Browse Articles

Personalized learning path recommendation for middle school students based on joint optimization of deep reinforcement learning and knowledge tracing

Abstract The rapid advancement of personalized education has created an urgent need for intelligent learning path recommendation systems that can dynamically adapt to students’ evolving knowledge states. However, existing learning path recommendation systems predominantly rely on static knowledge graphs or simple heuristic rules, failing to capture the temporal dynamics of student knowledge acquisition and the complex multi-objective nature of learning optimization. This paper proposes Path-Mentor, a novel framework that integrates a Temporal-Aware Graph Knowledge Tracing Network (TA-GKTN) with a Multi-Objective Curriculum Planner and Executor (MOCPE) through a Counterfactual Causal Inference-based Co-Training (CCI-CT) mechanism. The TA-GKTN module models students’ dynamic knowledge states as graph-structured representations by incorporating temporal convolution and gating mechanisms into heterogeneous graph neural networks. The MOCPE module employs a hierarchical multi-agent reinforcement learning framework to decouple high-level knowledge point sequencing from low-level learning activity execution, enabling systematic balancing of conflicting objectives including knowledge consolidation, novelty exploration, and cognitive load management. The CCI-CT method establishes an end-to-end joint optimization pipeline by generating counterfactual training data and enabling bidirectional knowledge transfer between the knowledge tracing and path recommendation modules. Comprehensive experiments conducted on four benchmark datasets (ASSISTments 2012-2013, Junyi Academy, EdNet, and Eedi) demonstrate that Path-Mentor significantly outperforms baseline models across six evaluation metrics, achieving a 12.4% improvement in AUC-ROC compare with DKT model for knowledge state prediction, 23.7% reduction in learning path efficiency, and 18.6% enhancement in long-term mastery improvement. These results validate the effectiveness of the proposed joint optimization framework in advancing personalized learning path recommendation for middle school students.

Willingness toward kidney donation among patients’ relatives at Muhimbili National Hospital, Dar es Salaam, Tanzania: A cross-sectional study

Background Kidney transplantation provides superior long-term survival and quality of life over dialysis for patients with end-stage kidney disease; however, its use is limited by the availability of donors. In Tanzania, only living-related kidney transplantation has been performed since 2017 at Muhimbili National Hospital (MNH). Success relies on donor readiness, yet little is known about public willingness to donate. Aim To assess willingness of patients’ relatives to donate kidneys at MNH. Methods Cross-sectional study design among 424 in-patient relatives at MNH from May to June 2023. Systematic random sampling was used to recruit participants. Data were collected using a questionnaire comprised of inquiries on socio-demographics, knowledge, attitudes, and willingness to donate and analyzed using the Stata 18 software. Frequency distribution tables summarized descriptive statistics. Modified Poisson regression with robust variance identified factors associated with willingness. Results Of the 424 participants, Mean age 36 ± 11 years; 240(56.6%) female, 362(85.4%) urban, 400(94.3%) educated. While 361(85%) heard of organ donation, only 32(7.5%) had adequate knowledge, 289(68.2%) positive attitude, 200(47.2%) willing to donate. Age 35–44 years (aRR = 0.69 [95% CI: 0.49–0.99], p = 0.046), female gender (aRR = 0.82 [95% CI: 0.67–0.99], p = 0.042), informal traders/farmers (aRR = 0.77 [95% CI: 0.60–0.99], p = 0.041) had lower willingness versus counterparts. Positive attitude showed 74% higher likelihood (aRR = 1.74 [95% CI: 1.31–2.30], p < 0.001). Conclusion Low kidney donation willingness was influenced by attitude, age, gender, and occupation. Majority of the participants lacked adequate knowledge. Educational programs needed to improve knowledge, attitude, and willingness.

Low resource word sense disambiguation in Oromo with fine tuned small transformers

Abstract A key task in natural language processing is word sense disambiguation (WSD), which attempts to determine the accurate meaning of ambiguous words based on their context. While transformer-based designs have achieved significant results in high-resource languages, WSD for low-resource languages such as Oromo remains hard due to inadequate annotated corpora and lexical resources. Contextual representation learning has been greatly enhanced by recent advancements in transformer-based language models, allowing for more reliable disambiguation in situations with limited input. This study uses a manually created dataset from the Oromo–English Dictionary to examine the efficacy of transformer-based models for lexical-sample WSD in Oromo. The dataset contains sentences annotated by two native speakers, attaining an inter-annotator agreement of 0.82, indicating good annotation reliability. The dataset was filtered for experimental usage following preprocessing, normalization, and elimination of noisy cases. 472 training sentences, 71 validation sentences (15%), and 140 test sentences made up the final dataset. The dataset has a highly unbalanced long-tail distribution and encompasses 43 sense classes. BERT-base-cased gets the best performance with an accuracy of 0.862 and a macro-F1 score of 0.2897, according to an experimental evaluation of transformer-based models, including BERT, RoBERTa, DistilBERT, Davlan/afro-xlmr-base, and multilingual variations. Significant differences between models with χ 2  = 34.03 and p  = 4.0 × 10 −1 are confirmed by statistical analysis using the Friedman test. BERT-base-cased performs much better than most transformer variations and classical baselines, according to post-hoc Wilcoxon signed-rank tests. These results show that contextual transformer representations are quite successful for low-resource WSD, although there is still a significant class imbalance that limits performance.

Characteristics, management, and outcomes of segmental and subsegmental pulmonary embolism in ICU patients: A retrospective cohort study

Objective To describe the incidence, management, and outcomes of segmental and subsegmental pulmonary embolism (PE) in intensive care unit (ICU) patients and to explore associations between therapeutic-dose anticoagulation and clinical outcomes. Design Single-center retrospective cohort study. Setting Tertiary academic hospital ICU between January 2019 and June 2025. Patients Critically ill adults (≥18 years) who underwent computed tomography pulmonary angiography (CTPA) during ICU admission and had radiologically confirmed segmental or subsegmental PE. Interventions None. Measurements and main results Radiology reports of all CTPA examinations performed in ICU-admitted patients were screened to identify the most proximal level of thrombus. Clinical records were reviewed for demographics, illness severity, radiologic characteristics, anticoagulation practice, bleeding, venous thromboembolism (VTE) recurrence, and mortality. Among 896 CTPA examinations performed in 804 patients, 164 examinations (18.3%) identified PE. Of these, 115 scans (12.8% of all CTPAs) demonstrated distal PE only, corresponding to 104 patients (12.9%) (61 segmental, 43 subsegmental). Overall, 96% of patients with distal PE received anticoagulation and 86% of anticoagulated patients received therapeutic-dose regimens. Bleeding occurred in 15% (major bleeding 12%), 90-day VTE recurrence in 7.8%, and 90-day mortality in 24%. No statistically significant association was found between the use of therapeutic-dose anticoagulation and 90-day mortality (adjusted odds ratio [OR], 0.70; 95% CI, 0.21–2.45), bleeding episodes (adjusted OR, 2.34; 95% CI, 0.47–19.2), or VTE recurrence (adjusted OR, 0.69; 95% CI, 0.11–6.22). Conclusions In critically ill adults, segmental and subsegmental PE are commonly detected on CTPA and are usually treated with therapeutic-dose anticoagulation. Although VTE recurrence was less frequent than bleeding episodes and mortality, our study did not find a significant association between therapeutic-dose anticoagulation and bleeding episodes, recurrent VTE, or mortality. Larger prospective studies are needed to define optimal anticoagulation strategies for ICU patients with distal PE.

Serum NO₃⁻ and NO₂⁻ levels among pregnant women from agricultural communities: associations with self-reported dietary and environmental exposures

Abstract Nitrate (NO₃⁻) and nitrite (NO₂⁻) exposure can lead to adverse health impacts, yet data on direct biomarkers are limited. To assess serum NO₃⁻ and NO₂⁻ levels among pregnant women in two agricultural regions of the Jordan Valley in relation to dietary and environmental factors. This cross-sectional study recruited 346 pregnant women aged 18 years or older from four hospitals in the Ghor region between 2023 and 2024. Serum NO₃⁻ and NO₂⁻ levels were measured using a Griess assay. Socio-demographic data, dietary habits, and water source information were gathered through questionnaires. The median serum NO₂⁻ level was 0.63 µM and the median serum NO₃⁻ level was 19.13 µM in the overall sample. Serum NO₂⁻ was significantly higher among women from North Ghor compared to South Ghor (0.83 µM vs. 0.45 µM, p  < 0.001), while serum NO₃⁻ levels did not significantly differ between regions. In multivariable models adjusting for maternal age, BMI, hypertension, anemia, and smoking, we found that variables of region of residency and eating green beans once a week or more were significantly associated with higher NO₂⁻ levels in the whole sample ( p  < 0.001 and p  = 0.031, respectively) and in North Ghor ( p  = 0.032), while hypertension was associated with lower NO₂⁻ levels ( p  = 0.032 whole sample; p  = 0.013 North Ghor). For NO₃⁻, smoking during pregnancy was associated with lower levels in the whole sample ( p  = 0.003) and in South Ghor ( p  = 0.004), while employment in crop harvesting/collection was associated with higher NO₃⁻ levels in South Ghor ( p  = 0.024). Regional differences in serum NO₂⁻ levels exist among pregnant women in rural Jordan, influenced primarily by dietary habits, while occupational exposure to crop harvesting and smoking were associated with serum NO₃⁻ levels. Targeted recommendations and public health strategies are needed to mitigate NO₃⁻ and NO₂⁻ exposure.

Prevalence, virulence profiles and antibiotic susceptibility patterns of Shiga toxin producing Escherichia coli O157:H7 among children 6–59 months in Longido, Arusha-Tanzania

Shiga toxin producing Escherichia coli (STEC) is a zoonotic pathogen associated with diarhoeal disease and severe complications in children, yet its epidemiology in pastoral settings of Tanzania remains insufficiently characterized. This study determined the prevalence, virulence gene profiles and antibiotic susceptibility patterns of STEC among children aged 6‍-59 months with diarhoea in Longido District, northern Tanzania. A hospital based cross-sectional‍ study was conducted between July and August 2025, enrolling 150 participants from four health facilities. Stool samples were collected and analyzed using culture, serological conformation and multiplex polymerase chain reaction targeting five genes; rfbE , stx1 , stx2 , eae A  and hly A . STEC was operationally defined by detection of stx1 and/or stx2 . Antibiotic susceptibility was assessed using the Kirby-Bauer disk diffusion method. The prevalence of STEC was‍ 13.3% (20/‍150). All stx2 positive isolates co-occurred with stx1 . Virulence genes showed a heterogeneous but significantly clustered distribution, with rfbE (20.0%) and stx1 (13.3%) predominating. Significant co-occurrence was observed‍ between stx1 and eae and between stx1 and hlyA (p < .001). Animal contact, raw milk consumption and use of untreated water were significantly associated with STEC infection. Firth penalized logistic regression confirmed these exposures as independent predictors. Antibiotic susceptibility profiles were uniform, with complete susceptibility to ciprofloxacin,‍ gentamicin, cefotaxime and ceftazidime while ampicillin and trimethoprim showed complete resistance. These findings indicate that STEC transmission in pastoral communities is strongly driven by zoonotic and environmental exposures, characterized by clustered virulence determinants and consistent antibiotic profiles. Limitations include the cross-sectional design and short sampling period, which may not capture seasonal variation. Strengthened surveillance and integrated One Health interventions are needed to reduce disease burden.

Comparative analysis of artificial intelligence models for predicting oil viscosity in the in-situ catalytic oil upgrading process: a study of influential parameters

Intelligent kinematic physics engine construction for hyper-redundant cable-driven flexible manipulator

To solve the problems of difficult kinematics modeling and poor real-time performance for a hyper-redundant cable-driven flexible manipulator, a construction method of an intelligent kinematics physics engine based on Bayesian-optimized Long Short-Term Memory (BO-LSTM) network is proposed in this paper. Firstly, the multilevel kinematics equation of the flexible manipulator is established based on the Denavit-Hartenberg (D-H) parameter method, and an analytical model of the forward solution is established. Secondly, the inverse kinematic model of the Newton iteration method is proposed based on the kinematics equation, and the stability of the numerical solution method is proven. Subsequently, the kinematics database is established based on the Monte Carlo method, the analytical model of forward solution and the high-precision numerical model of inverse solution, and the database is visualized. Finally, an intelligent kinematics physics engine is established based on the kinematics database and the BO-LSTM neural network model. This model combines the global search ability of Bayesian optimization and the advantages of LSTM neural network in processing time series data, which provides a new solution for the kinematic modeling of hyper-redundant cable-driven flexible manipulator. The calculation speed of the forward and inverse BO-LSTM models are 0.018 s and 0.017 s, which are faster than the numerical method. The MSEs of the forward and inverse BO-LSTM models are 0.065 and 0.056 respectively, which are lower than those of the BP, RBF and LSTM neural networks. The SimMechanics module in MATLAB is used to simulate the hyper-redundant cable-driven flexible manipulator. The experimental result shows that the intelligent kinematics physics engine based on the BO-LSTM neural network model has high precision and high efficiency in kinematics solution.

Data-driven exploration of electronic nose technology to differentiate bacteria in blood cultures under biofilm-promoting conditions

Abstract Biofilms are a major cause of delayed wound healing, yet current biofilm identification methods are limited by invasiveness, processing times, or specificity. This study investigates the potential of metal-oxide electronic noses for identifying bacterial cultures of Staphylococcus aureus , Pseudomonas aeruginosa , Enterococcus faecium , and Staphylococcus epidermidis grown in blood-based growth medium. We conducted in-vitro experiments to capture volatilome signatures from cultures grown under biofilm-promoting conditions and analyzed data using an interpretable machine learning workflow to disentangle algorithmic limitations from biological variability. This workflow incorporated feature extraction and selection, correlation-based clustering, and Shapley analysis. Six classification models were evaluated using cross-validation. Considering all five classes, classification accuracy reached at most 55.6%, which Shapley-based interpretation attributed mainly to biological factors: E. faecium and S. aureus exhibited high signal similarity to control samples and strong inter-day variability. Accuracy increased to 100.0% for species with distinct volatile signatures, and dimensionality reduction resulted in a model using two constructed features. These findings demonstrate that classification performance in biological sensing cannot be explained solely by algorithmic factors. Interpretable machine learning workflows help for distinguishing biological sources of complexity from algorithmic ones. We provide a proof-of-principle for electronic nose-based identification of bacterial growth under biofilm-promoting conditions.

The relationship between esports and cognitive function: A scoping review

The aim of this scoping review was to map and synthesize empirical evidence on the relationship between esports participation and cognitive function, with particular attention to executive control, attentional processes, visuospatial working memory, decision-making, and lifestyle-related factors. Searches were conducted in PubMed and Web of Science following PRISMA-ScR guidance. Nine empirical studies met the inclusion criteria and were included in the final synthesis. The included evidence consisted mainly of cross-sectional comparisons, together with a small number of intervention, acute experimental, and qualitative studies. The studies were organized into three thematic domains: cognitive differences across expertise levels, lifestyle and intervention effects on cognitive performance, and decision-making or training mechanisms. Five studies primarily examined expertise-level differences, three addressed lifestyle, health load, or intervention-related factors, and one focused on decision-making and training processes. Overall, the available evidence suggests that competitive esports players may show advantages in selected cognitive domains, particularly executive control, attentional flexibility, and visuospatial processing. However, findings were heterogeneous across game genres, participant classifications, cognitive tasks, and reporting practices. Because only nine studies were eligible and most designs were cross-sectional, the current evidence base remains insufficient for strong causal conclusions about whether esports participation improves cognitive function or whether pre-existing cognitive abilities contribute to esports expertise. Future research should use longitudinal, experimental, and cross-cultural designs with standardized cognitive measures and transparent reporting of statistical results.

Fabrication and characterization of ecofriendly Taro starch bioplastic composite films reinforced with coffee husk and enset fiber for packaging applications

Translating injury prevention evidence into safer padel: Protocol of a TRIPP-guided scoping review

Padel is a rapidly growing sport with high injury incidence rates and substantial consequences. However, no comprehensive overview exists of the evidence on padel-related injuries and injury prevention strategies. A scoping review can help identify gaps in the emerging field of padel research. The objective of this scoping review is to systematically map the available evidence on padel‑related injuries and injury prevention strategies across all player populations and settings, from epidemiological description to intervention effectiveness. Studies describing injuries or injury risk factors or mechanisms in padel players of any kind and regarding primary injury prevention strategies or interventions involving stakeholders at any socio-ecological level will be considered. Studies that have been conducted in any geographic location, at any playing level, in any setting and in any implementation context will be considered. The proposed scoping review will be conducted in accordance with the JBI methodology for scoping reviews and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) will be followed. A three-step search strategy will be utilized. The search strategy will be adapted for PubMed, SPORTDiscus, Web of Science, Scopus, EMBASE and CINAHL. A structured gray‑literature search will also be conducted. The data extracted from included papers will include specific details about the participants, concept, context, study methods and key findings relevant to the review questions. No critical appraisal of individual sources of evidence will be performed. The extracted data will be presented according to the six stages of the Translating Research into Injury Prevention Practice (TRIPP) framework. The review has been registered through Open Science Framework (osf.io/4t69j).

Novelty stressor induces differential brain oxidative stress and antioxidant profiles between proactive and reactive stress coping styles

Edge-intelligent safelink-V2X: A low-latency cooperative framework for real-time vulnerable road user protection

The protection of Vulnerable Road Users (VRUs) remains a major challenge in modern transportation safety, as onboard line-of-sight and adverse weather conditions limit conventional onboard sensors. Existing systems that rely solely on vehicle-based sensing or on isolated communication struggle to provide timely, accurate alerts in dynamic urban environments. To address these shortcomings, this paper introduces SafeLink-V2X, a comprehensive Vehicle-to-Everything Cooperative Warning Framework designed to enhance safety for pedestrians, cyclists, and scooter riders. SafeLink-V2X employs Cellular Vehicle-to-Everything (C-V2X) and Dedicated Short-Range Communications (DSRC) protocols to enable direct data exchange of location, velocity, and heading between connected vehicles, smart infrastructure, and VRUs via smartphones or wearable tags. By applying sensor fusion and machine learning–based conflict prediction, the system identifies potential collision points and issues real-time, context-aware warnings through vehicle HMIs and VRU devices, promoting immediate evasive action. Evaluation on urban intersection simulations (detailed in Section 5) demonstrates that SafeLink-V2X reduces simulated collision probability by up to 91.4%, increases situational awareness measures by 44%, and lowers end-to-end alert latency by 30% compared to baseline onboard-only and communication-only systems under the same conditions.

Influencing factors of functional exercise adherence in stroke survivors: a cross-sectional study based on structural equation modeling

A late fusion multi-task learning for respiratory waveform and rate estimation from photoplethysmography

Continuous respiratory monitoring enables early detection of physiological deterioration, yet conventional capnography remains impractical for prolonged use. Photoplethysmography (PPG) offers a non-invasive alternative that encodes respiratory information through baseline wander (respiratory-induced intensity variation; RIIV), amplitude modulation (respiratory-induced amplitude variation; RIAV), and frequency modulation (respiratory-induced frequency variation; RIFV) of the pulsatile waveform. Existing PPG-based deep learning approaches, whether operating on the raw signal or on these physiological modulations, are limited to single-task architectures that estimate either respiratory rate or reconstruct the respiratory waveform in isolation, without jointly addressing both outputs. We propose a late fusion multi-task framework in which dedicated encoder branches independently process each modulation before fusion, and dual decoders simultaneously reconstruct the respiratory waveform and estimate the respiratory rate. The framework was evaluated on the CapnoBase ( n  = 42) and BIDMC ( n  = 52) benchmarks across multiple training strategies. For respiratory-rate estimation, the best transfer-learning configurations achieved a mean absolute error (MAE) of 2.27 bpm on CapnoBase and 1.33 bpm on BIDMC. For waveform reconstruction, the corresponding MAE values were 19.00% and 20.90%, with moderate correlations ( r  = 0.662 and r  = 0.591, respectively). Sequential transfer learning consistently outperformed all other strategies, whereas pooled training degraded both outputs, demonstrating that capnography-derived and impedance-derived waveforms are not interchangeable training targets. These findings establish that short-window PPG can simultaneously support respiratory-rate estimation and waveform reconstruction, when reference signal compatibility is explicitly addressed in multi-task training.

Controlled blasting technology for high in-situ stress and biased pressure tunnels: research and application

The impact of transparency and imitation over complex networks in strategic classification

Classification algorithms are widely used in critical domains such as healthcare, bank loans, credit and fraud detection. These systems should be transparent, yet it remains unclear how individuals will use explanations to adjust their own features. Individuals often access multiple sources of information, from insights provided by institutions to experiences shared among peers. Based on the information received, individuals may decide to strategically adapt to obtain a favourable outcome, honestly improving or attempting to game the system. This paper studies the impact of transparency and social information on strategic classification. We assume that agents adapt based on best response and behavioural imitation along the edges of social networks. We observe that increasingly opaque decision rules can negatively impact the utility of institutions, especially in dense social networks. The number of False Positives is reduced in networks with a lower average degree, when users imitate the average behaviour, as opposed to the most extreme behaviours. When imitating the most extreme behaviour among their connections, users change their features to a large extent in networks with a higher average degree (i.e., higher density). This applies to both honest improving and gaming, with more pronounced impacts in the case of the latter, creating an additional source of risk for institutions. Our model and results reveal that behavioural imitation patterns and social network effects influence the downstream effects of algorithmic transparency.

A robust binary secretary bird optimization method for high-dimensional data classification

Abstract Feature selection is a key step in machine learning–based decision systems, especially in medical and biomedical applications, where datasets often contain a large number of features that can negatively affect both accuracy and interpretability. In this study, we introduce the binary secretary bird optimization algorithm (B-SBOA), a binary version of the secretary bird optimization algorithm specifically developed for feature selection tasks. The proposed approach translates the hunting and escape behaviors of the secretary bird into effective binary search strategies, allowing a well-balanced trade−off between exploration and exploitation. B-SBOA was tested on twenty-five benchmark datasets from the UCI repository and compared with nine well-known binary metaheuristic algorithms, including PSO, GWO, MPA, HBO, SMA, SFOA, DOA, SCA, and MSO. The experimental results show that B-SBOA consistently delivers performance that is either superior to or competitive with existing methods across F-score, precision, recall, and other standard metrics. B-SBOA was evaluated on 25 benchmark datasets from the UCI repository and compared with nine well-known binary metaheuristic algorithms. The results show that B-SBOA achieves superior performance, with average improvements reaching 3–8% in F-score and 2–6% in precision and recall compared to competing methods. In several high-dimensional datasets such as Arrhythmia and Hillvalley, the proposed method achieved the highest classification accuracy while reducing the number of selected features.

Identification of a HOMA-IR cut-off point for cardiometabolic risk and modifiable risk factors in peruvian adolescents

Background Although HOMA-IR is widely used to assess insulin resistance, reported cut-off values vary substantially across population, particularly during adolescence. The aim of this study was to determine the distribution of HOMA-IR values. identify a HOMA-IR cut-off associated with metabolic syndrome (MS), and assess modifiable risk factors of IR in a longitudinal cohort of Peruvian adolescents. Methods We performed a secondary data analysis from a longitudinal adolescent’s study. A sample of 371 adolescents (14.5 ± 0.1 years old) from low- medium socioeconomic status. ROC curve analysis was used to identify the specific cut-off point to classify IR using the sensitivity and specificity values in comparison with the MS. Multiple logistic regression analysis including diet, physical activity and body composition from adolescence, excess weight during infancy and family history of non-chronic disease was included to identify risk factors (FHCD) associated with IR. Results The HOMA-IR was 3.29 (SD 1.71) with no differences by sex. We identified 3.9 for HOMA-IR as the cut-off point with sensitivity (72.4%) and specificity (75.4%) for predicting MS. IR was present in 28.6% (95% CI 24.2;33.4%); 84% had at least one cardiometabolic risk factor and low HDL and abdominal obesity were the most prevalent (62 and 35%, respectively). Adolescents with higher fat mass index (OR 16.03, 95% CI 6.79 to 37.86), and those physically inactive (OR 2.08 95% CI 1.06 to 4.07) were more likely to have IR. No association was found with diet, excess weight at infancy and FHCD. Conclusions A cut-offs point of 3.9 for HOMA-IR allows to identify adolescents with high metabolic risk. Strategies to promote lower FMI and improve the physical activity levels could reduce the risk of IR in adolescents.