Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 3 appraisals
JMIR human factors
Dynamic Balance Control and Postural Adaptation in Human-Robot Collaborative Manipulation: Within-Subject Experimental Study
BACKGROUND: The integration of robots into industrial settings has rapidly advanced, aiming to reduce human involvement in demanding tasks while improving overall efficiency. As collaborative robots (cobots) become more prevalent, assessing the physical strain during joint tasks is essential to promote long-term well-being in the workplace. OBJECTIVE: This study aimed to investigate how human-robot collaboration influences workers' postural control and musculoskeletal load during manipulation tasks performed in parallel. METHODS: Fourteen healthy male participants performed manipulation tasks under 3 conditions: without robotic assistance, with a cobot providing load support (Robot Free [RF]) and a cobot constrained to horizontal movement (Robot Plane [RP]). Center of pressure trajectories were computed, and nonlinear recurrence quantification analysis indicators (recurrence rate [REC], determinism [DET], and their ratio) were calculated in the anteroposterior, mediolateral, and anteroposterior-mediolateral planes. RESULTS: Statistical analysis showed greater postural sway in robot-assisted conditions compared to Free. Mean distance increased from 1.7 (SD 0.6) cm in Free to 2.4 (SD 0.6) cm in RF (P<.001) and 2.3 (SD 0.6) cm in RP (P<.001). Mean velocity increased from 2.9 (SD 0.9) cm/s in Free to 4.3 (SD 1.4) cm/s in RF and RP. Confidence ellipse area increased from 7.6 (SD 4.1) cm2 in Free to 24.9 (SD 14.2) cm2 in RF and 23.1 (SD 13.4) cm2 in RP. Sway area increased from 1.5 (SD 0.7) cm2/s in Free to 2.9 (SD 1.2) cm2/s in RF and RP. Nonlinear metrics revealed lower recurrence rates in robot-assisted conditions, decreasing from 0.31 (SD 0.08) in Free to 0.2 (SD 0.08) in RF and 0.2 (SD 0.04) in RP in the anteroposterior-mediolateral plane (P<.001), from 0.33 (SD 0.08) in Free to 0.28 (SD 0.07) in RF (P=.02) and 0.16 (SD 0.03) in RP (P=.007) in the mediolateral direction, and from 0.36 (SD 0.07) in Free to 0.3 (SD 0.06) in RF (P=.009) and 0.26 (SD 0.03) in RP (P<.001) in the anteroposterior direction. Determinism remained stable (values close to 1), leading to higher determinism-to-recurrence ratios for robot-assisted conditions, increasing from 3.41 (SD 0.87) in Free to 5.41 (SD 1.69) in RF and 5.51 (SD 1.11) in RP in the anteroposterior-mediolateral plane (P<.001), from 3.11 (SD 0.63) in Free to 3.64 (SD 0.72) in RF (P=.02) and 3.92 (SD 0.48) in RP (P=.007) in the mediolateral direction, and from 2.82 (SD 0.48) in Free to 3.47 (SD 0.57) in RF (P=.009) and 3.46 (SD 0.45) in RP (P<.001) in the anteroposterior direction. No significant differences were found between the robot-assisted conditions. CONCLUSIONS: Interaction increases postural sway, indicating reduced stability and higher physical demand. This could reflect impaired balance or adaptation. Nonlinear analysis reveals that postural control remains structured. Results also suggest that the mere presence of the cobot is the primary driver of these postural changes.
26 Apr 2026
Read appraisal →PloS one
Unsupervised industrial image defect detection based on autoencoder and GANs
In the process of intelligentization in modern manufacturing, especially in industrial fields such as automobile manufacturing, semiconductor production, and electronic product assembly, product quality control is crucial. Traditional defect detection methods face problems such as supervised learning methods relying on a large amount of labeled data, weak generalization ability, and high cost. In the process of intelligent manufacturing, industrial product quality control is a key link to ensure production safety and product consistency. Especially in typical industrial scenarios such as automobile manufacturing, semiconductor production, and electronic product assembly, traditional defect detection methods are difficult to meet the needs of actual production lines for high-precision and high-efficiency detection due to their reliance on a large amount of labeled data, weak generalization ability, and high cost. To solve the problem of defect detection in small and zero sample scenarios in these industries, improve detection sensitivity and localization accuracy, and enhance the model's generalization ability to unknown defects, a unsupervised industrial image defect detection method based on autoencoder and Generative Adversarial Networks (GANs) is proposed. This study constructs an Multi-level Deep feature Adaptive fusion AutoEncoder (MDAAE) module, extracts multi-scale features through Pre-Trained Convolutional Neural Backbone Networks (PTCNBN), introduces Attention Mechanism (AM) to dynamically calculate feature weights, and achieves feature fusion and reconstruction. Meanwhile, the self-AM is fused to improve the GANs. The self-attention module efficiently captures long-range dependencies and generates an adversarial network training objective function to optimize the generator's unsaturated loss. The results showed that the Area Under the Curve (AUC) of the Recall-Precision curve of the research method reached 93.6 ± 0.5%, and its F1-Score value exceeded 0.890 ± 0.003 in defect types such as scratches and dents. In practical applications, the inference time of the research method remained stable at 3.0 ± 0.2 GB of central processor memory, and the Fréchet Inception Distance (FID) value fully converged at the 2230th iteration, with a stable FID value of 3. The false alarm rate was only 8.6 ± 0.7% under strong light conditions. The proposed UIIDD method based on autoencoder and GANs had good robustness, generalization ability, reliability, and efficiency. This study effectively solves the problems of poor robustness, weak generalization ability, and high cost of traditional defect detection methods.
12 Apr 2026
Read appraisal →PloS one
A computer vision approach for the grading of cotton base load ages in measuring the performance of washing machine.
The performance testing standards for washing machines specify clear requirements regarding the age of the base load used. To enable non-contact detection of the service life of the test fabric and thereby improve the consistency of washing machine performance test results, this study employed computer vision techniques to investigate the feasibility of using image features of the base load for age grading. Base load samples underwent 1-100 accelerated washing cycles were categorized into five degradation stages (C1-C5 were used to represent base loads with ages of 1-20 cycles, 21-40 cycles, 41-60 cycles, 61-80 cycles, and 81-100 cycles, respectively). Wrinkle information and plain weave structure information were extracted from base load images, from which color, texture and area features were obtained. In addition, k-nearest neighbors (kNN), multilayer perceptron (MLP), linear discriminant analysis (LDA), and logistic regression (LR) classifiers were trained to grade the age of base load. As a result, LR classifier demonstrated robust overall performance, achieving accuracy of 0.75, 0.75, 0.62, 0.75, and 1.00 for C1, C2, C3, C4, and C5, respectively. Models utilizing exclusively plain weave structure-derived features consistently outperformed those using only wrinkle-derived features across all classifiers. These results validate computer vision as an effective tool for objective base load aging assessment, offering significant potential to streamline washing machine testing protocols and enhance sustainability. Future work can be focused on expanding sample sizes and exploring mobile-based implementation.
12 Apr 2026
Read appraisal →