Search Results (10,126)

Remote sensing (RS) images are often degraded by atmospheric haze, which compromises both visual interpretation and downstream applications. To address this, we introduce CSGL-Former, a novel Cross-Stripes Global–Local Fusion Transformer for RS image dehazing. Our model efficiently captures anisotropic long-range dependencies using cross-stripes attention (CSA) and aggregates hierarchical global semantics via a Multi-Layer Global Aggregation (MLGA) module. In the decoder, global context is adaptively blended with fine-grained local features to restore intricate textures. Finally, inspired by the atmospheric scattering model, a soft reconstruction head restores the clear image by predicting spatially varying affine parameters, strictly preserving content fidelity while effectively removing haze. Trained end-to-end, CSGL-Former demonstrates a compelling balance of accuracy and efficiency. Extensive experiments on the RRSHID and SateHaze1K benchmarks show that our model achieves state-of-the-art or highly competitive performance against representative baselines. Ablation studies further validate the effectiveness of each proposed component. Full article

This paper presents a Load-Dependent Multimodal Vibration Signal Enhancement and Fusion Framework (LD-MVSEFF) for load-specific condition monitoring, building on the Customised Load Adaptive Framework (CLAF). The proposed approach enhances the classification of CLAF load-dependent subclasses, namely, Healthy, Mild, Moderate, and Severe, by integrating complementary information from raw vibration signals and encoded signal representations. Three input channels are employed, combining time–frequency domain features with Continuous Wavelet Transform (CWT) and Gramian Angular Difference Field (GADF) image encodings, with each channel independently trained and evaluated to identify its most effective classifiers. To address the reduced separability of the Mild and Moderate fault subclasses under varying load conditions, a weighted decision-fusion strategy is introduced, assigning classifier contributions according to their class-specific strengths. Experimental evaluation over five runs demonstrates high and stable performance, with the best configuration achieving an overall accuracy of 99.04% ± 0.22% and an average training time of 18 min and 30 s. The results confirm the effectiveness of LD-MVSEFF as a robust multimodal methodology for load-specific condition monitoring. Full article

In real-world scenarios, multivariate time series data typically presents a variety of complex characteristics simultaneously, including long-term trends, multiple seasonality, sudden event disturbances and random noise. Owing to remarkable discrepancies among different variables in dimensions, periodic stability and other aspects, and the gradual evolution of these periodic characteristics over time, models are confronted with numerous challenges in handling non-stationarity, multi-scale dynamic variations and heterogeneous fusion of variables. To tackle these problems, this paper proposes a time–frequency parallel fusion framework—TFDG-Net (Time–Frequency Dual-Branch Gated Fusion Network). This framework models the prior information in the frequency domain and the temporal query network in the time domain in parallel, and introduces a channel-wise gating mechanism to achieve more flexible adaptive fusion after data inverse normalization. Such a design enables the model to operate collaboratively on the original physical scale, which not only improves the long-term prediction capability for periodically stable variables, but also effectively suppresses the interference of noise and event-driven factors, thus significantly enhancing prediction accuracy and the robustness of the training process. In multiple long-term prediction benchmark tests covering fields such as energy and finance, compared with various mainstream models, TFDG-Net reduces the mean squared error and mean absolute error by an average of 12.0% and 7.8% respectively. Full article

Magnetic resonance imaging (MRI) reconstruction from undersampled k-space data enables accelerated acquisition but leads to a severely ill-posed inverse problem. Although supervised deep learning methods have achieved strong performance, they typically rely on large paired datasets that are difficult to obtain in clinical practice. To address these limitations, we propose a dictionary-learning-inspired dAta-fRee deep generative modeling with Meta-Attention and implicit precoNditIoning for compressively sampled MRI (CS-MRI), termed ARMANI. Specifically, a meta-attention-augmented deep image prior (MA-DIP) generator performs a joint optimization over the latent input $\eta$ and the network parameter $\theta$, where $\eta$ is regularized via gradient-domain sparsity and $\theta$ is constrained by a ridge penalty, mirroring the adaptive estimation of sparse coefficients and an empirical sparsifying dictionary. Furthermore, we integrate a single-step pseudo-orthogonal projection to achieve implicit preconditioning, which modulates the loss landscape and mitigates ill-conditioning of the forward operator. Experimental results demonstrate that ARMANI consistently outperforms existing SOTA data-free and self-supervised methods, and, with limited training data, achieves performance comparable to or slightly better than the supervised benchmark MoDL, with effective artifact suppression and faithful recovery of fine structural details. Overall, ARMANI shows strong scalability and potential for practical deployment in fully data-free CS-MRI reconstruction scenarios. Full article

Background: Extracellular vesicles (EVs) released from skeletal muscle mediate metabolic communication via microRNAs (miRNAs). While both circadian rhythms and exercise influence metabolism, the joint modulation of the muscle-derived EV miRNA landscape by circadian rhythms and chronic exercise remains undefined, particularly under the metabolic stress of obesity. Methods: Employing a 2×2 factorial design (Phase: ZT3 vs. ZT15; Condition: sedentary vs. exercise; ZT, Zeitgeber Time), EV-enriched fractions were isolated from ex vivo quadriceps muscle (QUA) cultures of high-fat diet-fed mice following an 8-week treadmill training regimen using polymer-based precipitation, and comprehensive miRNA profiling was performed by small RNA sequencing. Results: Principal component analysis (PCA) revealed that circadian phase accounted for a greater proportion of global variance in EV miRNA profiles than exercise. Differential expression analysis identified miR-1a-3p and miR-1b-5p as upregulated across both composite phase and exercise contrasts; however, condition-specific analyses indicated that this signal was primarily driven by the sedentary-phase comparison (ZT15-sed vs. ZT3-sed), in which the miR-29 family was also prominently co-upregulated, rather than constituting independent phase and exercise effects; this phase-associated signature was absent in the corresponding exercise-condition comparison. Exploratory functional enrichment of experimentally validated targets revealed phase-preferential association with metabolic and iron–heme pathways, whereas exercise-associated miRNAs mapped to signaling, inflammatory, and transcription-related networks. Conclusions: Circadian phase was the dominant contributor to global variance in muscle-derived EV-enriched miRNA profiles in obesity, as reflected by the phase-associated separation along principal component 1 (PC1, 33.47% of total variance), with exercise introducing context-dependent adaptive modulation. This study provides a foundational basis for investigating the temporal regulation of muscle secretome dynamics under high-fat diet conditions, highlighting temporal specificity as a key dimension in EV-mediated exercise physiology research. Full article

The COVID-19 pandemic transformed healthcare education, increasing the shift to digital tools and establishing a hybrid curriculum blending online learning with traditional clinical practice. This study aims to understand how this shift impacts the educational growth and skill building of nursing students. A qualitative approach was conducted to understand the experience of Hong Kong nursing students adapting to online learning during the pandemic and beyond. Fifty nursing students were interviewed, and Colaizzi’s phenomenological method revealed key themes in their learning narratives. The analysis revealed four distinct themes characterizing the students’ experiences: (i) Learning on their terms: the mandated shift in healthcare reflecting a lack of agency during the educational transition; (ii) Knowledge without touch: the perceived incompetence of the COVID-19 nursing cohort, highlighting anxieties regarding a lack of hands-on clinical proficiency; (iii) Words left unsaid: The weight of insecurity, indicating a decline in interpersonal skills due to isolation; and (iv) Beyond the perfect algorithm: the unrehearsed art of care, describing the difficulty in translating digital simulations to complex, human-centric patient care. Findings show that while digital progress ensured continuity in education, it also contributed to reduced clinical confidence, weaker communication skills, and shifts in how nursing students approached their learning. Consequently, the post-COVID environment demands that training programs evolve to address these specific deficits. Advancing the existing pandemic-era nursing literature, this study emphasizes the need for diverse, targeted teaching methods to mitigate these gaps. By intentionally bridging theoretical knowledge with hands-on clinical practice, educators can better support student wellbeing and help restore the confidence and competence required of future graduates. Full article

Objective: The objective of this study was to map and critically synthesize empirical evidence on ML/AI applications across surface mining unit operations, and to characterize models, validation practices, and evidence gaps. Eligibility criteria: Our eligibility criteria comprised peer-reviewed studies (2020–2025) applying ML/AI to surface mining activities, training/validating models on empirical datasets, and reporting quantitative performance metrics. Information sources: Scopus, ScienceDirect, Dimensions, and Web of Science were our information sources, last searched December 2025 and supplemented by website and citation snowballing. Risk of bias: Risk of bias was assessed using an adapted domain-based approach based on PROBAST, used to interpret findings without excluding studies. Synthesis method: Our research employed a narrative synthesis (no meta-analysis due to heterogeneity in datasets, algorithms, contexts, and metrics), grouped by application domain. Results: From 5317 records, 57 studies were included, concentrated in blasting (43), followed by load and haul (6), post-dismantling management (4), extraction (2), and overall exploitation (2). Studies predominantly reported statistical metrics (e.g., R², RMSE, and MAE), with limited operational performance indicators; validation was frequently site-specific. Dataset sizes were not reported consistently across studies. Limitations: This study’s limitations were database coverage, restricted timeframe, and incomplete reporting (e.g., software/tooling). Conclusions: ML/AI shows strong potential, especially in blasting, but scalable deployment is constrained by site specificity, inconsistent reporting, and heterogeneous validation; standardized reporting and operational indicators are priorities. Registration: The systematic review protocol was registered in OSF with DOI 10.17605/OSF.IO/5UMKB. Funding: EU Erasmus+ STRIM project (1010832727). Full article

The detection of surface defects on smartphone components is a critical step in quality assurance for industrial manufacturing. However, existing deep learning-based methods struggle with the extreme variations in defect morphology and scale, while labeled training data remains scarce due to the high cost of expert annotation. To address these challenges, we propose a twofold solution. First, we introduce MSAdaNet, a Multi-Scale Adaptive Defect Detection Network, which integrates three novel modules: a Parallel Multi-Scale Feature Aggregation (PMSFA) backbone, a Focusing Diffusion Pyramid Network (FDPN) neck, and a Scale-Adaptive Shared Detection (SASD) head. Second, to combat data scarcity, we propose a novel data generation pipeline, creating the synthetic Smartphone Camera Bezel Dataset (SCBD) of 4936 images. Extensive experiments on both real-world and synthetic datasets validate our approach. On the challenging public SSGD, MSAdaNet achieves a state-of-the-art mAP@0.5 of 54.8%, outperforming prominent frameworks and improving upon the strong YOLOv11m baseline by +10.6 points in mAP@0.5 and +18.3 points in recall. Furthermore, on our synthetic SCBD, the model achieves an impressive 94.0% mAP@0.5, confirming the quality of our data generation pipeline and the robustness of our architecture across different data distributions. Ablation studies systematically confirm the significant contribution of each proposed module, validating MSAdaNet as an effective and efficient solution for industrial defect detection. Full article

Collaborative robots, or cobots, are designed to improve productivity and safety in industrial settings. However, effective Human–Robot Collaboration (HRC) relies heavily on the human operator’s trust in the robotic partner. This study posits that trust is significantly enhanced by the robot’s ability to adapt to unpredictable human behavior. To achieve this adaptability, we propose applying an Adversarial Reinforcement Learning (ARL) framework to the robot’s activity planning. We model the assembly process as a Markov Decision Process (MDP) on a Directed Acyclic Graph (DAG). The robot learns an assembly policy using an on-policy algorithm while a simulated human agent, trained with the same algorithm, acts as an adversary that introduces disturbances and delays. We applied the proposed approach to a simple industrial case study and evaluated it on complex assembly sequences generated synthetically. Although the ARL-trained robot did not outperform conventional assembly optimization algorithms in terms of task completion time, it guaranteed robustness against human variability. This ensured task completion within a bounded timeframe regardless of human actions. By demonstrating consistent performance and adaptability in the face of uncertainty, the robot exhibits the Ability and Benevolence components of the ABI model of trust. This fosters a more resilient and trustworthy collaborative environment. Full article

Automated surface defect detection often faces a “cold-start” problem due to limited annotated data for new anomalies. Traditional object detectors struggle to converge in such few-shot settings. To address this, we adapt Vision–Language Models (VLMs), specifically YOLO-World. We use semantic pre-training to mitigate data scarcity. We evaluate this approach on the MVTec AD dataset in bounding-box format. We use a strict 1:9 train-validation split, resulting in an average of 11.8 defect instances per category. YOLO-World surpasses traditional baselines, like YOLOv11s and YOLOv26s, in 12 of 15 categories. The optimized VLM pipeline achieves up to 64.9% mAP@50 on texture-heavy categories, such as Tile, with only nine training instances. Ablation studies show standard optimization techniques are limited under 10-shot constraints. We find a critical augmentation divide. Disabling spatial distortions (Mosaic) is vital to preserving rigid-object geometry. The Normalized Wasserstein Distance (NWD) improves the localization of microscopic anomalies. Varifocal Loss (VFL) often causes model collapse. Ultimately, VLMs offer a superior foundation for cold-start inspection but require carefully tailored pipelines for robustness. Full article

This study explores how demystifying Earth Observation (EO) through co-creation pathways and local language can enhance flood resilience and environmental governance in African informal cities. Using case studies from Maiduguri and Hadejia, Nigeria, the research employed a transdisciplinary mixed-methods design combining rapid evidence assessment, surveys, participatory workshops (n = 50 stakeholders) integrating simplified Sentinel-1/2 demonstrations, indigenous knowledge mapping, and pre-/post-engagement surveys on EO familiarity. Non-expert participants were trained to interpret satellite data using local language, linking distant teleconnections with local flood experiences. The findings revealed significant gains in EO literacy and improvements in interpretive confidence, gender-inclusive participation, and policy engagement. Localizing the curriculum enabled participants to translate technical EO concepts into locally meaningful narratives, fostering cognitive empowerment and practical application in flood preparedness and advocacy. The study demonstrates that data democratization is not only a matter of open access but also of open understanding. It advances a conceptual model linking Demystification, Literacy, Empowerment, Co-Production and Resilience, positioning EO as a social technology that bridges scientific and indigenous knowledge systems. The findings contribute to debates on decolonizing environmental science and propose a potential participatory framework for integrating EO into community-based adaptation, legal accountability, and policy reform across Africa’s rapidly urbanizing landscapes. Full article

Background: Infections caused by multidrug-resistant bacteria and inadequate vaccine coverage against opportunistic pathogens highlight the need for interventions that broadly and durably enhance host defense beyond antigen-specific adaptive immunity. Trained immunity, driven by metabolic and epigenetic reprogramming of innate immune cells, has been predominantly characterized using Bacille Calmette–Guérin and β-glucan, whereas its induction by Gram-negative bacteria remains poorly defined. To address this gap, we aimed to determine whether heat-killed Klebsiella pneumoniae (HK Kp) induces trained immunity through metabolic and hematopoietic reprogramming to confer heterologous antibacterial protection. Methods: HK Kp-trained murine bone marrow-derived macrophages and HK Kp-immunized C57BL/6 mice were employed to interrogate functional, metabolic, and transcriptomic reprogramming in vitro, hematopoietic progenitor remodeling in vivo, and protective efficacy against systemic Salmonella Typhimurium and Staphylococcus aureus infection. Results: HK Kp-trained macrophages showed markedly enhanced IL-1β secretion across all restimulation conditions, stimulus-dependent amplification of TNF-α responses, increased phagocytosis, and improved intracellular control of S. typhimurium, together with sustained upregulation of the glycolytic enzymes-encoding genes Hk2 and Pfkfb3. Transcriptomic profiling revealed extensive reprogramming enriched in glycolysis/gluconeogenesis and hematopoietic cell lineage pathways. In vivo, HK Kp immunization shifted bone marrow stem/progenitor compartments toward a myeloid-biased state. HK Kp-trained mice challenged with lethal S. typhimurium or S. aureus exhibited less weight loss, improved survival rates, and reduced bacterial burdens. Conclusions: Inactivated K. pneumoniae orchestrates metabolic and hematopoietic reprogramming to establish enhanced innate immune responsiveness and confer heterologous protection in murine S. typhimurium and S. aureus sepsis models, supporting its potential as a potent inducer of trained immunity. These findings establish HK Kp-based trained immunity as a promising strategy for combating multidrug-resistant and vaccine-evading pathogens. Full article

Accurate cell and nuclei segmentation remains challenging due to the sensitivity of classical graph-cut methods to parameter tuning. While deep learning models like U-Net offer strong performance, they require large annotated datasets and substantial GPU resources. This work presents a cost-effective alternative: a genetic programming (GP) framework that jointly optimizes unary cost functions and regularization parameters for graph-cut segmentation, coupled with automatic seed selection. Evaluation is conducted under two distinct protocols: (1) oracle-guided per-image optimization, establishing upper-bound performance (mean Dice 0.822, IoU 0.733), and (2) true generalization via train/test split, where expressions learned on 50 images are applied to 50 unseen images (mean Dice 0.695, IoU 0.588). The fixed-model generalization still significantly outperforms the baseline graph cut ($+0.158$ Dice, $p<0.001$). Cross-dataset validation on MoNuSeg (H&E histopathology) achieves a Dice score of 0.823 with the fixed GP model, significantly outperforming the baseline (+0.272). This result uses a single fixed model—the best-performing expression from BBBC038 training—applied in a zero-shot manner to MoNuSeg without any retraining or domain adaptation. All 100 images showed non-negative improvement under oracle optimization in the experiments. The method requires no GPU training, runs in 550 s per image for oracle search, and offers interpretable symbolic cost functions. Code and annotations are provided to ensure reproducibility. This approach offers a practical, interpretable alternative in resource-constrained biomedical imaging settings. Full article

Image captioning using deep learning bridges computer vision and natural language processing, enabling machines to generate human-like textual descriptions for images. While significant progress has been made in English, in Arabic, the image captioning field remains under-explored due to the language’s morphological complexity, right-to-left script, and scarcity of annotated datasets. This paper addresses this gap by adapting the BLIP-2 (Bootstrapped Language—Image Pre-training) model for Arabic caption generation, leveraging machine-translated datasets, like Flickr 30k, to overcome resource limitations. BLIP-2 combines a vision transformer (ViT) for image encoding and a CamelBERT large language model (LLM) for text generation, enhanced by a lightweight Querying Transformer (Q-Former) for cross-modal alignment. Despite challenges such as translation artifacts and linguistic nuances, our experiments demonstrate promising results in generating coherent Arabic captions. In short, this study highlights the potential of BLIP-2 for multilingual applications while underscoring the need for native Arabic datasets and further optimization. Ultimately, this work contributes to advancing inclusive artificial intelligence technologies for Arabic-speaking communities, with applications in assistive tools, education, and content creation. Full article

The menstrual cycle represents a dynamic infradian rhythm characterized by coordinated fluctuations in ovarian steroids that extend beyond reproductive function and influence systemic metabolism. This narrative review synthesizes current evidence on how menstrual cycle phase modulates energy balance, macronutrient metabolism, micronutrient handling, and responses to dietary bioactive compounds. Across phases, small-to-moderate but consistent differences emerge in energy intake, resting energy expenditure, substrate utilization, and protein turnover, with a tendency toward increased energy intake and lipid oxidation during the mid-luteal phase compared with the early follicular and peri-ovulatory phases. Emerging metabolomics data further reveal coordinated cyclical variation in amino acids, B vitamins, and lipid species, suggesting temporally sensitive windows in which low energy availability or micronutrient insufficiency may more readily impair performance, recovery, or symptom burden. Importantly, menstrual cycle-related metabolic variability reflects not only estradiol and progesterone oscillations but also integrated adaptations across the hypothalamic–pituitary–adrenal axis, autonomic nervous system, immune signaling, and gut microbiota. These interconnected systems contribute to inter- and intra-individual heterogeneity in metabolic phenotype. From a clinical and applied perspective, the evidence supports “cycle-aware” but non-dogmatic nutritional strategies, particularly in contexts of metabolic dysfunction, high training loads, or reproductive disorders. Future research should systematically verify cycle phase, incorporate multi-system biomarkers, and adopt sex-specific analytical frameworks to improve translational relevance. Recognizing the menstrual cycle as a biologically meaningful metabolic variable may enhance precision nutrition, exercise prescription, and metabolic risk assessment in women. Full article