Search Results (700)

Rapid antimicrobial resistance (AMR) prediction from MALDI-TOF mass spectrometry (MS) remains challenging, particularly when training artificial intelligence (AI) models under small-sample constraints. Performance is often hampered by the high dimensionality of spectral data and the subtle nature of resistance-related signals: full-spectrum approaches risk overfitting to high-dimensional noise, whereas peak-selection strategies risk discarding structurally informative, low-intensity signals. Here, we propose ReShuffle-MS, a region-guided data augmentation framework for MS data. Each spectrum is partitioned into a Main Discriminative Region (MDR) and a Peripheral Peak Region (PPR). By recombining signals within the PPR across samples of the same class while keeping the MDR intact, ReShuffle-MS generates structure-preserving augmented samples. On a clinical dataset for Escherichia coli (E. coli) levofloxacin resistance prediction, ReShuffle-MS delivered significant and consistent performance gains. It improved the average accuracy of classical machine learning models by 3.7% and enabled a one-dimensional convolutional neural network (CNN) to achieve 83.25% accuracy and 97.28% recall. Visualization using Grad-CAM revealed a shift from sparse, peak-dependent attention toward broader and more meaningful spectral patterns. Validation on the external DRIAMS-C dataset for ceftriaxone resistance further demonstrated that the method generalizes to a distinct laboratory setting and a different antibiotic target. These findings suggest that ReShuffle-MS can enhance the robustness and clinical utility of AI-based AMR prediction from routinely acquired MALDI-TOF spectra. Full article

In this study, an adaptive event-driven coordination paradigm is proposed for achieving consensus in nonlinear multi-agent systems (MASs) over directed networks. First, a newly dynamic event-triggered mechanism with single-point historical information is introduced to minimize unnecessary network communication. And a more general form of an event triggering mechanism with moving window historical information is designed for further saving network resources. Considering that the use of historical information over a long period of time may cause deviations, an event-triggered mechanism that can adjust the maximum memory length is proposed in this work to minimize unnecessary network communication. Secondly, the unknown nonlinearities in the MAS model are addressed using the universal approximation capability of neural networks. Then, a methodology for distributed adaptive control under event-triggered mechanisms is introduced leveraging the memory-based command-filtered backstepping methodology, and the proposed scheme resolves the complexity explosion problem. Finally, a case study is conducted to validate the feasibility of the proposed method. Full article

Modern port operations face increasing challenges from rapidly changing weather and environmental conditions, requiring accurate short-term forecasting to support safe and efficient maritime activities. This study presents a sensor fusion-based machine learning framework for real-time multi-target nowcasting of wind gust speed, sustained wind speed, and wind direction using heterogeneous data collected at the Port of Livorno from February to November 2025. Using an IoT architecture compliant with the oneM2M standard and deployed at the Port of Livorno, CNIT integrated heterogeneous data from environmental sensors (meteorological stations, anemometers) and vessel-mounted LiDAR systems through feature-level fusion to enhance situational awareness, with gust speed treated as the primary safety-critical variable due to its substantial impact on berthing and crane operations. In addition, a comparative performance analysis of Random Forest, XGBoost, LSTM, Temporal Convolutional Network, Ensemble Neural Network, Transformer models, and a Kalman filter was performed. The results show that XGBoost consistently achieved the highest accuracy across all targets, with near-perfect performance in both single-split testing (R² ≈ 0.999) and five-fold cross-validation (mean R² = 0.9976). Ensemble models exhibited greater robustness than deep learning approaches. The proposed multi-target fusion framework demonstrates strong potential for real-time deployment in Maritime Autonomous Surface Ship (MASS) systems and port decision-support platforms, enabling safer manoeuvring and operational continuity under rapidly varying environmental conditions. Full article

Grain aging during storage leads to quality deterioration and significant economic losses. Traditional analytical approaches are often labor-intensive, slow, and inadequate for modern intelligent grain storage management. This review summarizes recent advances in the intelligent discrimination of grain aging using volatile organic compound (VOC) fingerprints combined with machine learning (ML) techniques. It first outlines the biochemical mechanisms underlying grain aging and identifies VOCs as early and sensitive biomarkers for timely determination. The review then examines VOC determination methodologies, with a focus on headspace solid-phase microextraction coupled with gas chromatography-mass spectrometry (HS-SPME-GC-MS), for constructing volatile fingerprinting profiles, and discusses related method standardization. A central theme is the application of ML algorithms, including Partial Least Squares Discriminant Analysis (PLS-DA), Support Vector Machines (SVM), Random Forest (RF), and Convolutional Neural Networks (CNN)) for feature extraction and pattern recognition in high-dimensional datasets, enabling effective discrimination of aging stages, spoilage types, and grain varieties. Despite these advances, key challenges remain, such as limited model generalizability, the lack of large-scale multi-source databases, and insufficient validation under real storage conditions. Finally, future directions are proposed that emphasize methodological standardization, algorithmic innovation, and system-level integration to support intelligent, non-destructive, real-time grain quality monitoring. This emerging framework provides a promising powerful pathway for enhancing global food security. Full article

Background. The aim of this study was to develop a predictive model of anticancer drug therapy toxicity in patients with gastric cancer. Methods. The retrospective study included 100 patients with stage II–IV gastric cancer who underwent 4 chemotherapy cycles. Initial significant toxicity factors included age, gender, height, body mass, body mass index, disease stage, skeletal muscle index (SMI), as well as plasma levels of trace elements (copper, zinc, selenium, manganese) and thyroid-stimulating hormone, cancer histology type and treatment regimen. The CTCAE v5.0 scale was employed to assess the severity of adverse events. Statistical analysis and building of mathematical neural network models were carried out in SPSS Statistics (v19.0). Results. Lower SMI values were associated with higher rates of toxicity-related complications of anticancer drug therapy (p < 0.05): leukopenia, hypoproteinemia, nausea, vomiting, cardiovascular events. Anemia, thrombocytopenia, hepatic cytolysis syndrome, nausea, diarrhea, constipation and stomatitis showed a weaker correlation with SMI. An increase in TSH was associated with higher rates of thrombocytopenia, nausea and vomiting. A decrease in Cu/Zn in plasma correlated with the severity of leukopenia and diarrhea, whereas Se/Mn showed an inverse correlation with the severity of anemia. Conclusions. Sarcopenia, abnormal thyroid status and imbalances in copper, zinc, selenium and manganese in blood plasma of patients with gastric cancer may be used as predictors of increased toxicity of anticancer drug therapy. Full article

Acoustic metamaterials are artificially engineered materials composed of subwavelength structural units, whose effective acoustic properties are primarily determined by structural design rather than intrinsic material composition. By introducing local resonances, these materials can exhibit unconventional acoustic behavior, enabling enhanced sound insulation beyond the limitations of conventional structures. In this study, a thin plate (thin sheet) refers to a structural element whose thickness is much smaller than its in-plane dimensions and can be accurately described using classical thin-plate vibration theory. When resonant mass blocks are attached to a thin plate, a thin-plate acoustic metamaterial is formed through the coupling between plate bending vibrations and local resonances. Thin-plate acoustic metamaterials exhibit excellent sound insulation performance in the low- and mid-frequency ranges. Multilayer configurations and the combination with porous materials can effectively broaden the insulation bandwidth and improve overall performance. However, the large number of structural parameters in multilayer composite thin-plate acoustic metamaterials significantly increases design complexity, making conventional trial-and-error approaches inefficient. To address this challenge, a neural-network-based inverse design framework is proposed for multilayer composite thin-plate acoustic metamaterials. An analytical model of thin-plate metamaterials with multiple attached cylindrical masses is established using the point matching and modal superposition methods and validated by finite element simulations. A multilayer composite unit cell is then constructed, and a dataset of 30,000 samples is generated through numerical simulations. Based on this dataset, a forward prediction network achieves a test error of 1.06%, while the inverse design network converges to an error of 2.27%. The inverse-designed structure is finally validated through impedance tube experiments. The objective of this study is to establish a systematic theoretical and neural-network-assisted inverse design framework for multilayer thin-plate acoustic metamaterials. The main novelties include the development of an accurate analytical model for thin-plate metamaterials with multiple attached masses, the construction of a large-scale simulation dataset, and the proposal of a neural-network-assisted inverse design strategy to address non-uniqueness in inverse design. The proposed approach provides an efficient and practical solution for low-frequency sound insulation design. Full article

With the increasing power output of aero-engines, combustor hot-gas mass flow rate and temperature continue to rise, posing more severe challenges to combustor structural cooling design. To enhance the film-cooling performance of the Z-profile feature in a reverse-flow combustor, this study performs a multi-parameter numerical optimization by integrating computational fluid dynamics (CFD), a radial basis function neural network (RBFNN), and a genetic algorithm (GA). The hole inclination angle, hole pitch, row spacing, and the distance between the first-row holes and the hot-side wall are selected as design variables, and the area-averaged adiabatic film-cooling effectiveness over a critical downstream region is adopted as the optimization objective. The RBFNN surrogate model trained on 750 CFD samples exhibits high predictive accuracy (correlation coefficient (R > 0.999)). The GA converges after approximately 50 generations and identifies an optimal configuration (Opt C). Numerical results indicate that Opt C produces more favorable vortex organization and near-wall flow characteristics, thereby achieving superior cooling performance in the target region; its average adiabatic film-cooling effectiveness is improved by 7.01% and 9.64% relative to the reference configurations Ref D and Ref E, respectively. Full article

The model of a liquid desiccant dehumidification air conditioning (LDAC) system is one of the key foundations for achieving efficient cooling, dehumidification and regeneration, and saving energy consumption. The data-driven modeling method does not need to understand the complex heat and mass transfer mechanism and equipment physical information, thus the modeling complexity is greatly reduced. This paper proposes a temperature and humidity prediction model integrating the Black Kite Algorithm (BKA), Bidirectional Temporal Convolutional Network (BiTCN), Bidirectional Long Short-Term Memory (BiLSTM), and Self-Attention mechanism (SA). The model extracts local spatiotemporal features from sequence data through BiTCN, enhances the understanding of contextual dependencies in temporal data using BiLSTM, and employs the SA to assign dynamic weights to different time steps. Furthermore, BKA is adopted to optimize the hyperparameter combinations of the neural network, thereby improving prediction accuracy. To validate the model performance, an experimental platform for an LDAC system was established to collect operational data under multiple working conditions, constructing a comprehensive dataset for simulation analysis. Experimental results demonstrate that compared to conventional time-series prediction models, the proposed model achieves higher accuracy in predicting outlet temperature and humidity across various operating conditions, providing reliable technical support for system real-time control and performance optimization. Full article

The environmental persistence of antibiotic residues in aquatic systems represents a critical global challenge, with photolysis serving as a primary abiotic degradation pathway. Traditional approaches to studying antibiotic photodegradation and transformation product (TP) identification face significant limitations, including complex reaction mechanisms, multiple concurrent pathways, and analytical challenges in characterizing unknown metabolites. The integration of artificial intelligence (AI) and machine learning (ML) technologies has begun to transform this field, offering new capabilities for predicting photodegradation kinetics, elucidating transformation pathways, and identifying novel metabolites. This comprehensive review examines current applications of AI/ML in antibiotic photolysis research, analyzing developments from 2020 to 2025. Key advances include quantitative structure–activity relationship (QSAR) models for photodegradation prediction, deep learning approaches for automated mass spectrometry interpretation, and hybrid computational–experimental frameworks. Machine learning algorithms, particularly Random Forests, support vector machines, and Neural Networks, have demonstrated capabilities in handling multi-dimensional environmental datasets across diverse antibiotic classes, including fluoroquinolones, β-lactams, tetracyclines, and sulfonamides. Despite progress in this field, challenges remain in model interpretability, standardization of datasets, validation protocols, and integration with regulatory frameworks. Future directions include machine-learning-enhanced quantum dynamics for improving mechanistic understanding, real-time AI-guided experimental design, and predictive tools for environmental risk assessment. Full article

Background: Folic acid supplementation (FAS) before and in early pregnancy prevents neural tube defects, but the benefits of extending FAS to late pregnancy on pregnancy outcomes remain unclear. We aimed to investigate the associations between duration of FAS and a spectrum of pregnancy outcomes, and to determine whether the associations were modified by maternal age or pre-pregnancy body mass index (BMI). Methods: This prospective multicenter study included 15,694 singleton pregnancies. We used mixed-effects log-binomial regression models to estimate the adjusted risk ratios (aRRs) and 95% confidence intervals (CIs) for gestational diabetes mellitus (GDM), gestational hypertensive disorders (GHDs), pre-eclampsia, preterm birth, macrosomia, small (SGA) and large for gestational age (LGA), and the interaction effects of advanced maternal age and pre-pregnancy BMI. Results: Of 15,694 women, 4523 (28.8%) did not take FAS before or during pregnancy, 2854 (18.2%) took FAS only during peri-pregnancy, 921 (5.9%) took FAS from peri- to mid-pregnancy, and 7396 (47.1%) took it through late pregnancy. Compared with women without FAS, those supplemented until mid-pregnancy were associated with lower risks of GHDs (aRR 0.84, 95% CI 0.74, 0.96) and pre-eclampsia (aRR 0.81, 95% CI 0.67, 0.97). Supplementation until late pregnancy was associated with lower risks of preterm birth (aRR 0.67, 95% CI 0.59, 0.76), SGA (aRR 0.74, 95% CI 0.63, 0.87), and LGA (aRR 0.88, 95% CI 0.79, 0.97). Among women of advanced maternal age or with overweight/obesity, supplementation until mid-pregnancy was associated with higher risk of GDM. Conclusions: Extending FAS until mid-pregnancy is associated with lower risks of GHDs and preeclampsia, and extending it until late pregnancy is associated with lower risks of preterm birth, SGA, and LGA. However, women of advanced maternal age or with overweight/obesity should be cautious about prolonging FAS. Full article

Glacier surface albedo controls solar energy absorption and Arctic mass balance, yet comprehensive modelling approaches remain limited. This study develops and validates multiple modelling frameworks for glacier albedo prediction using automatic weather station (AWS) data from Hansbreen and Werenskioldbreen in southern Svalbard during the 2011 ablation season. We compared three point-based approaches across elevation zones. At lower elevations (190 m), linear regression models emphasising snowfall probability or temperature controls achieved excellent performance (R² = 0.84–0.86), with snowfall probability contributing 65% and daily positive temperature contributing 86.3% feature importance. At higher elevations (420 m) where snow persists, neural networks proved superior (R² = 0.65), with positive degree days (72.5% importance) driving albedo evolution in snow-dominated environments. Spatial modelling extended point predictions across glacier surfaces using elevation-dependent probability calculations. Validation with Landsat 7 imagery and multi-algorithm comparison (n = 5) revealed that while absolute albedo values varied by 12% (0.54–0.60), temporal dynamics showed remarkable consistency (27.8–35.2% seasonal decline). Point-to-pixel validation achieved excellent agreement (mean absolute difference = 0.03 ± 0.02, 5.3% relative error). Spatial validation across 173,133 pixel comparisons demonstrated good agreement (r = 0.62, R² = 0.40, RMSE = 0.15), with an accuracy of reproducing temporal evolution within 0.001–0.021 error. These findings demonstrate that optimal glacier albedo modelling requires elevation-dependent approaches combining physically based interpolation with machine learning, and that temporal pattern reproduction is more reliably validated than absolute values. The frameworks provide tools for understanding albedo-climate feedback and improving mass balance projections in response to Arctic warming. Full article

Background and Objectives: Osteoporosis is a prevalent skeletal disorder characterized by decreased bone mass and compromised bone microarchitecture, leading to an elevated risk of fractures and significant morbidity, particularly among aging populations. Early diagnosis and personalized management are critical to reducing fracture incidence and associated healthcare burdens. Recent advances in artificial intelligence (AI) and machine learning (ML) have led to potential improvements in enhancing osteoporosis care by enabling accurate diagnostic imaging analysis, robust fracture risk prediction, and personalized therapeutic strategies. Materials and Methods: We performed a narrative review to summarize and critically evaluate the current literature on AI and ML applications in osteoporosis diagnosis and management. We searched relevant literature from inception to January 2025 to provide a comprehensive perspective, focusing on key themes, methodological approaches, and clinical implications. Results: Deep learning models, especially convolutional neural networks, facilitate rapid and accurate bone mineral density assessment from routine radiographs, expanding screening capabilities beyond conventional dual-energy X-ray absorptiometry (DXA). Machine learning algorithms harness clinical and demographic data to generate fracture risk models that often outperform traditional tools, enabling timely identification of high-risk individuals. Furthermore, AI-driven analyses of historical treatment responses coupled with real-time monitoring through wearable technologies and mobile applications allow for personalized therapeutic optimization and enhance patient engagement. Despite these promising advances, challenges remain regarding ethical considerations, data privacy, legal liability, incomplete model validation, lack of standardization, and the need for critical appraisal of real-world clinical efficacy for widespread clinical adoption. Conclusions: This narrative review indicates that AI and ML hold significant promise to revolutionize osteoporosis management by enabling early detection, precise risk stratification, and tailored interventions. However, the current evidence is heterogeneous, often lacking robust external validation and quantitative synthesis. Critical gaps include insufficient evaluation of model robustness across diverse populations, discussion of negative or conflicting results, and a comprehensive assessment of the limitations inherent in current AI evidence. Strategic efforts to validate, regulate, and critically integrate these technologies into routine clinical workflows are essential to realize their full potential and address the growing burden of osteoporosis worldwide. Full article

Background/Objectives: The ability to perceive and process emotional faces quickly and efficiently is essential for human social interactions. In recent years, humans have started to interact more regularly with robotic faces in the form of virtual or real-world robots. Neurophysiological research regarding how the brain decodes robotic faces relative to human ones is scarce and, as such, warrants further research to explore these mechanisms and their social implications. Methods: This study uses event-related potentials (ERPs) to examine the neural correlates during an emotional face categorization task involving human and robotic stimuli. We examined differences in brain activity elicited by viewing robotic and human faces expressing both happy and neutral emotions. ERP waveforms’ amplitudes for the P100, N170, P300, and P600 components were calculated and compared. Furthermore, mass univariate analysis of ERP waveforms was carried out to explore effects not limited to brain regions previously reported in the literature. Results: Results showed robotic faces evoked increased waveform amplitudes at early components (P100 and N170) as well as at the later P300 component. Further, only mid-latency and late cortical components (P300 and P600) showed amplitude differences resulting from emotional valences, aligning with dual-stage models of face processing. Conclusions: These results advance our understanding of face processing during human–robot interaction and contribute to our understanding of brain mechanisms underlying interactions when viewing social robots, setting new considerations for their use in brain health settings and broader cognitive impact. Full article

In this study, a multi-objective optimization framework based on an artificial neural network (ANN) was developed for an inlet perforated distributor plate in a 24S24P 10 kWh cylindrical lithium-ion battery module using immersion cooling. A combined Newman, Tiedeman, Gu and Kim with Computational Fluid Dynamics (NTGK-CFD) model was used to generate a symmetrically designed space by varying the input variables, including hole size A (mm), hole spacing ΔH (mm), and coolant mass flow rate V_in (kg/s). A three-level full factorial design was used to generate 27 cases, then CFD simulations were performed to provide a training data for the ANN model to predict the output variables, including maximum temperature T_max, maximum temperature difference ΔT_max, and pressure drop ΔP. The results show that the ANN model provides a reliable predictive model, capable of reproducing the thermal-hydraulic behavior of the immersion-cooled battery module with high fidelity via correlation coefficients R of 0.997 for all three output variables. In addition, Pareto-based optimization shows designs that balance cooling efficiency and pumping power. The selected optimal solution maintains T_max within the optimal range at 37.97 °C while reducing ΔP by up to 44%, providing a practical solution for large-scale battery module thermal management in EVs. Full article

Conservatively perturbed equilibrium (CPE) experiments yield transient concentration extrema that surpass steady-state equilibrium values. A physics-informed neural network (PINN) framework is introduced to simulate these over-equilibrium dynamics in linear chemical reaction networks without reliance on extensive time-series data. The PINN incorporates the reaction kinetics, stoichiometric invariants, and equilibrium constraints directly into its loss function, ensuring that the learned solution strictly satisfies physical conservation laws. Applied to three- and four-species reversible mechanisms (both acyclic and cyclic), the PINN surrogate matches conventional ODE integration results, reproducing the characteristic early concentration extrema (maxima or minima) in unperturbed species and the subsequent relaxation to equilibrium. It captures the timing and magnitude of these extrema with high accuracy while inherently preserving total mass. Through the physics-informed approach, the model achieves accurate results with minimal data and a compact network architecture, highlighting its parameter efficiency. Full article