Search Results (1,833)

To address the problems of the limited feature extraction capability and insufficient training efficiency of the traditional Continuous Bag-of-Words (CBOW) model in Natural Language Processing (NLP), the Quantum Neural Network-enhanced CBOW model (QNN-CBOW) integrates Quantum Neural Networks (QNN) with the CBOW model, effectively enhancing training performance. This work aims to systematically investigate the sensitivity and influence patterns of key factors (activation function type, number of quantum feature extraction layers, context window size, and quantum gate noise level) on model behavior under controlled small-scale simulation conditions. Comparative experiments are carried out using the control variable method to clarify the influence mechanism of each factor. This paper presents a NISQ-era proof-of-concept study, which provides a theoretical basis and practical reference for the fusion and optimization of quantum neural networks and traditional NLP models. Full article

To overcome low bioavailability and high trauma in inner ear therapies, targeted delivery across the round window membrane (RWM) via hollow microneedles (HMNs) offers a promising solution. However, the fabrication of high-aspect-ratio, small-size HMNs remains challenging. This study demonstrates the successful fabrication of small-outer-diameter HMNs using a 10 μm resolution digital light processing (DLP) system. Finite element analysis (FEA) identified a double tangent-arc transition as the optimal structural design for minimizing stress concentration. To manage the heightened parameter sensitivity at sub-feature-scale fabrication, a corrected curing index (CCI) model was established via a physics-informed regression approach incorporating polymerization kinetics and nonlinear spatial intensity distribution, achieving high fitting accuracy (R² > 0.96). Under optimized parameters, the fabricated HMNs possessed mean dimensions of 805.13 μm in height, 37.54 μm in inner diameter, and 79.36 μm in outer diameter. Compressive tests exhibited a robust structural strength of up to 141 mN per needle following post-curing. Combined in silico and in vitro experiments demonstrated excellent penetration performance. Furthermore, the HMNs achieved stable, pressure-dependent delivery with volumetric flow rates rising from 0.14 mL∙min⁻¹ to 0.39 mL∙min⁻¹ as driving pressure escalated from 50 kPa to 300 kPa, validating their functional capacity for controlled drug administration. Full article

Background: Low-dose aspirin (LDA) reduces preeclampsia (PE) risk by up to 40%, yet its molecular effects on chorion trophoblast cells (CTCs), a fetal membrane lineage at the feto-maternal interface, remain obscure. CTCs form a structural and immunoregulatory barrier whose dysfunction drives inflammation-associated membrane pathology in PE. Extracellular vesicles (EVs) released by CTCs may encode cellular stress and adaptation states, offering a molecular window into aspirin’s timing-dependent effects on PE risk modification. Methods: Human CTCs were challenged with cigarette smoke extract (CSE) to model oxidative stress-driven PE pathology. Two paradigms were tested: (1) prophylactic aspirin (4 and 40 µg/mL) before and/or flanking the CSE, and (2) therapeutic aspirin after the CSE challenge. The EVs were isolated via ultracentrifugation and size-exclusion chromatography, characterized by nanoparticle tracking and immunoblotting, and profiled by quantitative mass spectrometry. A network pathway analysis and machine learning biomarker selection defined the EV-encoded molecular states. Results: The CTC-derived EVs from the CSE-exposed cells carried a PE-like proteomic signature marked by suppressed VEGF/ECM remodeling, activated TNF-p53 apoptotic signaling, and heightened inflammation. Prophylactic low-dose aspirin shifted the EV cargo toward an EV-encoded signature consistent with preserved angiogenic potential (enrichment of VEGFA, COL1A1, and MMP14) and predicted attenuation of apoptotic and NF-κB pathway activity by an Ingenuity Pathway Analysis. High-dose aspirin produced broad transcriptional suppression without an accompanying pro-angiogenic EV signature. Therapeutic (post-injury) aspirin partially attenuated the injury-associated EV cargo but did not restore the angiogenic EV signature. An exploratory machine learning analysis of EV proteomes identified a candidate prophylactic biomarker panel anchored by HSPA8, SERPINF2, COL4A1, and PLOD1, mapped to the predicted angiogenic recovery and redox-balance pathways. These EV cargo readouts represent the predicted molecular states and require functional validation before clinical interpretation. Conclusions: The CTC-derived EV proteomic signatures capture the dose- and timing-dependent aspirin effects in this in vitro CTC model, positioning the chorion as a candidate pharmacological “secondary responder” favoring cellular resilience over classical anti-inflammatory suppression. As an exploratory hypothesis-generating study, EV-based molecular profiling could provide a foundation for future investigations aimed at stratifying aspirin responders from non-responders, although clinical validation in maternal plasma cohorts will be required before any translational application. Full article

Guttation plays an important role in increasing rice yield, preventing crop diseases and improving soil fertility, and is an important index to measure the water status in the field. However, the real-time detection of droplets generated by guttation is a difficult task. Droplets are small and dense targets. In the deep learning model, the detection of small objects in high-resolution images is a key problem to be solved. A Droplet-YOLO method is proposed, which combines the YOLOv8 model and multi-instance learning. The original image is divided into multiple sub-images using a sliding window and detected by YOLOv8. The detection of each sub-image is carried out by an independent repeated test, so multi-instance learning ensures 99.99% probability of detecting droplets. By comparing various models, Droplet-YOLO performs best in terms of precision, recall, and mean precision (mAP). In addition, through the hyper-parameter adjustment experiment, the configuration with a batch size of 32 and epoch of 200 was finally selected. The accuracy and recall rate of the model reached 96.8% and 95.7%, respectively, and the mAP reached 99.0%. Experiments show that this method has significant advantages for small object detection in high-resolution images, and the proposed Droplet-YOLO model is superior to the most advanced methods. Full article

In the context of urban distribution, given the complexity of express delivery and the variability of distribution conditions, vehicle routing problems with time-dependent characteristics have received increasing attention. This study incorporates a cross-period travel time estimation method for road segments that accounts for temporal and weather-dependent variations in vehicle speed. Building upon this foundation, this study establishes an multi-objective optimization model for the green vehicle routing problem that systematically incorporates intricate constraints, including time-varing vehicle speed, fuel consumption, carbon emissions, and customer servive time windows. This model aims to achieve three primary objectives: (1) minimizing the fleet size, (2) minimizing the overall delivery expenses, which include fuel consumption and carbon emissions, and (3) maximizing the average customer satisfaction. To solve this model, we develop an improved Non-Dominated Sorting Genetic Algorithm III (INSGA-III). To effectively prevent the algorithm from becoming trapped in local optima, we propose a dual-criteria selection mechanism. Meanwhile, we introduce a destroy-and-repair variable neighborhood search strategy to enhance the algorithm’s optimization capability under complex constraints. Experimental evaluations conducted on Solomon benchmark instances as well as real-world case studies indicate that the proposed INSGA-III algorithm surpasses widely utilized multi-objective optimization methods across all assessed performance metrics. This highlights the significant potential of the presented INSGA-III algorithm for practical applications in urban delivery scenarios, which is closely linked to the sustainable development of cities. Full article

Automatic video anomaly detection remains challenging because abnormal events are infrequent, visually heterogeneous, and weakly bounded in time. This study proposes an unsupervised framework trained only with normal video segments. The framework integrates sliding-window segment construction, dual-view perturbation, a two-branch spatio-temporal encoder, exponential moving-average prototype updating, prototype-guided contrastive optimization, and a robust anomaly score composed of prototype deviation, second-order temporal residual, and local-neighborhood sparsity. Experiments were conducted on UCSD Ped2, CUHK Avenue, and ShanghaiTech under the same input size, segment length, optimizer, and threshold protocol. The proposed model achieved AUC values of 97.4%, 91.8%, and 83.7% on the three datasets, respectively, with an average AUC of 91.0% and an average F1 score of 88.1%. Relative to the baseline contrastive model, the average AUC increased by 2.4 percentage points, and the average F1 score increased by 2.8 percentage points. Across three independent runs, the improvement over the contrastive baseline was statistically significant (paired two-sided t-test, p = 0.018). Ablation and sensitivity analyses indicate that the performance gain is mainly attributable to spatio-temporal joint encoding, prototype traction, temporal residual scoring, and local-neighborhood support. These results show that contrastive representation learning, explicit prototype updating, and temporal-aware scoring can jointly produce a stable representation of normal behavior without using abnormal samples during training. Full article

Arthritis (RA) is characterized by immune-mediated chronic inflammation and endothelial dysfunction, ultimately resulting in clinically overt cardiovascular complications. As a prototypical disease of microvascular dysfunction, RA represents an ideal model to study microvascular alterations. The dermal capillary network offers an easily accessible window to the peripheral microcirculation, whose function can be easily assessed using Nailfold videocapillaroscopy (NVC) or laser techniques. Whereas the clinical significance of structural alterations is not always clear, functional abnormalities may provide more direct insight into the dynamic status of the microvasculature and endothelial integrity. The present narrative review aims to provide an integrative overview of available studies evaluating functional abnormalities of the dermal microcirculation in RA, with particular emphasis on the emerging role of NVC as a dynamic vascular assessment tool. Several studies in RA have assessed the structure and morphology of the peripheral microvasculature using NVC, but far fewer data exist on functional alterations assessed with this method. The study of functional alterations of the dermal microvascular network in RA has largely been based on laser techniques, which consistently point towards altered microvascular reactivity. By contrast, functional NVC-related approaches remain limited, despite their potential ability to simultaneously assess structural and dynamic capillary abnormalities in vivo. Available evidence supports that NVC may be reframed as a promising functional vascular biomarker in RA. However, the available literature is characterized by small sample sizes, predominantly cross-sectional designs, and methodological heterogeneity, highlighting the need for standardized prospective studies. Full article

Background: This study investigated the clinical timeline, patient monitoring behaviors, and cumulative bilateral treatment burden in patients with bilateral neovascular age-related macular degeneration. Methods: We retrospectively analyzed follow-up patterns and treatment intensity from first-eye (FE) diagnosis to second-eye (SE) conversion. Results: SE conversion occurred within a mean of 2.0 years in the FE-active group (62.5%) while the FE remained exudative, contrasting with 6.2 years in the FE-inactive group (37.5%). Upon SE conversion, the total annual intravitreal injection burden escalated 3.4-fold (p = 0.002). Notably, the FE-inactive group exhibited numerically lower annual outpatient visit counts (4.40 ± 2.71 vs. 10.29 ± 5.02; p = 0.116), which potentially widened the monitoring window. Additionally, baseline SE retinal pigment epithelium (RPE) abnormalities independently predicted progression (aOR: 19.04; p = 0.032). Conclusions: While previous literature focuses on individual eyes, our findings highlight a vigilance gap in SE detection based on FE status. Clinicians must maintain proactive surveillance for patients with baseline SE RPE abnormalities, particularly when FE stability or next-generation long-acting therapies extend clinic intervals. Due to the limited sample size, these preliminary findings warrant validation in larger prospective cohorts. Full article

Night ventilation coupled with thermal mass is an effective means of reducing overheating in passive buildings. Successful systems require a high airflow rate coupled with enhanced convective heat transfer to the thermal mass. This work presents results for enhanced convection when the primary thermal mass is in the ceiling. Such mass distribution occurs, for example, in multi-story apartments in developing economies. Experimental results are measured in a scale model of a typical room. The original contribution is the use of upward-directed ventilation at an angle of 30° to 40° from a window located at a typical distance below the ceiling. At scaled air change rates of 4.9 air changes per hour, the measured convective heat transfer coefficient at the ceiling was 7.7 W/m² K. In contrast, when air flowed horizontally from the window, the heat transfer coefficient was 3.5 W/m² K or less, indicating that substantial improvement was gained by directing airflow toward the ceiling. To link the experimental results to an application in a full-size building, an approximate model is presented to estimate the impact of directed night ventilation on the thermal mass (specifically the concrete slab ceiling) and room air temperatures. Coupling angled flow with nighttime ventilation, the ceiling slab and peak daytime air temperature can be reduced by 5 °C compared to horizontal ventilation from a window at conventional height. These results have enabled collaborators in Gujarat, India, to launch tests in a full-scale home serving a low-income community without access to air conditioning. Full article

Genotype imputation (GI) plays a critical role in predicting missing genetic information for genomic studies and breeding applications. Although recent reference-free deep learning approaches have demonstrated promising performance, they often fail to exploit local genomic information, which limits further improvements in prediction accuracy and stability. In this study, we developed ALGI, a novel method based on a sparse convolutional denoising autoencoder, which uniquely integrates local genomic window information with group-specific feature learning. Unlike conventional convolutional or autoencoder-based approaches, ALGI first applies K-means clustering to group samples according to local genomic windows, then learns hidden genotype configurations specific to each group, capturing fine-scale local patterns and complex haplotype structures. Systematic evaluation was conducted across yeast, human, and pig MHC regions under multiple scenarios, including different window sizes, missing rates, sample sizes, and numbers of variants. Results show that ALGI demonstrates consistent improvements over conventional methods (Beagle) and state-of-the-art deep learning approaches (AE, SCDA) under the evaluated settings, with enhanced accuracy, stability, and robustness. In addition, ALGI is user-friendly and publicly available. While evaluated on highly polymorphic MHC regions, its strong performance suggests applicability to less complex regions, though broader genome-wide validation is needed. This approach provides a powerful tool for genomic selection and advancing complex trait genetics in livestock and other species. Full article

Brain–computer interfaces (BCIs) based on motor imagery (MI) aim to convert electroencephalographic (EEG) activity into reliable device commands across users and recording setups. However, low signal-to-noise ratio and strong inter-subject variability still limit true “plug-and-play” deployment without lengthy calibration. To address these challenges, we propose a multi-branch convolutional long short-term memory (CNN-LSTM) architecture that jointly performs multi-scale temporal feature extraction and within-trial sequence modeling. The model employs four parallel 1D convolutional branches with distinct kernel sizes, each followed by an LSTM module and late fusion, combined with group normalization and supervision over sequences of sub-windows within each trial. We evaluate the approach on the EEG Motor Movement/Imagery (EEGMMI) dataset from PhysioNet under strictly subject-independent conditions, and on the ISLab-MI Dataset, a 32-channel wearable-EEG collection designed to assess cross-setup robustness. On EEGMMI, the network achieves up to 82.63% accuracy for binary left/right MI and 74.10% for a four-class task using 4 s trials under 5-fold cross-validation, outperforming an EEGNet-style baseline by 1–10% depending on class count and window length. Under a leave-one-subject-out protocol, the model attains 74.9% mean accuracy for a three-class MI task. Zero-shot transfer to ISLab-MI yields 64.60% and 63.02% accuracy in three- and four-class settings, respectively, while brief subject-specific fine-tuning using only 20% of each session improves performance to 81.38% and 73.48%. These findings show that combining multi-scale convolutional feature extraction with explicit sequence modeling and robust normalization yields accurate, data-efficient, and portable MI decoders suitable for practical BCI applications. Full article

Background and Objectives: Conditioning activities (CAs) are commonly used to induce post-activation performance enhancement (PAPE); however, it remains unclear whether load-dependent responses established in bilateral, predominantly isotonic models extend to unilateral split squat conditions. In particular, evidence regarding holding isometric muscle action (HIMA) is limited, and it is unknown how external load and holding duration interact to influence both performance outcomes and phase-specific muscle activation. Therefore, this study examined the acute effects of HIMA duration and external load during unilateral split squat CA on jump performance and phase-specific electromyographic (EMG) activity. Materials and Methods: Twenty recreationally active men completed a randomized, counterbalanced crossover design involving four split squat CA conditions, unloaded 3 s, unloaded 5 s, 3 s loaded (60% 1RM), and 5 s loaded (60% 1RM), each performed as three sets of three repetitions. To minimize fatigue effects, standardized rest intervals and familiarization sessions were implemented prior to testing. Single-leg jump (SLJ) and countermovement jump (CMJ) were assessed before and after CA, with post-activation measurements conducted at 3 min (SLJ) and 4 min (CMJ), consistent with established PAPE time windows. Surface EMG was time-normalized across the split squat cycle and analyzed using phase-specific area under the curve. Results: CMJ significantly increased following both loaded conditions (p < 0.05; moderate to large effect sizes), whereas no differences were observed between unloaded durations. External load consistently elevated EMG amplitude across all measured muscles (moderate to large effects). Extending duration under load further increased activation in the hamstrings, stabilizers, vastus medialis, and gastrocnemius medialis (p < 0.05; small to moderate effects), whereas unloaded conditions showed minimal neuromuscular differences. Conclusions: External load, rather than isometric holding duration, appears to be a key factor influencing acute PAPE responses in unilateral split squat HIMA, whereas prolonged holding duration may primarily modulate muscle recruitment patterns without additional performance gains. However, given the acute experimental design and a recreationally active sample, these findings should be interpreted with caution and considered exploratory. Further studies are warranted to confirm these effects across different populations and longer-term training conditions. Full article

Individual tree crown (ITC) detection from remotely sensed data is fundamental to forest inventory and ecological monitoring, but deep learning approaches remain constrained by limited labelled training data. We systematically evaluate three transformer detectors (the Detection Transformer (DETR), Deformable DETR, and DETR with Improved DeNoising Anchor Boxes (DINO)) paired with two backbones, ImageNet-pretrained ResNet-50 and a Masked Autoencoder (MAE) pretrained on unlabelled Canopy Height Model (CHM) data. These are benchmarked against a classical local maximum and watershed pipeline and Faster R-CNN across four test sets spanning boreal, temperate mixed-wood, and diverse North American forest types at 0.25–1.0 m resolution. Spatially held-out test regions with a one-patch buffer band eliminate sliding-window leakage; headline configurations are reported as mean ± standard deviation across three random seeds. With multi-resolution MAE pretraining, the practical lower bound for non-degenerate single-dataset transformer detection lies between ∼200 and ∼1200 patches. Without MAE pretraining, DETR fails at every dataset size we tested. Multi-dataset joint training reaches $F_1=0.84\pm 0.01$ on the boreal test set and 0.45–0.68 across the temperate-mixed-wood and NEON test sets, while Faster R-CNN narrowly wins on the smallest training pool. Standard DETR with ResNet-50 collapses regardless of the length of training schedule, but the same architecture with an MAE backbone reaches $F_1=0.83\pm 0.01$ at that schedule, showing that DETR’s reported instability is conditional on the combination of backbone initialization and training budget rather than architectural. Resolution and backbone interact: ResNet-50 wins at 0.25 m, and MAE wins at 0.5–1.0 m, consistent with the eight-pixel MAE patch-matching crown scale only at coarser resolutions. Full article

Despite the growing importance of the Saudi capital market, sectoral-level volatility connectedness within Tadawul remains largely unexplored. This study contributes to the literature by applying the Diebold–Yılmaz framework to examine volatility connectedness across 16 Tadawul sectors over the period January 2017 to December 2024. Total, directional, and net pairwise volatility spillovers are quantified from daily closing prices using a VAR(4) model combined with generalized forecast error variance decomposition. The static analysis reveals a high overall connectedness of 80.49%, indicating that cross-sectoral spillovers account for the majority of volatility fluctuations. Materials, Transportation, and Real Estate Management and Development are identified as the dominant net transmitters of volatility, while Utilities and Telecommunication Services are persistent net receivers. The dynamic analysis shows that sectoral connectedness is highly time-varying, peaking at 93.70% during the COVID-19 period, with additional episodes of elevated spillovers during 2022–2023. The network analysis reveals that the strongest pairwise linkages exist among Materials, Transportation, Real Estate Management and Development, and Banks, forming the core of the spillover network. While block-bootstrap results reinforce the identification of dominant net transmitters and receivers, they reveal substantial uncertainty in the rank-order of intermediate sectors, necessitating a more nuanced interpretation. The results are robust to alternative rolling window sizes and forecast horizons. These findings have important implications for portfolio diversification, sectoral risk monitoring, and macroprudential policy in the Saudi capital market. Full article

Metallic glasses can be transformed into nanocrystalline–amorphous alloys via controlled crystallisation with fast nucleation and slow grain growth. This can be achieved either through appropriate chemical composition of amorphous precursors or by applying ultra-rapid annealing (URA). Typically, heating between preheated copper blocks is used to ensure the URA conditions. In this work, ribbons were heated by an electric current flowing along their length, and the temperature was monitored using pyrometers. The investigated alloys were Fe_86-xNi_xB₁₄ (at. %), where x = 4, 6 or 10. Properly adjusted isothermal annealing at 380–410 °C for 1–20 s induced crystallisation, with the nanocrystalline bcc-Fe(Ni) phase occupying 0–55% of the volume. With increasing annealing time, the coercive field increased from 9 A/m in the amorphous state to 25 A/m and 17 A/m for x = 4 and x = 10, respectively. Transmission electron microscopy confirmed that samples annealed at higher temperatures for shorter times exhibited smaller grain sizes compared to those annealed at lower temperatures for longer times, which resulted in improved magnetic softness. An increase in nickel content reduced coercivity, improved ductility, and offered a wider window for the choice of annealing temperature. Full article