Search Results (33)

Infrared small-UAV detection remains difficult because the target often appears as a weak thermal point rather than a clear object. This problem is clear in the SIDD dataset, where most test targets are smaller than 32 × 32 pixels. To address this case, this paper proposes S-Drone-YOLO, a compact YOLO-based detector that maintains a high-resolution P2 prediction path and leverages it carefully during classification. The model starts from a lightweight YOLOv5-style detector. It adds a stride-4 P2 path and replaces the C3 neck blocks with C2fAttn to improve feature reuse before prediction. Two components are then added to the Architecture II design. The Coordinate-Aware Residual C2f Block, CAR-C2f, strengthens the P2 branch using coordinate attention and residual scaling. The P2-Guided Quality-Aware Detection Head (P2-QADH) combines local P2 details with nearby P3 context. It produces a quality map that adjusts the classification logits. The regression branch, output tensor format, and training loss interface remain unchanged. On the SIDD infrared drone dataset, S-Drone-YOLO reaches 0.988 precision, 0.939 recall, 0.699 mAP50-95, and 0.962 F1-score. It uses 6.45 M parameters and 31.3 GFLOPs. Compared with the Architecture I model, recall increases by 0.8 percentage points and mAP50-95 increases by 0.4 percentage points. At the same time, the parameter count decreases by 20.3%, and GFLOPs decrease by 43.7%. Fine-tuning on five RGB UAV datasets and a second thermal dataset (ThermalUAV2UAV) yields F1 scores ranging from 0.941 to 0.999, with an mAP50-95 of 0.843 on the thermal dataset. The background analysis also shows stable F1-scores across sky, sea, city, and mountain scenes. These results suggest that controlled P2 guidance can improve infrared small-UAV detection while keeping the model size practical. Full article

Large Language Models (LLMs) used for clinical decision support must not only make accurate predictions but also generate rationales that are consistent with, and sufficient for, those predictions. Building on Reason2Decide, a two-stage rationale-driven multi-task framework, we propose Reason2Decide-C (R2D-C, where C denotes cycle consistency), which augments Reason2Decide’s stage 2 training with confidence-adaptive scheduled sampling and cycle-consistent rationale-to-label training. In stage 1, we pretrain our model on rationale generation. In stage 2, we jointlytrain on label prediction and rationale generation, gradually replacing gold labels with model-predicted labels based on confidence. Simultaneously, we feed the rationale logits back into the model to recover the label, thus enforcing explanation sufficiency. We evaluate R2D-C on one proprietary triage dataset, as well as public biomedical QA and reasoning datasets. Across model sizes, R2D-C substantially improves rationale–prediction consistency (where stage 1 and stage 2 predictions agree) and sufficiency (where the rationale alone recovers the ground-truth label) over other baselines while matching or modestly improving predictive performance (F1); in several settings R2D-C surpasses $40\times$ larger foundation models. Ablations confirm that the full combination is optimal, maximizing alignment and LLM-as-a-Judge rationale quality. These results demonstrate that confidence-adaptive scheduled sampling and cycle-consistent rationale-to-label training substantially enhance explanation alignment without sacrificing accuracy. Full article

Interchange merging areas are critical safety hotspots in urban road networks, where complex vehicle interactions challenge traffic safety and efficiency. Improving safety performance at these locations is essential for developing sustainable, resilient, and intelligent urban transportation systems. To overcome the limitations of single-driver simulators, this study developed a multi-driver simulation platform based on Unity3D (Version 2022.3.1f1c1), enabling real-time interaction among multiple human drivers. High-resolution trajectory data were collected from 231 valid interaction events. An eight-direction relative position model was employed to classify behaviors into four patterns: longitudinal, lateral, front cut-in, and rear cut-in. Risk was quantified using time-exposed and time-integrated Anticipated Collision Time metrics, with events subsequently clustered into low (n = 138), medium (n = 67), and high-risk (n = 26) categories. An ordered logit regression model identified key risk factors. The results quantitatively demonstrate that interaction risk escalates significantly with abrupt speed changes (OR = 16.22) and late-stage occurrence of speed extremes (OR = 6.76) in the interacting vehicle, as well as large initial speed differences (OR = 2.45). Conversely, stable speed regulation and adaptive acceleration by the subject vehicle proved to be potent mitigating factors. These findings provide actionable insights for the development of intelligent collision warning systems and the sustainable design of interchange infrastructure. Full article

Developing tools to assess university lecturers on the quality of their teaching practice is a priority for achieving Sustainable Development Goal (SDG) 4. The aim of this research is therefore to propose a typology of teaching profiles and analyse the factors that influence each of them, with a view to proposing a model for identifying lecturers who need support. Based on 10,938 observations made between August 2021 and June 2025, a heuristic method was used to classify all observations into six quadrants according to student academic performance and student evaluations of their teachers. Based on the classification of all observations into six quadrants, called teacher profiles, logistic regressions were performed in Gretl software version 2024c to identify which characteristics inherent to the teacher, the educational context, and the curricular stage explain the classification of the teacher into a particular profile. The results indicate that full-time teachers and those with higher academic qualifications tend to obtain higher scores on the SET, while online or hybrid modalities are associated with lower scores on the SET. In addition, the six teaching profiles obtained an accuracy of over 80% in the logit models, highlighting significant effects of age, modality, and curricular level (p < 0.05) on the probability of belonging to each quadrant. It is concluded that factors related to curricular advancement, the educational context, and those inherent to the teacher can explain the proposed typology. This typology could serve as a management tool to identify teachers who need specific support to move towards profiles that are ideal for the university institution. Among the main limitations of this study are the heuristic methodology and the fact that the data were obtained from a single educational institution. Full article

Background: Locus of control (LoC) may shape how older adults appraise aging, particularly in acute geriatric rehabilitation. Evidence linking internal/external LoC to domain-specific Views on Aging (VoA, containing Physical Loss, Social Loss, Personal Growth, Self-awareness/Gains) remains limited. Methods: We analyzed a cross-sectional cohort of patients aged 70 and above from an acute geriatric rehabilitation unit (N = 103) and contextualized findings with a 1:1 Mahalanobis-matched subsample from the German Ageing Survey. Internal and external LoC and covariates (age, sex, Barthel, cognitive function, depressive symptoms, health satisfaction) were standardized (z). Associations were estimated using (i) ordinary least squares (OLS) regression across eight LoC effects, as well as (ii) proportional-odds ordinal models (quartiles; logit link), as a complementary, distribution-robust approach. Results: For the Physical VoA domain, higher internal LoC related to more positive appraisals (OLS β = 0.133, 95% CI 0.043–0.223, p = 0.035; OR = 3.52), whereas higher external LoC related to less positive appraisals (β = −0.165, 95% CI −0.285 to −0.045, p = 0.035; OR = 0.274). Internal LoC also increased the odds of more positive Personal Growth (OR = 1.64, 95% CI 1.04–2.72), while effects on Social Loss (external LoC OR = 0.649, 95% CI 0.418–0.991) and Gains were smaller. Univariate Spearman correlations were directionally consistent. In the DEAS comparison, older patients showed greater endorsement of both physical losses and gains. Conclusions: In acute geriatric rehabilitation, internal control beliefs align with more positive views of physical aging and growth, whereas external control aligns with less positive physical (and modestly social) views. The results position LoC as a clinically relevant correlate of aging appraisals. Full article

AmpC β-lactamases are class C enzymes that hydrolyze penicillins, cephalosporins, and monobactams. The WHO recently classified third-generation cephalosporin-resistant and carbapenem-resistant Enterobacterales as critical pathogens. We conducted a systematic review and meta-analysis to evaluate AmpC prevalence in hospital isolates across South America. We searched PubMed/MEDLINE, SciELO, and Google Scholar. We included 69 observational studies that phenotypically or genotypically identified AmpC producers. A random-effects generalized linear mixed model with logit transformation estimated pooled prevalence; heterogeneity and moderators were explored through subgroup analyses and meta-regression. Seventy studies, including 48,801 isolates, were eligible. AmpC β-lactamases were detected in 11.7% of isolates (95% CI 11.4–12.0), with extreme heterogeneity (I² ≈ 97%). Enterobacter species showed the highest prevalence (~46%), whereas Escherichia spp. had the lowest (~4.5%) prevalence of AmpC positivity within each genus. Meta-regression indicated that studies focusing on a single genus reported higher prevalence and that including pediatric patients was associated with a lower prevalence of AmpC-positive microorganisms among isolates. Quality of evidence was rated low due to inconsistency, moderate risk of bias, and indirectness of data. AmpC producers are entrenched in South American hospitals, and species-aware surveillance and harmonized detection are critical to guide empiric therapy and antimicrobial stewardship. Full article

The widespread deployment of large language models (LLMs) has raised urgent demands for verifiable content attribution and misuse mitigation. Existing text watermarking techniques often struggle in black-box or sampling-based scenarios due to limitations in robustness, imperceptibility, and detection generality. These challenges are particularly critical in open-access settings, where model internals and generation logits are unavailable for attribution. To address these limitations, we propose CWS (Contrastive Watermarking with Semantic Modeling)—a novel keyless watermarking framework that integrates contrastive semantic token selection and shared embedding space alignment. CWS enables context-aware, fluent watermark embedding while supporting robust detection via a dual-branch mechanism: a lightweight z-score statistical test for public verification and a GRU-based semantic decoder for black-box adversarial robustness. Experiments on GPT-2, OPT-1.3B, and LLaMA-7B over C4 and DBpedia datasets demonstrate that CWS achieves F1 scores up to 99.9% and maintains F1 ≥ 93% under semantic rewriting, token substitution, and lossy compression (ε ≤ 0.25, δ ≤ 0.2). The GRU-based detector offers a superior speed–accuracy trade-off (0.42 s/sample) over LSTM and Transformer baselines. These results highlight CWS as a lightweight, black-box-compatible, and semantically robust watermarking method suitable for practical content attribution across LLM architectures and decoding strategies. Furthermore, CWS maintains a symmetrical architecture between embedding and detection stages via shared semantic representations, ensuring structural consistency and robustness. This semantic symmetry helps preserve detection reliability across diverse decoding strategies and adversarial conditions. Full article

Background/Objectives: Omega-3 highly unsaturated fatty acids (HUFAs), particularly EPA and DHA, have anti-inflammatory and lipid-modulating properties for treating atherosclerosis. However, the relationship between plasma fatty acid profiles, omega-3 status, and statin efficacy in carotid atherosclerosis remains poorly defined. Objectives: This study evaluates plasma and plaque fatty acid (FA) composition, explores their associations with plaque stability, and examines the relationship of omega-3 levels, lipid biomarkers (VLDL-C, LDL-C, HDL-C, total cholesterol, and triglycerides) with statin and β-blocker treatment. A mathematical model was developed to predict the erythrocyte omega-3 index from plasma. Methods: In this case–control study, plasma and carotid plaques of 52 patients undergoing carotid endarterectomy were analyzed. Plasma was compared with that of 50 healthy controls. FAs were quantified by LC-MS/MS. Plaques were histologically classified as stable or unstable. Results: Atherosclerotic patients showed disturbed FA metabolism, including decreased plasma omega-3 EPA + DHA, SFAs and HUFAs, increased MUFAs, and impaired desaturase and elongase activity. Unstable plaques had higher MUFA and lower HUFA content compared with stable plaques. Significant correlations between plasma EPA + DHA and HDL-C and triglycerides were observed in statin-naïve patients, whereas statins appeared to attenuate these associations. Co-treatment with β-blockers had no significant effect. A validated logit-based model accurately predicted the erythrocyte omega-3 index from plasma (R² = 0.782). Conclusions: Altered plasma and plaque FA profiles correlate with atherosclerosis’s plaque instability and inflammatory lipid profiles. Statins significantly influence these associations, suggesting their complex interaction with lipid metabolism. Plasma measurements of omega-3 fatty acids in combination with predictive modelling may be beneficial for diagnostic and therapeutic monitoring in carotid atherosclerosis. Full article

Background: Recently, elevated levels of plasma erythritol have been associated with major adverse cardiovascular events (MACE). It is known that people with HIV (PWH) have a higher cardiovascular disease burden. Whether PWH have higher levels of plasma erythritol has not been evaluated. This study aimed to assess if blood erythritol levels are elevated in PWH and to examine relationships between erythritol and dietary, cardiometabolic, inflammatory, and gut health markers. Methods: Plasma erythritol levels were measured using frozen samples from 162 participants, including 109 PWH and 53 people without HIV (PWoH) in a parent study. General linear models were used to assess the linear relationship between characteristics, cardiovascular measures, markers of body composition, inflammation, and gut integrity with plasma erythritol. Logistic regression was used to assess risk factors associated with PWH, and cumulative logit models were used to investigate which factors were associated with having the highest plasma erythritol levels among PWH. Results: Compared to PWoH, PWH had higher plasma erythritol levels (p = 0.03). Every 10% increase in VLDL (p = 0.01), visceral adipose tissue (p < 0.0001), or TNFrI (p = 0.01) was associated with an approximately 1% increase in plasma erythritol. Among PWH, HgbA1c (p = 0.003), TNFrI (p = 0.002), and IFAB-P (p = 0.004) were associated with having the highest tertile of plasma erythritol (≥3.6 μM). Compared to PWoH, PWH were more than two times as likely (p = 0.03) to have plasma erythritol ≥ 3.6 μM. Conclusions: We identified positive associations between plasma erythritol levels and several factors, including HIV status, BMI, adipose tissue, TNFr1, HbA1c, and VLDL. These results underscore the importance of further investigating the role of elevated plasma erythritol levels in people with HIV, particularly in light of their increased vulnerability to cardiovascular and metabolic diseases. Full article

This study aims to provide valuable insights into consumer preferences for delivery services in online shopping in Chile. The COVID-19 pandemic has accelerated the evolution of delivery and logistics services, leading to increased competition among online stores. Chile, with its highly digitally enabled population and a competitive landscape of online retailers, serves as an ideal reference case for Latin America. By analyzing key delivery attributes such as delivery time, order arrival time range, compensation policies for delivery delays, and delivery prices, we offer valuable insights into consumer behavior. These insights will, in turn, inform the formulation of effective strategies within the online shopping industry. We examine the following aspects: (a) The willingness of consumers to pay for the service attributes; (b) The relative importance assigned to these attributes by consumers; and (c) The relationship between consumer preferences and socioeconomic characteristics. Using Multinomial Logit Models and a database from a Discrete Choice Experiment, we have discovered that the most significant attributes of delivery service are the time until product arrival and the existence of compensation in case of delivery delays. Additionally, we found that consumers are willing to pay more for the same delivery service if the product is large, as large products generally have higher prices. Furthermore, we observed that delivery time preferences vary by gender and for small products, and price sensitivity varies according to educational level, household size, and socioeconomic status. To the best of our knowledge, no previous research of this kind has been conducted for Chile. Full article

This study investigated the interplay between exit selection models and local pedestrian movement patterns within floor field frameworks. Specifically, this investigation analysed the performance of a multinomial logit exit choice model, incorporating both expected utility theory and cumulative prospect theory frameworks when coupled with three distinct local-level pedestrian movement models (FF-Von Neumann, FF-Moore, and NSFF). The expected utility theory framework considers the deterministic component as a linear relationship, while the cumulative prospect theory framework further considers the decision-maker’s risky attitudes by transforming objective terms into subjective terms using a power value function. The core objective was to comprehend how local movement dynamics, as represented by the floor field models, influence decision-making during exit selection. Comparative analyses revealed intriguing variations between the three local models, despite their shared expected utility theory-based exit choice framework. These discrepancies stemmed from the diverse pedestrian trajectory behaviours generated by each model. Consequently, these local dynamics impacted the decision-maker’s assessment of critical factors, such as the number of evacuees close to the decision-maker ($NCDM$) and the number of evacuees close to an exit ($NCE$), which the exit choice model incorporates. These assessments, in turn, significantly affected higher-level decision-making. The integration of the three models with the multinomial logit exit choice model, using either cumulative prospect theory and expected utility theory frameworks, further strengthened the observed bilateral relationship. While the specific nature of this relationship varied depending on the chosen framework and its implementation details, these consistent findings demonstrate the robustness of the results. This reinforced the influence of local-level pedestrian dynamics on higher-level exit selection, highlighting the importance of accurate crowd dynamics modelling, especially when advanced exit choice models consider local movement factors. Full article

The unmet need for assistance in activities of daily living (ADLs) accentuates older persons’ risk of falls, ill health, hospitalisation, and mortality. In Nigeria, the family arrangements through which older persons derive support are changing due to modernisation, migration, and economic challenges. How the family dynamics explain the unmet needs is poorly understood. This study investigates the influence of family and household structures on older persons’ unmet needs in ADLs in southwestern Nigeria. The study analysed the data of 827 older adults aged ≥65 years selected from Oyo State, southwestern Nigeria, using a multi-stage sampling design. Associations were examined using the Poisson–logit hurdle regression model. From the results, 65% of older persons with difficulties had unmet needs in instrumental ADLs and 59% in basic ADLs. Increased unmet needs were associated with older persons living with non-family members (β = 0.19; p < 0.01; 95% C.I. = 0.05–0.32) and widows (β = 0.27; p < 0.01; 95% C.I. = 0.13–0.42). Conversely, unmet needs decreased with higher family size (β = −0.06; p < 0.001; 95% C.I. = −0.08–−0.03), living in rich households (β = −0.29; p < 0.001; 95% C.I. = −0.42–−0.17), not being the household head (β = −0.27; p < 0.001; 95% C.I. = −0.40–−0.15), close family bonds, and proximity to children/caregivers. The study recommends alternative or complementary home-based support mechanisms for seniors with vulnerable family settings in southwestern Nigeria. Full article

Introduction: Early detection of type 2 diabetes is essential for preventing long-term complications. However, screening the entire population for diabetes is not cost-effective, so identifying individuals at high risk for this disease is crucial. The aim of this study was to compare the performance of five diverse machine learning (ML) models in classifying undiagnosed diabetes using large heterogeneous datasets. Methods: We used machine learning data from several years of the National Health and Nutrition Examination Survey (NHANES) from 2005 to 2018 to identify people with undiagnosed diabetes. The dataset included 45,431 participants, and biochemical confirmation of glucose control (HbA1c) were used to identify undiagnosed diabetes. The predictors were based on simple and clinically obtainable variables, which could be feasible for prescreening for diabetes. We included five ML models for comparison: random forest, AdaBoost, RUSBoost, LogitBoost, and a neural network. Results: The prevalence of undiagnosed diabetes was 4%. For the classification of undiagnosed diabetes, the area under the ROC curve (AUC) values were between 0.776 and 0.806. The positive predictive values (PPVs) were between 0.083 and 0.091, the negative predictive values (NPVs) were between 0.984 and 0.99, and the sensitivities were between 0.742 and 0.871. Conclusion: We have demonstrated that several types of classification models can accurately classify undiagnosed diabetes from simple and clinically obtainable variables. These results suggest that the use of machine learning for prescreening for undiagnosed diabetes could be a useful tool in clinical practice. Full article

The purpose of this study was to develop a prediction model for stroke-associated pneumonia (SAP) based on risk factors for SAP and to suggest nursing interventions to prevent SAP. In addition, a nomogram was developed to enhance its utility in nursing practice. The retrospective cohort study included 551 patients hospitalized for acute ischemic stroke at a university hospital in South Korea. Data were collected through a structured questionnaire and a review of the electronic medical record (EMR). In the development of a predictive model for SAP, multivariate logistic regression analysis showed that independent risk factors for SAP were age ≥ 65 years, National Institute of Health Stroke Scale (NIHSS) score ≥ 7, nasogastric tube feeding, and C-reactive protein (CRP) ≥ 5.0 mg/dL. The logit model was used to construct the SAP prediction nomogram, and the area under the curve (AUC) of the nomogram was 0.94. Furthermore, the slope of the calibration plot was close to the 45-degree line, indicating that the developed nomogram may be useful for predicting SAP. It is necessary to monitor the age, NIHSS score, nasogastric tube feeding status, and CRP level of stroke patients and identify high-risk groups using the developed nomogram to provide active nursing interventions to prevent SAP. Full article

African Swine Fever (ASF) is a highly infectious disease, severely affecting domestic pigs and wild boar. It has significantly contributed to economic losses within the pig farming industry. As a critical component of biosecurity measures, the selection of cleaning and disinfection (C&D) procedures is a dynamic and long-term decision that demands a deeper knowledge base among pig farmers. This study uses a binary logit model to explore the effect of epidemic prevention training on the adoption of C&D procedures among pig farmers with irregular and regular C&D procedures based on micro-survey data obtained from 333 pig farmers from Sichuan. The endogeneity issue was handled using propensity score matching, resulting in solid conclusions. In addition, the critical mediating impact of biosecurity cognition was investigated using a bootstrap analysis. The empirical study demonstrated that epidemic prevention training encourages pig farmers to adopt C&D procedures, with biosecurity cognition significantly mediating. Furthermore, epidemic prevention training was more likely to promote the adoption of C&D procedures among pig farmers with shorter breeding experiences and those having breeding insurance. Our study emphasized the importance of implementing epidemic prevention training to improving pig farmers’ biosecurity cognition and promoting the adoption of C&D procedures. The results included suggested references for preventing ASF and the next epidemic of animal diseases. Full article