of this model shows it aligns with existing anatomical data, indicating its potential use in robotic rehabilitation for optimizing training programs based on ligament kinematics.
Article Synopsis
- Ankle orientation and stiffness measurements are crucial for monitoring rehabilitation progress, but current assessment methods have limitations, especially those relying on manual manipulation.
- This study introduces a new ankle assessment technique utilizing position sensors and multi-axis load cells, suitable for both manual and automatic testing across multiple degrees-of-freedom.
- Preliminary evaluations with nine subjects showed the device's measurements of ankle orientation and stiffness were consistent with established methods and demonstrated high reliability, with strong statistical correlations.
Assessment of movement quality in robot- assisted upper limb rehabilitation after stroke: a review.
J Neuroeng Rehabil
September 2014
Article Synopsis
- - The paper reviews studies on stroke patients who underwent robot-assisted rehabilitation, focusing on kinematic parameters that reflect the quality of upper limb movement.
- - It highlights that while movement accuracy parameters are well-documented for chronic stroke patients, those related to sensorimotor control are more common in sub-acute patients, often lacking correlation with clinical assessments.
- - The authors suggest that the current assessments overlook coordinated movement and proximal upper limb functionality, recommending further research to better understand joint movement complexities in stroke rehabilitation.
Article Synopsis
- A decade of research has led to the development of various gait rehabilitation robots and control strategies, moving beyond traditional industrial robotic methods to ones specifically designed for rehabilitation needs.
- The assisted-as-needed (AAN) control strategy is highlighted as a significant advancement, supported by principles of motor learning and clinical evidence, making it a popular focus for researchers worldwide.
- This article reviews and analyzes control algorithms for gait rehabilitation robots, distinguishing between trajectory tracking and AAN control, while also exploring the potential for improving AAN strategies based on individual patient assessments.
Article Synopsis
- A new robotic orthosis using pneumatic muscle actuators (PMA) was created to help neurologically impaired individuals train on a treadmill.
- It features hip and knee movements controlled by PMA in an antagonistic setup, allowing natural trunk movements and foot clearance during walking.
- Testing showed the orthosis effectively supports physiological walking patterns, making it a valuable advancement in robotic assistance for rehabilitation.