Abstract: The bystander effect describes the induction of responses in non-targeted cells through cell signaling by directly stimulated cells. While this phenomenon has been extensively studied in the context of ionizing radiation, its occurrence following electrical stimulation (ES) remains poorly understood. Conditioned medium from N2a neuroblastoma cells exposed to voltage-controlled biphasic pulses at 500 mV/mm and 100 Hz induced neuronal differentiation in non-stimulated cells through an ES bystander effect. Bystander medium promoted morphological changes associated with neuronal differentiation, including increased neurite outgrowth and a reduction in the proliferation marker KI-67, indicating that the effects of ES extend to neighboring non-targeted cells. Molecular analysis revealed increased expression and secretion of interleukin-6 (IL-6) following ES, while neutralization of IL-6 receptor inhibited the effects of ES, highlighting the role of IL-6 as a key mediator of this effect. We provide the first evidence that ES promotes a differentiation-associated bystander effect mediated by IL-6.

Abstract: Inductively coupled plasma mass spectrometry (ICP–MS) was used to measure 24 elements in occipital scalp hair from 134 adults living in Campania, southern Italy (67 women, 67 men; age range 20–72 yr). Continuous models were restricted to the ten elements detected in at least 70% of adults: Al, Ba, Ca, Cr, Cu, K, Mg, Pb, Se and Zn. Multivariable log-linear models with HC3-robust standard errors adjusted for sex, age, fish consumption, medication and supplement use, current and passive smoking, hair treatments, cosmetic product use, tap-water use, occupational/environmental risk and sample mass, with multiple testing controlled by Benjamini–Hochberg false discovery rate (FDR) correction. Ca (median 1279.7g/g), Zn ( 153.5g/g), Mg ( 101.7g/g), K ( 14.8g/g) and Cu ( 13.9g/g) dominated the elemental profile. Pb had a median of 0.81g/g but an extreme right tail reaching 887.1g/g; Cd was detected in only 22.4% of adults, with a population median of zero. The most reproducible pattern was lower hair concentrations in males for Ca (Δ=−61.6%, q<0.001), Mg (Δ=−52.2%, q<0.001), Cu (Δ=−37.7%, q=0.003), Cr (Δ=−33.0%, q=0.011) and Zn (Δ=−23.9%, q=0.011). Sample mass remained inversely associated with Mg and Zn, indicating a potential analytical or censoring-related effect requiring quality-control attention. No fish, medication, supplement, smoking, tap-water or environmental-risk predictor survived FDR correction. An exploratory territorial display used medians for Al, Cr, Fe, Mn, Pb and Zn, and detection frequencies for the low-detection elements As, Cd and V, to visualize potential sentinel patterns. Exploratory unsupervised clustering separated a large mineral-enriched profile from a smaller mineral-depleted profile and, in a three-cluster sensitivity analysis, isolated a four-participant high-Pb/high-Ba sentinel subgroup; however, residual clustering after adjustment for sex, age, hair-care variables and sample mass was weak. These findings support hair biomonitoring as a screening tool in mixed-exposure regions and align with recent European calls for harmonised human biomonitoring, provided that sex, robust territorial aggregation, sample mass, non-detects, outliers and external contamination are handled explicitly.

Abstract: This research aims to help design artifacts that better fit their users, to promote their well-being. In urban spaces, well-being is affected by physical factors (such as the air quality) and by mental factors (the feeling the space affords its users). Focusing on mental well-being, a growing body of research focuses on the visual experience of urban spaces, and specifically on the aspect of visual impression. Studies of visual impression seek to link visual features of spaces to linguistic descriptions used by their users (e.g., referring to a space as “inviting”). Such works can be used to identify relations between visual features and our experience of space that are potentially generalizable, and thus useful for designing. However, as an emerging subfield, the study of visual impression of urban spaces is currently fragmented and would benefit from syntheses of the work done to date. Here, we systematically review studies of visual impression in urban settings, and or-ganize these across three dimensions: (1) purpose, (2) method and (3) findings. We identify research gaps in the field, challenges, and potential paths for moving forward.

Abstract: AlSb film has attracted the attention for its excellent properties, and many preparation methods have been explored. Herein, AlSb thin films were prepared by DC magnetron co-sputtering method and the interfacial behavior between the films and air molecules were investigated by X-ray diffraction（XRD）, Auger electron spectroscopy (AES) testing and density functional theory (DFT) calculations to elucidated the deliquescence process of AlSb thin films and its underlying mechanism. The results revealed that AlSb thin film exhibited Sb₂O₄ and Sb₂O₅ phases while the thin films doped Cu no longer showed any Sb oxide phases after the film exposed to air for one day. The chemical state of aluminum in the film remained stable along the depth direction, whereas antimony exhibited a pronounced gradient in chemical state from the surface to the interior. The oxidation state of Sb ions varied from -3 in the interior to +5 at the surface. The interaction between the (111) crystal plane of the AlSb film and air molecules is an exothermic process, with water molecules exhibiting the highest adsorption energy on the film surface, followed by oxygen molecules. The adsorption energies for nitrogen and carbon dioxide molecules were the lowest. Consequently, AlSb molecules readily combine with H₂O molecules. Furthermore, doping the AlSb film with copper or zinc atoms effectively reduced the adsorption energy for water and oxygen molecules, offering a new approach to suppress the deliquescence and oxidation of AlSb thin films. This study provides an important theoretical foundation for subsequent research on this material system.

Abstract: This paper examined a machine-learning technique for classifying riverine land covers in satellite images at different times with their normalized indices. The satellite images captured a river course in Kurobe River, Japan, in November 2017 and November 2018. The machine learning technique examined in this study was RF (random forests). The riverine land covers were classified into the following classes: tree, grass, bare gravel/sand bed, and water surface. This study trained RF using satellite image features in November 2017. Then, it applied the trained RF to those in November 2018 to examine the RF applicability to new satellite images with different radiance. A permutation importance analysis tried to detect essential satellite image features between the following ones: four multispectral band components, i.e., red, green, blue, and near-infrared, as well as three normalized indices, i.e., NDVI (normalized difference vegetation index), NDWI (normalized difference water index), and BNDVI (blue NDVI). The results showed that the F-measure of RF with the three normalized indices had a higher value, 0.74, than that without them, 0.63. In addition, the permutation importance analysis of the RF classification indicated that BNDVI had the highest value among all image features. These results supported the usefulness of the normalized indices for the machine-learning-based classification of riparian land covers at different times with different radiance of satellite images.

Abstract: Liquid-phase quartz crystal microbalance (QCM) biosensors are commonly quantified using scalar readouts such as frequency shift, dissipation, or bandwidth. Although these descriptors are useful, they can discard resonance line-shape morphology that contains additional information for quantitative inference. Here, we present a machine-learning-assisted impedance line-shape workflow for temporally robust quantitative 10 MHz QCM biosensing. A passive microfluidic mixer was used to generate controlled glycerol–water concentration gradients under constant total flow while complex impedance spectra were acquired in real time. Each sweep was parameterized using constrained Gaussian/Lorentzian models to obtain 52 physically interpretable descriptors spanning resistance, reactance, impedance magnitude, admittance components, and phase. The pipeline combines consensus outlier handling, training-fold mRMR feature ranking, and regression models evaluated under both corrected shuffled five-fold cross-validation and a stricter temporally blocked validation protocol. The shuffled reference model achieved R²=0.948 and RMSE = 0.162 %v/v, while seven-block temporal validation achieved R²=0.764, RMSE = 0.343 %v/v, and MAE = 0.246 %v/v. Under the same row-matched comparison, a Kanazawa frequency-shift baseline derived from the |Z| trough position yielded R²=0.331, RMSE = 0.578 %v/v, and MAE = 0.416 %v/v. Feature-set ablation further indicated that phase and conductance line-shape descriptors contain particularly strong temporally generalizable information. To demonstrate relevance for recognition-based biosensing, the workflow was also evaluated in an experimental DNA hybridization QCM assay in 1× PBS. Using all replicate-level predictions, the impedance line-shape ML pipeline reduced mean absolute percentage concentration error from 21.02% for conventional Δf calibration to 10.73%, corresponding to a 48.9% reduction in prediction error. These results show that impedance-resolved resonance morphology can improve quantitative QCM biosensing beyond scalar frequency-shift calibration and provides an interpretable basis for AI-assisted electronic biosensors.

Abstract: Prenatal detection of agenesis of corpus callosum (CC) anomalies is a challenge. Although a correct diagnosis of anomalies of the CC requires direct examination of the CC in the mid-sagittal plane of the fetal brain, the international guidelines on mid-trimester morphology scan do not recommend such direct examination of the CC in low-risk populations because of the associated technical difficulties. Recently, routine direct assessment of the CC with the mid-sagittal view has been recommended by several international experts in a consensus statement. The implementation of such routine direct assessment is not easy given the difficulties encountered in obtaining the mid-sagittal view of the CC. As such, it is the time to revisit the various two-dimensional and three-dimensional ultrasound techniques with a view to improve visualisation of the CC. The aim of this narra-tive review article is to discuss the use of various prenatal ultrasound screening methods of CC anomalies in-cluding standard axial views, a more detailed axial views, the mid-sagittal view, transvaginal approach, and 3D reconstruction. New insights on screening methods, and a pragmatic approach are also shared.

Abstract: Recent work has questioned both the theoretical foundations and empirical effectiveness of spectral graph neural networks (Spectral GNNs). This paper examines two aspects of the spectral graph learning paradigm. First, we review the scope and limitations of classical spectral graph theory, highlighting its emphasis on graph structure, extremal spectral quantities, and a narrow set of special graph families. Second, we trace the historical development of the Graph Fourier Transform (GFT), a key concept underlying Spectral GNNs. We identify three successive conceptual generalisations and show how concepts from Fourier and harmonic analysis were transferred to settings that lack the mathematical structures from which they originally derive their meaning. This perspective clarifies both the limitations inherited from spectral graph theory and the conceptual foundations on which Spectral GNNs were later built.

Abstract: Emergency and infrastructure-poor environments require coordination mechanisms that continue operating despite intermittent connectivity, limited bandwidth, and changing node participation. Existing low-bitrate mesh systems commonly rely on fixed timing configurations that either introduce unnecessary waiting under favourable conditions or become fragile as airtime and contention increase. This paper presents an adaptive quorum-based coordination framework for low-bitrate LoRa mesh networks that continuously adjusts execution timing from estimated LoRa airtime and observed participation. Each node executes a lightweight Monitor–Analyse–Plan–Execute–Knowledge (MAPE-K) control loop that derives slot spacing and coordination deadlines online. The framework combines quorum and full-participation finalisation with bounded single-retry recovery to keep local state deterministic. The approach was implemented on a four-node SX1276 LoRa mesh with opportunistic gateway support and evaluated under crash and omission faults with honest firmware. Across 2,514 coordination rounds, median completion time was 2.1 s in three-of-four quorum mode and 5.3 s under full participation, while all first-deadline misses completed through bounded recovery. Results indicate that adaptive timing reduces completion-time penalties associated with conservative static scheduling while preserving bounded execution behaviour on the evaluated testbed. The contribution is a cross-layer execution-timing framework that treats airtime as a runtime control signal for offline-first LoRa mesh coordination rather than a fixed deployment parameter.

Abstract: Small multirotor UAVs frequently operate in crosswind conditions, yet the aerodynamic interaction between windward and leeward propeller pairs remains incompletely understood. This study investigates the performance of a quadcopter propeller system under lateral crosswind using steadystate RANS simulations with the Transition SST turbulence model, validated against wind tunnel measurements (thrust and torque deviations within 5.4%). A parametric matrix of five rotational speeds (8,000 − 12,000 RPM) and six freestream velocities (0 − 10 m/s) is systematically examined. While thrust and power coefficients of all propellers increase monotonically with freestream velocity, the figure of merit (FM) of leeward propellers exhibits a previously unreported non-monotonic response: it decreases from hover, reaches a minimum near 6 m/s, and partially recovers at higher velocities. Windward propellers show no such degradation. Our velocity contour and streamline analyses reveal that this behaviour originates from windward wake ingestion into the leeward inflow region, which peaks at intermediate freestream velocities and is progressively alleviated as the stronger crosswind convects the wake downstream. The non-monotonic FM response is therefore a direct consequence of the competition between wake-induced inflow degradation and freestreamdriven aerodynamic augmentation. Our findings provide a systematic aerodynamic dataset essential for crosswind attitude control and propulsion system design in multirotor UAVs.

Abstract: Urban road networks in sub-Saharan Africa diverge from traditional sensor-rich, lane-disciplined environments. This paper established an empirical baseline assessment of network efficiency and velocity decay along the 8.0km Mile 17 to Governor's Roundabout corridor (8.0 km) in Buea, Cameroon. Characterized by radial geometry, a steep 25 m/km inbound gradient, unsignalized intersections, and a high mix informal transport (taxis comprising 59.5% and 65.27 of traffic for inbound and outbound flow, respectively). Fifteen-minute Passenger Car Unit (PCU) volume counts were recorded at three measurement points evaluated across three peak windows (morning, afternoon, and evening) over three consecutive weekday observation days. Volume-to-capacity ratios (v/c), peak-hour factors (PHF), equivalent hourly flow rates, space-mean speeds, and Mean Travel Time Indices (MTTI) were calculated for both inbound and outbound flow directions. Results reveal near-saturation and over-saturation at the Bonduma mid-corridor point during morning and afternoon peaks (v/c: 0.98 to 1.06). The highest operational stress occurred during the afternoon outbound window, where over-capacity conditions emerged simultaneously at Mile 17 and Bonduma. MTTI values ranged from 1.36 to 2.04, indicating that congestion is primarily driven by informal transport dynamics rather than absolute vehicle volume. Finally, the paper introduces the Directional Flow Asymmetry Index (DFAI) and Corridor Velocity Decay Rate (CVDR) to successfully quantify critical directional imbalances that traditional aggregate v/c analysis fails to resolve.

Abstract: Corn serves as one of China’s core staple crops which guarantees national grain security, yet frequent pest infestation severely cuts crop yields. Conventional crop protection measures suffer from low operating efficiency and inevitable environmental contamination, meanwhile existing mainstream detection algorithms are restricted by insufficient identification precision and high computational overhead. Targeting the above drawbacks, this research develops an improved RMS-YOLOv11 detection framework to achieve high-precision identification of corn pest individuals. The RFB component is embedded into the backbone’s feature extraction terminal of native YOLOv11 to strengthen feature extraction capacity for tiny pest targets; MobileNetV2 lightweight architecture is introduced to reduce model computational overhead; the original standard Conv layer is replaced with self-designed Conv-SWS module to balance multi-scale feature enhancement for targets with different sizes. Quantitative experimental outcomes reveal that the proposed RMS-YOLOv11 achieves 90% detection precision, 83% recall, 89.1% mAP@50, 65.5% mAP@75 and 59.3% mAP@50-95. These five indicators are separately raised by 3.1%, 4%, 4.5%, 2.4% and 3.6% compared with vanilla YOLOv11, and the proposed framework achieves better overall detection metrics against Faster-RCNN, SSD and other mainstream YOLO variants. Grad-CAM thermal visualization results verify that the optimized network can precisely lock the actual pest area and remedy the original network’s deficiency of inadequate feature attention toward small-size targets. The presented detection scheme supplies reliable technical support for field corn pest prevention and control and bears practical significance to safeguard domestic corn production safety.

Abstract: Railway transportation of hazardous materials (HazMat) exhibits a characteristic “low-probability, high-consequence” risk profile, compounded by spatially inequitable risk distribution across the network and increasingly stringent carbon emission constraints. Achieving synergistic optimization across safety, equity, and low-carbon dimensions constitutes a critical and unresolved scientific challenge. To address this, we formulate a robust multi-objective path optimization model (MORPO) that simultaneously minimizes three objectives: Conditional Value-at-Risk (CVaR) to capture extreme tail accident risk, the Gini coefficient to quantify regional risk-allocation equity, and traction-energy carbon emissions to represent ecological impact, all under freight-volume uncertainty. The Bertsimas–Sim robust counterpart theory is employed to equivalently transform the nonlinear uncertain constraints into deterministic linear constraints, ensuring computational tractability. To tackle the high-dimensional discrete nature, strong multi-objective conflicts, and non-convex Pareto-front characteristics of the problem, we propose an Adaptive Crossover-Mutation and Elite-preservation NSGA-II algorithm (ACE-NSGA-II), which integrates dual-strategy initialization, individual-level adaptive crossover and mutation operators, and a hierarchical elite preservation mechanism, thereby overcoming the premature convergence and front-degradation limitations of classical algorithms on three-dimensional non-convex fronts. The approach is validated on a representative North China railway freight network comprising 30 hub nodes and 50 mainline sections. Results demonstrate that: (1) ACE-NSGA-II significantly outperforms NSGA-II, NSGA-III, MOEA/D, and SPEA2 across IGD, HV, and Spread metrics, achieving a 58.7% reduction in IGD and a 23.7% increase in HV relative to standard NSGA-II; (2) under five uncertainty disturbance scenarios, the robust model yields a 13.3%–21.4% improvement in CVaR over the deterministic model, with larger advantages under stronger disturbances; (3) ablation experiments confirm that dual-strategy initialization and hierarchical elite preservation are the two most impactful components, and the four innovations exhibit significant positive synergistic effects; (4) parameter sensitivity analysis reveals that moderate robust conservatism (Γ∈[5,7]) paired with an appropriate population size (N∈[100,150]) achieves three-dimensional synergistic optimality. This work provides both a theoretical model and an algorithmic tool for safety–equity–low-carbon coordinated decision-making in railway hazardous materials transportation.

Abstract: This study examines whether renewable energy production is capable of covering energy costs in European Union family farms and identifies the determinants of the Energy Cost Coverage Ratio across different economic size classes. The analysis was based on Farm Sustainability Data Network (FSDN) data for 2014-2023 and combined descriptive statistics with panel data models. The results indicate that the Energy Cost Coverage Ratio remained relatively stable, fluctuating between 30.9% and 39.2%, despite a substantial increase in renewable energy production from €1594 to €2999 per farm. This limited improvement resulted from a simultaneous rise in energy costs, which increased from €5162 to €8465 per farm over the analysed period. Considerable differences in the Energy Cost Coverage Ratio were observed between economic size classes, while panel data models showed that its determinants vary across farm classes, with no single factor being significant for all classes. The findings suggest that renewable energy has strengthened the economic resilience of European farms, but its contribution remains insufficient to fully offset rising energy expenditures, highlighting the need for farm-size-specific policy support.

Abstract: Bioactive peptides are an important source of therapeutic molecules and molecular scaffolds involved in defense, signaling, and immune regulation. Despite the extraordinary diversity of Coleoptera, the structural landscape of beetle-derived bioactive peptides remains largely unexplored, limiting our understanding of their evolutionary diversity and biotechnological potential. Here, we performed a large-scale structural survey of predicted toxin-like peptide scaffolds across publicly available Coleoptera transcriptomes by integrating transcriptome mining, peptide maturation prediction, physicochemical characterization, AlphaFold 3 structural modeling, structural similarity analyses, and interpretable machine learning. We identified 291 candidate peptides, of which 155 contained canonical signal peptides and 273 produced mature peptides within the expected size range of known bioactive peptides. Structural analyses revealed that, despite extensive sequence diversity, many candidates converged toward a comparatively restricted repertoire of compact cysteine-rich architectures, suggesting that structural conservation exceeds primary sequence conservation during peptide diversification. Comparative structural analyses further identified recurrent protein architectures shared across multiple beetle lineages, while machine learning prioritization integrated structural and biochemical descriptors to identify high-confidence candidates for future functional characterization. These analyses establish the first structural atlas of predicted toxin-like peptides across Coleoptera and demonstrate that structure-guided transcriptome mining provides a powerful framework for uncovering evolutionarily conserved bioactive peptide scaffolds that would remain largely undetected using sequence-based approaches alone. Beyond expanding our understanding of peptide evolution in beetles, this resource is a foundation for future structural, functional, and biotechnological exploration of bioactive peptides in underexplored animal groups.

Abstract: The loss of a companion animal can provoke grief whose intensity rivals human bereavement, yet its biological basis remains largely uncharted. This structured narrative review synthesizes neurobiological, endocrinological, immunological, and psychosocial evidence into a mechanistic framework offered not as established fact but as a source of testable hypotheses. We trace the substrates of the human-animal bond across six interacting systems and propose how disruption of each upon an animal's death may drive grief. Three carry comparatively robust analogical support (oxytocinergic signaling, hypothalamic-pituitary-adrenal [HPA] axis regulation, and neuroinflammatory activation) and three are more exploratory (dopaminergic, serotonergic, and endocannabinoid pathways). Integrating extracted study-level data into transparent prioritization heuristics, we identify the HPA axis, assessed with its coupled oxytocinergic partner, as the most defensible first target for a direct biomarker study, a conclusion robust across three methodologically independent analyses. No direct pet-loss evidence yet supports any proposal, and this gap is the review's central motivation. As applied context, we address euthanasia-related guilt, disenfranchised grief under East Asian norms (a tentative "double disenfranchisement" hypothesis we want tested), assessment instruments, evidence-graded therapies, a Korean tool-development roadmap, and One Health research priorities.

Abstract: Hybrid renewable energy systems with photovoltaic generation, wind power, battery storage, and hydrogen pathways are becoming increasingly popular in off-grid and weak grid applications as they convert intermittent renewable energy sources into usable energy forms, thus decreasing dependency on fossil fuels. Traditional deterministic approaches to the system size may be unable to consider the risks associated with solar radiation, wind, ambient temperatures, and energy demand evolution due to climate-resource sensitivity analysis. In this research, HOMER Pro was used to apply least-cost sizing and resource availability sensitivity analysis for the sizing of hybrid photovoltaic–wind–battery-hydrogen system under four climate-resource sensitivity scenarios. Four different system layouts were compared based on the HOMER Pro optimization model using several evaluation criteria such as NPC, LCOE, renewable ratio, CO₂ emissions, unmet load ratio, and hydrogen energy production. Four different system layouts were compared based on the HOMER Pro optimization model using several evaluation criteria such as NPC, LCOE, renewable ratio, CO₂ emissions, unmet load ratio, and hydrogen energy production. The most appropriate design layout was identified as a PV-wind-battery-electrolyzer-hydrogen tank-fuel cell system with 900 kW PV, 350 kW wind power capacity, 1.85 MWh battery storage, 240 kW electrolyzer capacity, 520 kg hydrogen capacity and 180 kW fuel cell. At the base scenario, it produced the minimum LCOE equal to 0.246 USD/kWh, NPC of 4.38 million USD, renewable ratio of 98.7%, CO₂ emissions of 18 t/y and unmet load of 0.30%. In case of simultaneous pressure on resources due to climate, the LCOE rose to 0.253 USD/kWh while the unmet load did not exceed 0.43%. Thus, the inclusion of hydrogen allows one to ensure greater resilience compared to systems without hydrogen.

Abstract: Feature selection is one of the fundamental steps in producing simpler, more interpretable, and computationally efficient models while maintaining predictive power. In this study, rather than fixing the uncertainty relationship between the target variable and each candidate feature to a single entropy parameter, three volume-based indices are proposed that integrate the two-parameter structure of the Sharma-Mittal entropy over a defined parameter region. The indices, named Parameter-Integrated Conditional Sharma-Mittal Entropy, Parameter-Integrated Sharma-Mittal Entropy Information Gain, and Normalized Parameter-Integrated Sharma-Mittal Entropy Information Gain, represent, respectively, the conditional entropy volume, the information gain volume, and the form of this volume normalized relative to the total entropy volume of the target, respectively. Densities were estimated using Gaussian kernel density estimation; the alpha and beta parameters were numerically integrated over the range [0.05; 0.95] × [0.05; 0.95]. The methods were compared across six different regression datasets-Airfoil Self-Noise, AirQualityUCI, BodyFat, meteorology-based reference evapotranspiration, Concrete, and WineQualityWhite-using absolute Pearson correlation, absolute Spearman rank correlation, Shannon information gain, mutual information, and random forest variable importance. The comparison was conducted using raw importance scores, derived feature rankings, Spearman’s rho and Kendall’s tau rank correlations, bivariate rank scatter plots, and Mann–Whitney U test results. The results show that the proposed indices produced high or very high rank agreement with correlation- and information-based reference methods in the Airfoil Self-Noise, AirQualityUCI, BodyFat, and meteorological datasets. Specifically, the absolute Pearson correlation, absolute Spearman rank correlation, Shannon information gain, and mutual information, along with the average Spearman rho values across all datasets, were obtained as 0.884, 0.874, 0.883, and 0.903, respectively; the average agreement with random forest importance scores remained at 0.538. The findings reveal that the parameter-integrated Sharma–Mittal framework offers a density-based and discretization-independent filtering perspective for continuous data; however, they also highlight the need for additional sensitivity and out-of-sample prediction performance evaluations regarding absolute score scales, negative information-gain values, and the rank equivalence of the three proposed indices.

Abstract: Objective: To develop and evaluate EDBERT (Emergency Department Bidirectional Encoder Representations from Transformers), a BERT-based architecture built around a multi-input fusion network that enriches free-text triage notes with the additional context of presenting complaints and patient age to predict Emergency Department (ED) disposition, supporting early decision-making and reduced ED length of stay (LOS). Methods: A retrospective cohort of 570,143 ED presentations to the Royal Melbourne Hospital, Australia, was used. At the core of EDBERT, a fusion network integrates three complementary input sources—triage notes, presenting complaints, and patient age—through learnable weights, with the two free-text inputs encoded by customised 8-layer BERT encoder stacks pre-trained on triage text with Masked Language Modelling (MLM). The resulting architecture is also lightweight, containing 66 million parameters, substantially fewer than the 110 million of BERT-BASE. Eighty percent of the dataset was used for training and 20% for testing. Performance was benchmarked against BERT-Base configurations of varying depth. Results: EDBERT achieved an accuracy of 83.26%, a macro-averaged F1-score of 81.99%, and an area under the receiver operating characteristic curve of 0.91, outperforming all single-input BERT-Base variants, including a depth-matched 8-layer model (82.08%), indicating that the gain derives from the fusion of complementary inputs combined with domain-adaptive pre-training rather than from model capacity. Conclusion: A fusion-based BERT architecture that supplements the triage narrative with presenting complaint and age context predicts ED disposition more accurately than larger single-input BERT models, while requiring substantially less memory and computation, making it practical for deployment in hospitals with limited computing resources.

Abstract: Background: Glucagon-like peptide-1 (GLP-1) and dual GLP-1/glucose-dependent insulinotropic polypeptide (GIP) receptor agonists reduce cardiovascular events, but the underlying cardiac structural remodeling remains unclear. Objectives: This study evaluated longitudinal cardiac remodeling associated with semaglutide and tirzepatide in a real-world cohort and determined its weight-loss dependency. Methods: Using electronic health records from a federated network, we analyzed longitudinal echocardiograms of patients prescribed semaglutide or tirzepatide, stratified by 12-month weight loss into super-responders (>15%), moderate-responders (5%-15%), and minimal-responders (<5%). Additionally, GLP-1/GIP patients with >5% weight loss were propensity-matched with non-GLP-1/GIP weight-loss medication control patients on demographics and baseline body mass index. Results: Among 3,500 GLP-1/GIP patients (422 weight-loss super-responders, 1,426 moderate-responder, 1,652 minimal-responders), left ventricular (LV) mass decreased across all groups proportional to weight loss (super-responders: 198.4 ± 69.6 to 176.6 ± 66.0 g; moderate-responders: 210.7 ± 74.0 to 195.8 ± 68.1 g; minimal-responders: 216.1 ± 68.4 to 204.8 ± 67.4 g; p < 0.001). Left atrial volume and LV systolic and diastolic function did not change. Among weight-loss super-responders, semaglutide (n = 159) showed greater LV mass reduction than tirzepatide (n = 89) (Cohen’s d = -0.389 vs. -0.263), while tirzepatide showed small improvements in right ventricular function. Compared to matched non-GLP-1/GIP controls (n = 118/group), GLP-1/GIP patients exhibited significant LV mass reduction not observed in controls despite similar weight loss. Conclusions: Real-world findings demonstrate that incretin-based therapies are associated with reverse cardiac remodeling characterized by weight-loss-dependent and incretin-specific mechanisms, with potentially distinct structural targets among agents.