Search Results (1,019)

This study proposes Dual-Stream Former, a novel architecture that integrates a Video Swin Transformer and Conformer designed to address the challenges of visual speech recognition (VSR). The model captures spatiotemporal dependencies, achieving a state-of-the-art character error rate (CER) of 3.46%, surpassing traditional convolutional neural network (CNN)-based models, such as 3D-CNN + DenseNet-121 (CER: 5.31%), and transformer-based alternatives, such as vision transformers (CER: 4.05%). The Video Swin Transformer captures multiscale spatial representations with high computational efficiency, whereas the Conformer back-end enhances temporal modeling across diverse phoneme categories. Evaluation of a high-resolution dataset comprising 740,000 utterances across 185 classes highlighted the effectiveness of the model in addressing visually confusing phonemes, such as diphthongs (/ai/, /au/) and labio-dental sounds (/f/, /v/). Dual-Stream Former achieved phoneme recognition error rates of 10.39% for diphthongs and 9.25% for labiodental sounds, surpassing those of CNN-based architectures by more than 6%. Although the model’s large parameter count (168.6 M) poses resource challenges, its hierarchical design ensures scalability. Future work will explore lightweight adaptations and multimodal extensions to increase deployment feasibility. These findings underscore the transformative potential of Dual-Stream Former for advancing VSR applications such as silent communication and assistive technologies by achieving unparalleled precision and robustness in diverse settings. Full article

Multi-label arrhythmia classification from 12-lead ECG signals is a tricky problem, including spatiotemporal feature extraction, feature fusion, and class imbalance. To address these issues, a multi-scale temporal–dynamic graph convolutional with orthogonal gates method, termed MST-DGCN, is proposed for ECG arrhythmia classification. In this method, a temporal–dynamic graph convolution with dynamic adjacency matrices is used to learn spatiotemporal patterns jointly, and an orthogonal gated fusion mechanism is used to eliminate redundancy, so as to strength their complementarity and independence through adjusting the significance of features dynamically. Moreover, a multi-instance learning strategy is proposed to alleviate class imbalance by adjusting the proportion of a few arrhythmia samples through adaptive label allocation. After validating on the St Petersburg INCART dataset under stringent inter-patient settings, the experimental results show that the proposed MST-DGCN method can achieve the best classification performance with an F1-score of 73.66% (+6.2% over prior baseline methods), with concurrent improvements in AUC (70.92%) and mAP (85.24%), while maintaining computational efficiency. Full article

Background: Cluster sets (CSs) maintain velocity and power in compound movements by employing similar propulsion strategies or maintaining impulse through different mechanisms. This study aimed to explore the effect of four CS conditions on back squat (BS) propulsion and provide models for estimating changes in propulsion based on repetition and set number. Methods: Twenty male participants (age = 28.3 ± 3.1 years, stature = 1.74 ± 8.21 m, body mass = 84.80 ± 7.80 kg, BS 1RM = 140.90 ± 24.20 kg) completed four data collection sessions. Each session consisted of three sets of five repetitions at 80% 1RM BS with three minutes of unloaded inter-set rest, using varying intra-set rest intervals. Experimental conditions included 0 s (TRAD), 10 s (CS10), 20 s (CS20), and 30 s (CS30) inter-repetition rest, randomly assigned to sessions in a counterbalanced order. Ground reaction force data were collected on dual force platforms sampling at 1000 Hz, from which net propulsive impulse (J_PROP), mean force (MF), and propulsion time (t_PROP) were calculated. Conditions and sets were analysed using a 4 × 3 (CONDITION*SET) repeated-measures ANOVA to assess differences between conditions and sets, and linear mixed models (LMMs) were used to provide regression equations for each dependent variable in each condition. Results: The ANOVA revealed no significant interactions for any dependent variable. No main effects of CONDITION or SET were observed for J_PROP. The main effects of CONDITION showed that MF was significantly lower in TRAD than CS20 (g = 0.757) and CS30 (g = 0.749). t_PROP was significantly higher in TRAD than CS20 (g = 0.437) and CS30 (g = 0.569). The main effects of SET showed that MF was significantly lower in S2 (g = 0.691) and S3 (g = 1.087) compared to S1. t_PROP was significantly higher in S2 (g = 0.866) and S3 (g = 1.179) compared to S1. LMMs for CS20 and CS30 revealed no significant effect (p > 0.05) between repetition or set number and dependent variables. Conclusions: The results suggest that CS20 and CS30 maintain J_PROP by limiting MF and t_PROP attenuation. This is less rest than that suggested by the previous literature, which may influence programming decisions during strength and power mesocycles to maximise training time and training density. LMMs provide accurate estimates of BS propulsive force attenuation when separating repetitions by up to 30 s, which may help practitioners optimise training load for long-term adaptations. Full article

Air quality in the Po Valley (Northern Italy), one of Europe’s most polluted regions, remains a major concern due to its unfavorable orographic setting and intense anthropogenic emissions. Climate change may further hinder progress by modifying meteorological conditions that regulate pollutant dispersion and chemistry. This study applies a modeling framework combining regional climate simulations and chemical transport models to assess the climate penalty, i.e., the adverse impact of climate-driven meteorology on air quality independent of emissions. Three scenarios were analyzed: Baseline Reference Scenario (SRB) (2011–2015), Near-Future Medium Scenario (NF) (2028–2032), and Mid-Future Medium Scenario (2048–2052), with emissions held constant. A mitigation scenario (SC_MF_2050) under the Current Legislation was also tested to accomplish the new EU Ambient Air Quality Directive. Results show that PM₁₀ and NO₂ increase under future climates, mainly due to reduced wind speed and precipitation, enhancing pollutant accumulation. Multivariate analyses confirm the strong association between stagnant conditions and higher concentrations. Even with projected emission reductions, compliance with stricter EU targets may not be achieved everywhere. Climate penalty zones, especially in lowland and transport corridors, underscore the need to integrate climate resilience into air quality planning and adopt adaptive strategies for long-term effectiveness. Full article

Afatinib and osimertinib are current treatment options for non-small cell lung cancer (NSCLC) patients with uncommon epidermal growth factor receptor (EGFR) mutations, although their efficacy is limited. To explore potentially effective drugs for these patients, we evaluated the efficacy of conventional EGFR tyrosine kinase inhibitors (TKIs) and novel third-generation (3G) TKIs using in vitro models. Ba/F3 cells transformed with each of the five most frequent uncommon EGFR mutations, Del18 (delE709_T710insD), E709K, G719A, S768I, and L861Q, were used. The growth inhibitory effects of five novel 3G-TKIs, almonertinib, lazertinib, furmonertinib, rezivertinib, and befotertinib, in addition to currently available TKIs, were evaluated. We also explored for secondary resistant mutations to afatinib or osimertinib and TKIs that can overcome these resistances. Afatinib was active against all uncommon EGFR mutations tested. The 3G-TKIs were all active against the L861Q mutation and were inactive against the S768I mutation. Furmonertinib and befotertinib showed efficacy against exon 18 mutations (Del18, E709K, and G719A). In the acquired resistance models to afatinib or osimertinib, we found T790M or a novel T725M secondary mutation, respectively, both of which could be overcome by lazertinib. However, some afatinib-resistant cells acquired V769L/M secondary mutations that were refractory to all EGFR-TKIs tested. In conclusion, afatinib exhibited broad activity and some 3G-TKIs showed promising efficacy in the front-line setting. Lazertinib is a potential second-line option after acquisition of resistance to afatinib or osimertinib. Full article

This study investigates on-edge seizure detection that aims to resolve two major constraints that hold the deployment of deep learning models in clinical settings at present. First, centralized training requires gathering and consolidating data across institutions, which poses a serious issue of privacy. Second, a high computational overhead inherent in inference imposes a crushing burden on resource-limited edge devices. Hence, we propose NeuroFed-LightTCN, a federated learning (FL) framework, incorporating a lightweight temporal convolutional network (TCN), designed for resource-efficient and privacy-preserving seizure detection. The proposed framework integrates depthwise separable convolutions, grouped with structured pruning to enhance efficiency, scalability, and performance. Furthermore, asynchronous aggregation is employed to mitigate training overhead. Empirical tests demonstrate that the network can be reduced fully to 70% with a 44.9% decrease in parameters (65.4 M down to 34.9 M and an inferencing latency of 56 ms) and still maintain 97.11% accuracy, a metric that outperforms both the non-FL and FL TCN optimizations. Ablation shows that asynchronous aggregation reduces training times by 3.6 to 18%, and pruning sustains performance even at extreme sparsity: an F1-score of 97.17% at a 70% pruning rate. Overall, the proposed NeuroFed-LightTCN addresses the trade-off between computational efficiency and model performance, delivering a viable solution to federated edge-device learning. Through the interaction of federated-optimization-driven approaches and lightweight architectural innovation, scalable and privacy-aware machine learning can be a practical reality, without compromising accuracy, and so its potential utility can be expanded to the real world. Full article

Newcastle disease (ND) is a highly contagious and economically significant viral infection that affects poultry globally, with recurrent outbreaks occurring even among vaccinated flocks in Egypt. Caused by the Newcastle disease virus (NDV), the disease results in substantial losses due to high mortality rates, decreased productivity, and the imposition of trade restrictions. This study aimed to develop a rapid, sensitive, and field-deployable diagnostic assay based on real-time reverse transcription recombinase-aided amplification (RT-RAA) for the detection of all NDV genotypes in clinical avian specimens. Primers and an exo-probe were designed based on the most conserved region of the NDV matrix gene. After testing ten primer combinations, the pair NDV RAA-F1 and RAA-R5 demonstrated the highest sensitivity, detecting as low as 6.89 EID₅₀/mL (95% CI). The RT-RAA assay showed excellent clinical sensitivity and specificity, with no cross-reactivity to other common respiratory pathogens such as avian influenza virus, infectious bronchitis virus, Mycoplasma gallisepticum or infectious laryngotracheitis virus. All 25 field samples that were tested positive by real-time RT-PCR, including those with high CT values (~35), were detected by RT-RAA in 2–11 min, indicating superior sensitivity and speed. The assay requires only basic equipment and can be performed under isothermal conditions, making it highly suitable for on-site detection in resource-limited or rural settings. The successful implementation of RT-RAA can improve NDV outbreak response, support timely vaccination strategies, and enhance disease control efforts. Overall, the assay presents a promising alternative to conventional diagnostic methods, contributing to the sustainability and productivity of the poultry sector in endemic regions. Full article

Runners have a high probability of sports injuries due to improper landing patterns. This study aimed to investigate the effects of three different foot strike patterns on lower limb muscle activation in healthy young male university students without specialized sports training experience. Methods: Sixteen healthy male college students (age: 21 ± 1 years) participated in this study. They performed running with three different foot strike patterns: forefoot strike (FFS), midfoot strike (MFS), and rearfoot strike (RFS) at controlled speeds of 1.4–1.6 m/s. EMG signals from six lower limb muscles (vastus lateralis, vastus medialis, rectus femoris, tibialis anterior, lateral gastrocnemius, and medial gastrocnemius) during the stance phase were collected using a wireless EMG system (1000 Hz). Ensemble Empirical Mode Decomposition (EEMD) was employed to analyze the time–frequency characteristics of lower limb EMG signals and ankle joint co-activation patterns to investigate the corresponding neuromuscular control mechanisms. Statistical analyses were performed using repeated-measures ANOVA, and significance was set at p < 0.05. Results: The timing of maximum energy in lower limb muscles during the stance phase occurred earlier in RFS compared to FFS and MFS. At initial ground contact, the low-frequency component energy (below 60 Hz) of the medial gastrocnemius was significantly higher in MFS and RFS compared to FFS, while FFS exhibited significantly higher high-frequency component energy (61–200 Hz). The co-activation of ankle dorsiflexors and plantar flexors (TA/GM) was also significantly higher in MFS and RFS compared to FFS. During the 100 ms before foot contact, the low-frequency component energy (below 60 Hz) of the lateral gastrocnemius was significantly higher in MFS compared to FFS, and the degree of TA/GM co-activation was significantly higher in both MFS and RFS compared to FFS. Conclusions: The maximum frequency in lower limb muscles appeared earliest during the mid-stance phase in the rearfoot strike (RFS) pattern. Moreover, during the pre-activation and early stance phases, frequency differences were observed only in the medial gastrocnemius, with RFS showing significantly higher low-frequency power. Full article

Accurate crop mapping is vital for monitoring agricultural resources, food security, and ecosystem sustainability. Advances in high-resolution sensing technologies now enable precise, large-scale crop mapping, improving agricultural management and decision-making. However, in scenarios where balancing precision and computational resources is important, obtaining the optimal feature combination (especially newly proposed features) and strategies from the rich feature sets contained in multi-source remote sensing imagery remains one of the challenges. In this paper, we propose a hierarchical feature optimization method, incorporating a newly reported vegetation feature, for mapping crop types by combining the Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 optical imagery within the Google Earth Engine (GEE) platform. The method first calculates spectral features, texture features, polarization features, vegetation index features, and crop phenological features, with a particular focus on infrared band features and the newly developed Kernel Normalized Difference Vegetation Index (kNDVI). These 126 features are then selected to construct 15 crop type mapping models based on different feature combinations and a random forest (RF) classifier. Feature selection was performed using the feature correlation analysis and random forest recursive feature elimination (RF-RFE) to identify the optimal subset. The experiment was conducted in the Linhe region, covering an area of 2333 km². The resulting 10 m crop map, generated by the optimal model (Model 15) with 34 key features, demonstrated that integrating multi-source features significantly enhances mapping accuracy. The model achieved an overall accuracy of 90.10% across five crop types (corn, wheat, sunflower, soybean, and beet), outperforming other representative feature optimization methods, Relief-F (87.50%) and CFS (89.60%). The study underscores the importance of feature optimization and reduction of redundant features while also showcasing the effectiveness of red edge and infrared features, as well as the kNDVI, in mapping crop type. Full article

Let $\mathcal{E}$ and $\mathcal{F}$ be nonempty disjoint subsets of a metric space $(M,d)$. For a non-self-mapping $\phi :\mathcal{E}\to \mathcal{F}$, which is fixed-point free, a point $\varkappa \in \mathcal{E}$ is said to be a best proximity point for the mapping $\phi$ whenever the distance of the point $\varkappa$ to its image under $\phi$ is equal to the distance between the sets, $\mathcal{E}$ and $\mathcal{F}$. In this article, we establish new best proximity point theorems and obtain real extensions of Edelstein’s fixed point theorem in metric spaces, Krasnoselskii’s fixed point theorem in strictly convex Banach spaces, Dhage’s fixed point theorem in strictly convex Banach algebras, and Sadovskii’s fixed point problem in strictly convex Banach spaces. We then present applications of these best proximity point results to complex function theory, as well as the existence of a solution of a nonlinear functional integral equation and the existence of a mutually nearest solution for a system of integral equations. Full article

Edwardsiella tarda is a zoonotic facultative intracellular bacterium whose impact on farm-raised amphibians is still poorly defined. We recovered seven strains from American bullfrogs (Aquarana catesbeiana) on four farms in Guangdong, China, and combined field surveillance with molecular and pathological investigations. Phylogenetic analysis of 16S rRNA and rpoB sequences confirmed species identity. Quantitative PCR of 192 apparently healthy frogs revealed intestinal carriage at every farm, with prevalence ranging from 39 to 77 percent and bacterial loads of 10⁵–10⁶ CFU/mL, indicating widespread subclinical colonisation. Virulence profiling demonstrated a conserved core gene set (gadB, mukF, citC, fimA, ompA) and accessory variation confined to the flagellar gene fliC. The strains resisted trimethoprim, ampicillin, and tetracyclines, yet remained susceptible to third generation cephalosporins, carbapenems, and most aminoglycosides. Infection trials showed that although very high inocula caused acute fatalities, an inoculum of 10⁸ CFU/mL was sufficient to induce persistent enteritis characterised by suppressed tight junction proteins, elevated cytokine expression, and marked intestinal damage. These findings demonstrate that E. tarda circulates silently in bullfrog culture, carries an amphibian adapted virulence profile and still responds to key antimicrobials, providing a baseline for risk assessment, surveillance, and targeted control in amphibian aquaculture. Full article

The anaerobic co-digestion of agricultural residues emerges as a promising strategy for energy recovery and nutrient recycling within circular agricultural systems. This study aimed to optimize co-digestion parameters for vinegar residue (VR) and cattle manure (CM) using an orthogonal experimental design. Three key variables were investigated which are the co-substrate ratio (VR to CM), feedstock-to-inoculum (F/I) ratio, and total solids (TS) content. Nine experimental combinations were tested to evaluate methane yield, feedstock degradation, and digestate characteristics. Results showed that the optimal condition for methane yield comprised a 2:3 co-substrate ratio, 1:2 F/I ratio, and 20% TS, achieving the highest methane yield of 267.84 mL/g volatile solids (VS) and a vs. degradation rate of 58.65%. Digestate analysis indicated this condition generated the most nutrient-rich liquid digestate and solid digestate, featuring elevated N, P, and K concentrations, acceptable seed germination indices (GI), and moderate humification levels. While total nutrient content did not meet commercial organic fertilizer standards, the digestate is suitable for direct land application in rural settings. This study underscores the need to balance energy recovery and fertilizer quality in anaerobic co-digestion systems, providing practical guidance for decentralized biogas plants seeking to integrate waste treatment with agricultural productivity. Full article

A graph is called strongly Menger-edge connected (SME-connected) if any two vertices are connected by as many edge-disjoint paths as their smaller degree. For positive integers t and r, a graph G is called t-edge-fault-tolerant SME-connected (t-EFT-SME-connected) of order r if $G-F$ is SME-connected for any set F of edges in G with $\left|F\right|\le t$ and $\delta (G-F)\ge r$. We show that the n-dimensional folded crossed cube is $(n-1)$-EFT-SME-connected of order 1 and $(3n-5)$-EFT-SME-connected of order 2. Let $p(G,f)$ and $p^M(G,f)$ be the probabilities that G is connected and SME-connected when f edges are faulted randomly, respectively. We perform a numerical simulation on $p(G,f)$ and $p^M(G,f)$ for a five-dimensional folded crossed cube and folded hypercube. The numerical results show that, in addition to their same edge connectivity and SME connectivity, these two graphs have almost the same values of $p(G,f)$ and $p^M(G,f)$ for every f. This hints that, although the ‘edge-cross’ pattern in a hypercube-based graph can shorten the mean vertex distance, the ‘edge-cross’ is not a necessary pattern for strengthening the connectivity of the graph. Full article

Background: A short cervix in the second trimester significantly increases preterm birth risk, yet no reliable first-trimester prediction method exists. Current guidelines lack consensus on which women should undergo transvaginal ultrasound (TVUS) screening for cost-effective prevention. Therefore, it is vital to establish a highly accurate and economical method for use in the early stages of pregnancy to predict short cervix in mid-pregnancy. Methods: A total of 1480 pregnant women with singleton pregnancies and at least one risk factor for spontaneous preterm birth (<37 weeks) were recruited from January 2020 to December 2020 at the Shanghai First Maternity and Infant Hospital, Tongji University School of Medicine. Cervical length was assessed at 20–24 weeks of gestation, with a short cervix defined as <25 mm. Feature selection employed tree models, regularization, and recursive feature elimination (RFE). Seven machine learning models (logistic regression, linear discriminant analysis, k-nearest neighbors, support vector machine, decision tree, random forest, XGBoost) were trained to predict mid-trimester short cervix. The XGBoost model—an ensemble method leveraging sequential decision trees—was analyzed using Shapley Additive Explanation (SHAP) values to assess feature importance, revealing consistent associations between clinical predictors and outcomes that align with known clinical patterns. Results: Among 1480 participants, 376 (25.4%) developed mid-trimester short cervix. The XGBoost-based prediction model demonstrated high predictive performance in the training set (Recall = 0.838, F1 score = 0.848), test set (Recall = 0.850, F1 score = 0.910), and an independent dataset collected in January 2025 (Recall = 0.708, F1 score = 0.791), with SHAP analysis revealing pre-pregnancy BMI as the strongest predictor, followed by second-trimester pregnancy loss history, peripheral blood leukocyte count (WBC), and positive vaginal microbiological culture results (≥10⁵ CFU/mL, measured between 11⁺⁰ and 13⁺⁶ weeks). Conclusions: The XGBoost model accurately predicts mid-trimester short cervix using first-trimester clinical data, providing a 6-week window for targeted interventions before the 20–24-week gestational assessment. This early prediction could help guide timely preventive measures, potentially reducing the risk of spontaneous preterm birth (sPTB). Full article

Timely, objective pain recognition in dairy cattle is essential for welfare assurance, productivity, and ethical husbandry yet remains elusive because evolutionary pressure renders bovine distress signals brief and inconspicuous. Without verbal self-reporting, cows suppress overt cues, so automated vision is indispensable for on-farm triage. Although earlier systems tracked whole-body posture or static grimace scales, frame-level detection of facial micro-expressions has not been explored fully in livestock. We translate micro-expression analytics from automotive driver monitoring to the barn, linking modern computer vision with veterinary ethology. Our two-stage pipeline first detects faces and 30 landmarks using a custom You Only Look Once (YOLO) version 8-Pose network, achieving a 96.9% mean average precision ($mAP$) at an Intersection over the Union (IoU) threshold of 0.50 for detection and 83.8% Object Keypoint Similarity (OKS) for keypoint placement. Cropped eye, ear, and muzzle patches are encoded using a pretrained MobileNetV2, generating 3840-dimensional descriptors that capture millisecond muscle twitches. Sequences of five consecutive frames are fed into a 128-unit Long Short-Term Memory (LSTM) classifier that outputs pain probabilities. On a held-out validation set of 1700 frames, the system records 99.65% accuracy and an F1-score of 0.997, with only three false positives and three false negatives. Tested on 14 unseen barn videos, it attains 64.3% clip-level accuracy (i.e., overall accuracy for the whole video clip) and 83% precision for the pain class, using a hybrid aggregation rule that combines a 30% mean probability threshold with micro-burst counting to temper false alarms. As an early exploration from our proof-of-concept study on a subset of our custom dairy farm datasets, these results show that micro-expression mining can deliver scalable, non-invasive pain surveillance across variations in illumination, camera angle, background, and individual morphology. Future work will explore attention-based temporal pooling, curriculum learning for variable window lengths, domain-adaptive fine-tuning, and multimodal fusion with accelerometry on the complete datasets to elevate the performance toward clinical deployment. Full article