Search Results (28,382)

Traditional academic monitoring practices rely on retrospective data analysis, generally identifying at-risk students too late to take meaningful action. To address this, this paper proposes a real-time, rule-based decision support system designed to increase student achievement by early detection of disengagement, meeting the growing demand for prompt academic intervention in online and blended learning contexts. The study uses the Open University Learning Analytics Dataset (OULAD), comprising over 32,000 students and millions of virtual learning environment (VLE) interaction records, to simulate weekly assessments of engagement through clickstream activity. Students were flagged as “at risk” if their participation dropped below defined thresholds, and these flags were associated with assessment performance and final course results. The system demonstrated 72% precision and 86% recall in identifying failing and withdrawn students as major alert contributors. This lightweight, replicable framework requires minimal computing power and can be integrated into existing LMS platforms. Its visual and statistical validation supports its role as a scalable, real-time early warning tool. The paper recommends integrating real-time engagement dashboards into institutional LMS and suggests future research explore hybrid models combining rule-based and machine learning approaches to personalize interventions across diverse learner profiles and educational contexts. Full article

Accurate extraction of building features from remotely sensed data is essential for supporting research and applications in urban planning, land management, transportation infrastructure development, and disaster monitoring. Despite the prominence of deep learning as the state-of-the-art (SOTA) methodology for building extraction, substantial challenges remain, largely stemming from the diversity of building structures and the complexity of background features. To mitigate these issues, this study introduces DeepSwinLite, a lightweight architecture based on the Swin Transformer, designed to extract building footprints from very high-resolution (VHR) imagery. The model integrates a novel local–global attention module to enhance the interpretation of objects across varying spatial resolutions and facilitate effective information exchange between different feature abstraction levels. It comprises three modules: multi-scale feature aggregation (MSFA), improving recognition across varying object sizes; multi-level feature pyramid (MLFP), fusing detailed and semantic features; and AuxHead, providing auxiliary supervision to stabilize and enhance learning. Experimental evaluations on the Massachusetts and WHU Building Datasets reveal the superior performance of DeepSwinLite architecture when compared to existing SOTA models. On the Massachusetts dataset, the model attained an OA of 92.54% and an IoU of 77.94%, while on the WHU dataset, it achieved an OA of 98.32% and an IoU of 92.02%. Following the correction of errors identified in the Massachusetts ground truth and iterative enhancement, the model’s performance further improved, reaching 94.63% OA and 79.86% IoU. A key advantage of the DeepSwinLite model is its computational efficiency, requiring fewer floating-point operations (FLOPs) and parameters compared to other SOTA models. This efficiency makes the model particularly suitable for deployment in mobile and resource-constrained systems. Full article

Ovarian cancer remains a significant global health concern, and its diagnosis heavily relies on whole-slide images (WSIs). Due to their gigapixel spatial resolution, WSIs must be split into patches and are usually modeled via multi-instance learning (MIL). Although previous studies have achieved remarkable performance comparable to that of humans, in clinical practice WSIs are distributed across multiple hospitals with strict privacy restrictions, necessitating secure, efficient, and effective federated MIL. Moreover, heterogeneous data distributions across hospitals lead to model heterogeneity, requiring a framework flexible to both data and model variations. This paper introduces HFed-MIL, a heterogeneous federated MIL framework that leverages gradient-based attention distillation to tackle these challenges. Specifically, we extend the intuition of Grad-CAM to the patch level and propose Patch-CAM,which computes gradient-based attention scores for each patch embedding, enabling structural knowledge distillation without explicit attention modules while minimizing privacy leakage. Beyond conventional logit distillation, we designed a dual-level objective that enforces both class-level and structural-level consistency, preventing the vanishing effect of naive averaging and enhancing the discriminative power and interpretability of the global model. Importantly, Patch-CAM scores provide a balanced solution between privacy, efficiency, and heterogeneity: they contain sufficient information for effective distillation (with minimal membership inference risk, MIA AUC ≈ 0.6) while significantly reducing communication cost (0.32 MB per round), making HFed-MIL practical for real-world federated pathology. Extensive experiments on multiple cancer subtypes and cross-domain datasets (Camelyon16, BreakHis) demonstrate that HFed-MIL achieves state-of-the-art performance with enhanced robustness under heterogeneity conditions. Moreover, the global attention visualizations yield sharper and clinically meaningful heatmaps, offering pathologists transparent insights into model decisions. By jointly balancing privacy, efficiency, and interpretability, HFed-MIL improves the practicality and trustworthiness of deep learning for ovarian cancer WSI analysis, thereby increasing its clinical significance. Full article

Smart logistics systems generate massive amounts of data, such as images and videos, requiring real-time processing in edge clusters. However, the edge cluster systems face performance bottlenecks in reception and forwarding high-concurrency data streams from numerous smart terminals, resulting in degraded processing efficiency. To address this issue, a novel high-performance data stream model called CBPS-DPDK is proposed. CBPS-DPDK integrates the DPDK framework from Intel corporations with a content-based publish/subscribe model enhanced by semantic filtering. This model adopts a three-tier optimization architecture. First, the user-space data plane is restructured using DPDK to avoid kernel context switch overhead via zero-copy and polling. Second, semantic enhancement is introduced into the publish/subscribe model to reduce the coupling between data producers and consumers through subscription matching and priority queuing. Finally, a hierarchical load balancing strategy ensures reliable data transmission under high concurrency. Experimental results show that CBPS-DPDK significantly outperforms two baselines—OSKT (kernel-based data forwarding) and DPDK-only (DPDK). Relative to the OSKT baseline, DPDK-only achieves improvements of 37.5% in latency, 11.1% in throughput, and 9.1% in VMAF; CBPS-DPDK further increases these to 51.8%, 18.3%, and 11.2%, respectively. In addition, compared with the traditional publish–subscribe system NATS, CBPS-DPDK maintains lower delay, higher throughput, and more balanced CPU and memory utilization under saturated workloads, demonstrating its effectiveness for real-time, high-concurrency edge scenarios. Full article

In recent years, tail-sitter unmanned aerial vehicles (UAVs), capable of vertical take-off and landing (VTOL) and long-range flight, have garnered extensive attention. However, the challenge of yaw control, particularly for large-scale UAVs, has become a significant obstacle. It is challenging to generate sufficient yaw moments by motor differential thrust without compromising control authority in other channels or increasing mechanical complexity. Therefore, this paper proposes the concept of blown yaw, which utilizes the high-velocity airflow over rudders, induced by the propellers slipstream, to enhance the yaw control torque actively. An over-actuated, hundred-kilogram-class, tail-sitter UAV is designed to validate the effectiveness of the proposed method. To address the control allocation problem introduced by the implementation of blown yaw, an optimization-based control allocation module is developed, capable of precisely mapping the required forces and torques to all actuators. The proposed method, combined with computational fluid dynamics (CFD) simulations, accounts for the propeller model and the significant differences in actuator effectiveness across various flight conditions. Simulation results demonstrate that the proposed blown-yaw method significantly enhances the yaw control performance, achieving an overall energy savings of approximately 8.0% and a 60% reduction in the mean-squared error. Furthermore, the method exhibits robust performance against variations in control parameters and external disturbances. Full article

Moons tidally interact with their host planets and stars. A close moon is quickly synchronised by the planet or becomes captured in a higher spin–orbit resonance. However, the planet requires much more time to significantly alter its rotation rate under the influence of moon-generated tides. The situation becomes more complex for close-in planets, as star-generated tides come into play and compete with moon-generated tides. The synchronisation of the planet by its moon changes the tidal dynamics of the entire star–planet–moon system and can lead to long-term stable configurations. In this paper, we demonstrate that a certain initial condition must be met for this to occur. Based on the angular momentum conservation, the derived condition is universal and bears no dependence upon the planet’s internal structure or tidal dissipation model. It is applicable to dwindling systems as well as to tidally expanding orbits and cases of initially retrograde motion. We present calculations for specific planet–moon systems (Earth and the Moon; Neptune and Triton; Venus and its hypothetical presently extinct moon Neith; Mars, Phobos, and Deimos; and Pluto and Charon) to constrain dynamically plausible formation and evolution scenarios. Among other things, our analysis prompts the question of whether Pluto and Charon evolved into their current state from an initially more compact configuration (as is commonly assumed) or from a wider orbit—a topic that will be discussed at length elsewhere. Our results are equally applicable to exoplanets. For example, if asynchronous close-in exoplanets are detected, the possibility of tidal synchronisation by an exomoon should be considered. Full article

This study examines the deformation and failure mechanisms of two reservoir bank landslides: the traction-type Baijiabao landslide and the translational Baishuihe landslide. Based on long-term monitoring data and a hydro-mechanical coupled numerical model of rainfall infiltration, we investigate the impact of crack depth on landslide stability. Results show that the Baishuihe landslide exhibits translational failure, initiated at the rear by tension cracks and rear subsidence, followed by toe uplift, whereas the Baijiabao landslide displays traction-type progressive failure, starting with toe erosion and later developing rear-edge cracks. Rainfall induces similar seepage patterns in both landslides, with infiltration concentrated at the crest, toe, and convex terrain areas. As crack depth increases, soil saturation near the cracks decreases nonlinearly, while the base remains saturated. However, displacement responses differ: Traction-type landslides exhibit opposing lateral movements with minimal vertical displacement. In contrast, translational landslides show displacement increasing with crack depth, dominated by gravity. These findings guide targeted mitigation: traction-type landslides require crack control and toe protection, while translational landslides need measures to block thrust transfer and monitor deep slip surfaces. This study offers new insights into the effect of crack depth on landslide stability, contributing to improved landslide hazard assessment and management. Full article

As low-voltage distribution networks incorporate increasingly diverse loads, series arc faults exhibit weak characteristics that are easily masked by load currents, leading to high misjudgment rates in traditional detection methods. This paper proposes a series arc fault detection method based on an improved Light Gradient Boosting Machine (LightGBM) model. First, a test platform containing 12 household loads was built to collect arc data from both individual and composite loads. Composite loads refer to composite load conditions where multiple devices are running simultaneously and arcing occurs on some loads. To address the challenge of feature extraction, Variational Mode Decomposition (VMD) is employed to isolate the fundamental frequency component. To enhance high-frequency arc characteristics, singular value decomposition (SVD) is then applied. A multidimensional statistical feature set—comprising peak-to-peak value, kurtosis, and other indicators—is constructed. Finally, the LightGBM algorithm is used to identify arc faults based on these features. To overcome the LightGBM model’s limited ability to focus on hard-to-classify samples, a dynamic weighted hybrid loss function is developed. Experiments demonstrate that the proposed method achieves 98.9% accuracy across 223,615 sample groups. When deployed on STM32H723VGT6 hardware, the average fault alarm time is 83.8 ms, meeting requirements. Full article

Accurately predicting future activities in egocentric (first-person) videos is a challenging yet essential task, requiring robust object recognition and reliable forecasting of action patterns. However, the limited number of observable frames in such videos often lacks critical semantic context, making long-term predictions particularly difficult. Traditional approaches, especially those based on recurrent neural networks, tend to suffer from cumulative error propagation over extended time steps, leading to degraded performance. To address these challenges, this paper introduces a novel framework, Virtual Frame-Augmented Guided Forecasting (VFGF), designed specifically for long-term egocentric activity prediction. The VFGF framework enhances semantic continuity by generating and incorporating virtual frames into the observable sequence. These synthetic frames fill the temporal and contextual gaps caused by rapid changes in activity or environmental conditions. In addition, we propose a Feature Guidance Module that integrates anticipated activity-relevant features into the recursive prediction process, guiding the model toward more accurate and contextually coherent inferences. Extensive experiments on the EPIC-Kitchens dataset demonstrate that VFGF, with its interpolation-based temporal smoothing and feature-guided strategies, significantly improves long-term activity prediction accuracy. Specifically, VFGF achieves a state-of-the-art Top-5 accuracy of 44.11% at a 0.25 s prediction horizon. Moreover, it maintains competitive performance across a range of long-term forecasting intervals, highlighting its robustness and establishing a strong foundation for future research in egocentric activity prediction. Full article

Black fungus cultivation in bagged form requires frequent inspection of mycelial growth, a process that is labor-intensive and susceptible to subjective judgment. In addition, timely detection of contamination in low-light and high-humidity environments remains a significant challenge. To address these issues, this paper proposed an intelligent agriculture system for black fungus cultivation, with emphasis on practical deployment under real farming conditions. The system integrates a private 5G network with a YOLOv8-based deep learning model for real-time object detection and growth monitoring. Continuous image acquisition and data feedback are achieved through a multi-parameter environmental sensing module and an autonomous ground vehicle (AGV) equipped with IP cameras. To improve model robustness, more than 42,000 labeled training images were generated through data augmentation, and a modified C2f network architecture was employed. Experimental results show that the model achieved a detection accuracy of 93.7% with an average confidence score of 0.96 under live testing conditions. The deployed 5G network provided a downlink throughput of 645.2 Mbps and an uplink throughput of 147.5 Mbps, ensuring sufficient bandwidth and low latency for real-time inference and transmission. Field trials conducted over five cultivation batches demonstrated improvements in disease detection, reductions in labor requirements, and an increase in the average yield success rate to 80%. These findings indicate that the proposed method offers a scalable and practical solution for precision agriculture, integrating next-generation communication technologies with deep learning to enhance cultivation management. Full article

Purpose: This systematic review explores the ethical, legal, and regulatory frameworks governing the deployment of artificial intelligence technologies in radiology. It aims to identify key governance challenges and highlight strategies that promote the safe, transparent, and accountable integration of artificial intelligence in clinical imaging. This review is intended for both medical practitioners and AI developers, offering clinicians a synthesis of ethical and legal considerations while providing developers with regulatory insights and guidance for future AI system design. Methods: A systematic review was conducted, examining thirty-eight peer-reviewed articles published between 2018 and 2025. Studies were identified through searches in PubMed, Scopus, and Embase using terms related to artificial intelligence, radiology, ethics, law, and regulation. The inclusion criteria focused on studies addressing governance implications, rather than technical design. A thematic content analysis was applied to identify common patterns and gaps across ethical, legal, and regulatory domains. Results: The findings reveal widespread radiology-specific concerns, including algorithmic bias in breast and chest imaging datasets, opacity in image-based AI systems such as pulmonary nodule detection models, and unresolved legal liability in cases where radiologists rely on FDA-cleared AI tools that fail to identify abnormalities. Regulatory frameworks vary significantly across regions with limited global harmonization, highlighting the need for adaptive oversight models and improved data governance. Conclusion: Responsible deployment of AI in radiology requires governance models that address bias, explainability, and medico-legal accountability while integrating ethical principles, legal safeguards, and adaptive oversight. This review provides tailored insights for medical practitioners, AI developers, policymakers, and researchers: clinicians gain guidance on ethical and legal responsibilities, developers on regulatory and design priorities, policymakers (especially in the Middle East) on regional framework gaps, and researchers on future directions. Full article

Recent decades have been characterized by increasing numbers of bone tissue injuries and diseases resulting in the formation of bone defects. The number of such bone defects has also grown due to active surgical approaches implemented after surgical interventions for oncological, infectious, and dystrophic bone lesions. To repair such bone defects requires the use of bone tissue substitutes. Nowadays, constructs based on matrices of various compositions and structures, supplemented with the addition of biologically active components (including growth factors and cells), are the most promising approaches used in bone tissue engineering. The properties of the matrices are of the utmost importance in construct formation. This work presents the results of a comprehensive study of matrices of various natures intended for the formation of complex constructs for bone tissue engineering. Using a set of methods for studying the physical, mechanical, and biological characteristics, the total and associated porosity of the studied matrices, the structure, the mechanical parameters, and the level of cytotoxicity and cytocompatibility were determined. It was shown that all the studied materials were not cytotoxic (cytotoxicity rank of all matrices = 0–1). All matrices were porous, but samples of materials of biological origin had large pores ranging in size from 100 to 1000 μm, and pores of the hybrid polymer were sized from 0.1 to 100 μm. Total and open porosity ranged from 89% and 79% for the allogeneic matrix up to 67% and 48% for the hybrid polymer, respectively, while the σ values (compressive stress at break) of samples of all studied materials were close to each other. When human test culture MSCs interact with samples of these materials, it was shown that the cells adhere to the surface and structure of all materials and retain typical morphology, while also demonstrating the ability to proliferate and migrate along the surface and into the matrix structure, i.e., all materials are cytocompatible. Based on the data obtained, it can be assumed that all the studied matrices can be used for model biomedical studies and as a basis for constructs for bone tissue engineering. An adequate choice of research method at the earliest stages of the development of each material will ensure the most effective approaches for further work and subsequent use of this product. Full article

In high-humidity environments, the epoxy resin solid insulation materials of high-frequency transformers are prone to aging, resulting in varying degrees of deterioration in the material’s dielectric properties and other aspects. To enhance the adaptability of epoxy resin in high humidity environments, this paper, based on the molecular dynamics simulation method, establishes epoxy resin-based nanocomposites with doped nanofillers: a pure epoxy resin model and three epoxy resin models, respectively, doped with carbon nanotubes, graphene(GR), and SiO₂. Based on the above models, using LAMMPS-17Apr2024, the thermal diffusion coefficients (thermal conductivity and specific heat capacity), glass transition temperatures, and dielectric constants under different moisture contents are calculated. The results show that the various properties of the epoxy resin nanocomposites doped with nanofillers have been improved to varying degrees. Among them, the GR/epoxy resin composite model shows the most significant improvements in thermal conductivity, thermal diffusivity, and glass transition temperature, and the SiO₂/epoxy resin composite model has the best dielectric properties. Considering the high-temperature operation conditions and heat dissipation requirements of the high-frequency transformer, the GR-enhanced epoxy resin becomes the optimal filler choice. Full article

The development of underground space, as a critical strategy for enhancing urban land use efficiency, requires careful consideration of the effects that new construction may have on existing foundations and structures to prevent safety hazards such as foundation damage. This paper investigates the influence of shield tunnel construction on the pile foundations of adjacent bridges. Based on the shield tunnel project intersecting the Haiqin Bridge pile foundations along a segment of the Guangzhou–Zhuhai Intercity Railway as a case study, a finite element (FE) model was developed. The validity of the numerical method was confirmed through comparison with existing model test results. Building on this foundation, this paper analyzed the impact patterns of shield tunnel construction on existing bridge pile foundations. Additionally, the model was employed to assess how variables such as the relative spatial positioning between the pile foundations and the tunnel, as well as the stiffness coefficient of the pile foundations, affect the structural response of the piles. The findings reveal that shield tunnel construction crossing adjacent bridge pile foundations induces bending deformation of the piles toward the tunnel side. The maximum horizontal displacement and internal forces occur near the tunnel axis, whereas the peak vertical displacement is observed at the pile head. The zone most affected by tunnel excavation extends approximately one tunnel diameter (1D) before and after the pile foundation location. The vertical relative position between the tunnel and pile foundation governs the relative displacement behavior between the pile and surrounding soil during excavation. Specifically, when the pile toe moves downward relative to the tunnel, the excavation’s influence on the pile foundation shifts from being dominated by negative skin friction and settlement to positive skin friction and rebound, leading to substantial changes in the force distribution and displacement patterns within the pile. As the horizontal clearance between the tunnel and pile foundation increases, the internal forces and displacements within the pile foundation progressively diminish and eventually stabilize. Furthermore, an increase in pile stiffness coefficient decreases the maximum pile displacement and increases internal forces in the pile shaft. Pile diameter has a greater influence than Young’s modulus, which exhibits a relatively minor effect. Full article

This study proposes a novel structural deflection measurement method using a single smartphone with an innovative scale factor (SF) calibration technique, eliminating reliance on laser rangefinders and industrial cameras. Conventional off-axis digital image correlation (DIC) techniques require laser rangefinders to measure discrete points for SF calculation, suffering from high hardware costs and sunlight-induced ranging failures. The proposed approach replaces physical ranging by deriving SF through geometric relationships of known structural dimensions (e.g., bridge length/width) within the measured plane. A key innovation lies in developing a versatile SF calibration framework adaptable to varying numbers of reference dimensions: a non-optimized calculation integrates smartphone gyroscope-measured 3D angles when only one dimension is available; a local optimization model with angular parameters enhances accuracy for 2–3 known dimensions; and a global optimization model employing spatial constraints achieves precise SF resolution with ≥4 reference dimensions. Indoor experiments demonstrated sub-0.05 m ranging accuracy and deflection errors below 0.30 mm. Field validations on Beijing Subway Line 13′s bridge successfully captured dynamic load-induced deformations, confirming outdoor applicability. This smartphone-based method reduces costs compared to traditional setups while overcoming sunlight interference, establishing a hardware-adaptive solution for vision-based structural health monitoring. Full article