Abstract: To meet the requirements of high accuracy in image edge localization and strong noise resistance for computer vision calibration and precise measurement, an improved Zernike moment subpixel high-precision measurement method for circular hole-like workpieces is proposed. Firstly, the Canny operator is used as a coarse edge detection algorithm, with the traditional Gaussian filter in the Canny operator replaced by an improved Laplacian edge-adaptive median filter. This approach demonstrates improved edge preservation compared to traditional and adaptive median filtering, especially under high-concentration noise. Then, a subpixel edge detection algorithm is applied to refine the edges, thus enhancing the edge localization accuracy. An improved Zernike moment subpixel detection algorithm is employed for precise edge point detection. The improved algorithm selects a Zernike moment parameter template with higher detection accuracy. Finally, the inner and outer diameters of the circular hole-like part are measured by fitting the profile using the least squares method. Experimental results on several different workpieces demonstrate that the proposed algorithm achieves higher accuracy than the traditional Zernike moment subpixel method, with an error reduction of 75.1%, meeting the precision requirements in modern industrial part manufacturing processes.

Abstract: This scoping review examined how external clinical notes are obtained and used before scheduled primary care visits in high-income countries, with attention to continuity of care, workflow integration, value, healthcare delivery, policy, and risk. Within the health data ecosystem, the value of external clinical data depends not only on exchange capability but also on whether information is timely, easy to find, and usable in practice. Guided by Arksey and O’Malley, Joanna Briggs Institute guidance, and PRISMA-ScR, the review searched PubMed, CINAHL, and Google Scholar for English-language, peer-reviewed studies published from 2021 to 2026. Of 330 records identified, 15 studies were included and assigned Johns Hopkins evidence levels and quality ratings. The evidence base was dominated by Level III studies, indicating stronger support for conclusions about workflow barriers, usability, and care coordination than for causal or economic effects. Three patterns emerged: technical exchange alone did not ensure continuity of care, workflow integration shaped whether external information was useful, and the literature described clinical and operational value more clearly than direct financial return. Using the Sittig and Singh sociotechnical model, the review shows that value is produced or lost across infrastructure, clinical content, interface design, people, workflow, organizational conditions, external rules, and monitoring. Overall, external clinical notes function as high-value data only when they are available before the visit, routed appropriately, and usable within the routine primary care workflow. Future research should use stronger workflow-specific measures and assess cost implications and return on investment more consistently.

Abstract: The growing complexity of enterprise resource planning (ERP) systems necessitates intel-ligent approaches for dynamically identifying and evaluating key performance indicators (KPIs) that accurately reflect organizational performance. This paper proposes a mul-ti-agent architecture for dynamic KPI management over Oracle E-Business Suite (EBS). The core design combines a dynamic multi-agent analytics layer, an extendable dedicated EBS KPI Model Context Protocol (MCP) server layer, and a data layer. The dynamic multi-agent analytics layer defines a set of independent large language model (LLM) agents, each re-sponsible for a specific task determined by the business requirements of a particular com-pany. The EBS KPI MCP server layer defines the tools required to access and transform Oracle EBS data and exposes them to the AI agents in the upper layer. Above these layers is the user layer, where the user actively participates in the process through a hu-man-in-the-loop approach. Based on this general architecture, we proposed and imple-mented, as a proof of concept (PoC), a multi-agent system for dynamic business KPI selec-tion, evaluation, and quantification, in which three distinct agents for KPI selection, KPI quantification, and KPI forecasting were instantiated within the multi-agent analytics lay-er. This demonstrates the practical applicability of the proposed general architecture. The study contributes to intelligent business analytics by showing how coordinated LLM agents can automate KPI lifecycle activities within ERP ecosystems, enabling adaptive, data-driven performance management aligned with evolving organizational needs.

Abstract: Modern infocommunication, sensing, and cyber-physical systems increasingly rely on heterogeneous data streams originating from channels of different physical nature, sampling rates, reliability levels, and uncertainty characteristics. Direct fusion of such data in conventional artificial intelligence pipelines often yields decision outputs that are difficult to interpret, calibrate, and trust, especially in safety-related or security-related applications. This work proposes an event-probabilistic approach to the unification of heterogeneous sensor data for decision-support systems. The main idea is to transform heterogeneous sensor observations into a common space of event-oriented probability estimates, which can then be integrated using reliability-aware weighting. In this form, the system can generate not only a final recommendation, but also supporting metrics, including event likelihood, risk level, uncertainty, data quality, and inter-channel conflict. The paper formulates the conceptual and architectural basis of the proposed framework and discusses its compatibility with further Bernoulli encoding and stochastic processing. An illustrative numerical experiment involving four sensor channels and three representative scenarios is used to demonstrate the behavior of the framework. The results show that adaptive reliability-aware weighting improves the stability of the integrated event probability under channel degradation, while explicit conflict assessment prevents unjustified automatic decisions under contradictory sensor evidence. The proposed framework may serve as a basis for future stochastic and photonic-stochastic decision-support systems in access control, industrial monitoring, transport infrastructure, and critical-infrastructure applications.

Abstract: Background/Objectives: Acute pancreatitis (AP) is a heterogeneous disease with outcomes ranging from mild to severe. Early risk stratification is essential, but commonly used scoring systems are often complex for routine use. This study aimed to evaluate the predictive value of systemic inflammatory indices derived from complete blood count parameters in assessing disease severity in AP. Methods: This retrospective study included 454 patients with AP. Demographic, clinical, and laboratory data were obtained from electronic medical records. Systemic inflammatory indices (NLR, PLR, MLR, dNLR, SII, SIRI, and AISI) were calculated from admission laboratory values, with logarithmic transformation applied to selected indices. Disease severity was classified according to the Revised Atlanta Classification, and patients were grouped as mild or moderate-to-severe. Statistical analyses included ROC curve analysis and logistic regression. Results: Among 454 patients, 371 (81.7%) had mild and 83 (18.3%) had moderate-to-severe AP. Patients with more severe disease were older and showed significant differences in several laboratory parameters. NLR, PLR, SII, dNLR, and Log-SII were significantly higher in the moderate-to-severe group. However, all indices demonstrated limited discriminative performance, with dNLR showing the highest AUC (0.612). In multivariate analysis, only age and C-reactive protein (CRP) remained independent predictors of disease severity. Conclusions: Systemic inflammatory indices are associated with disease severity in AP; however, their predictive performance is limited. Conventional parameters such as age and CRP remain more reliable for risk stratification.

Abstract: Piles with continuous helix (referred to herein as "screw pile") is a new configuration of helical piles. It features a continuous helix spiraling several pitches around a smooth shaft forming a "threaded shaft". This study investigates the compressive capacity and behavior of helical and screw piles using 3D numerical models calibrated and validated against full-scale field testing. The bearing capacity factor, Nc, for helical piles is back-calculated from the numerical results and compared against standard theoretical assumptions to evaluate their accuracy in predicting ultimate capacity. Parametric studies are conducted considering screw piles configuration, including shaft diameter, pitch size, helix diameter, as well as soil strength. The results reveal that shaft resistance accounts for up to 89% of the total capacity. Analysis of load distribution, shear contours, and displacement contours at failure allowed for the identification of different failure modes of soil adjacent to the pile’s threaded shaft: Individual Bearing Mode (IBM), Cylindrical Shear Mode (CSM), and a combined mode. The study identifies specific parametric thresholds for these modes in both sand and clay layers. Furthermore, varying clay strength is found to alter the development of the shear surface, transitioning from localized bearing to continuous shearing along the threaded shaft. Finally, apparent shaft resistance factors, α and β, are back-calculated to provide ractical parameters for evaluating the resistance of threaded shafts in layered soil.

Abstract: This study presents an approach to quantifying impact-induced damage severity in composites, focusing on synthetic carbon fibre reinforced polymer (CFRP), natural flax fibre reinforced polymer (FFRP) and hybrid fibres reinforced polymer (HFRP) composite of carbon and flax. The investigation aims to quantitatively characterise impact damage under energies ranging from 10 to 70 J through acousto-ultrasonics (AU) testing, proposing an efficient technique for evaluating the integrity of various FRP composites under in-service conditions. AU testing was performed at azimuthal angles of 0°, 30°, 45°, 60° and 90°, utilising acousto-ultrasonic waveform indices (AUWIs), such as wave velocity, peak amplitude, energy content, centroid frequency and skewness factor. Damage severity index is correlated with the damage mode. The findings establish that wave velocity is a reliable parameter for quantifying damage severity across all composite material types considered, with high adjusted R² values of 0.92 for CFRP, 0.89 for FFRP and 0.90 for HFRP. Peak amplitude also shows considerable sensitivity. Finally, this research highlights the limitations of traditional non-destructive evaluation (NDE) techniques and demonstrates the potential of combining multi-damage metrics with advanced imaging methods, such as X-ray micro-computed tomography (X-ray µCT) and scanning electron microscopy (SEM), to provide comprehensive assessment of damage in various composite materials. The proposed methodology offers a promising approach for quantifying the impact damage severity in composite structures, as applicable to wind turbine blades, amongst other structural components.

Abstract: Schistosomiasis intermediate host snails’ data in the Democratic Republic of the Congo are limited and geographically dispersed. The objective of this study was to characterize snail habitats and identify environmental determinants of their presence. Monthly malacological surveys were conducted at 72 water contact sites. The morphological identification of the snails was complemented by the sequencing of the mitochondrial cox1 gene in order to guarantee confirmation of the species. The physicochemical parameters of the water, as well as human activities on the site, were recorded. The associations between environmental characteristics and snail presence were evaluated using generalized estimating equation models to account for repeated measurements. A total of 172,491 snails were collected, including 4,899 Schistosoma intermediate hosts (Bulinus spp., n = 3,812; Biomphalaria spp., n = 1,087). Biomphalaria pfeifferi, Biomphalaria sudanica, Bulinus truncatus, and Bulinus forskalii were identified. Biomphalaria species were detected in stagnant or slow-flowing waters; however, they occupied distinct habitats. The presence of snails was found to be independently associated with stagnant water and inversely associated with cassava retting, dishwashing/laundry, and river crossing. These findings provide baseline evidence on the distribution and ecological determinants of the Schistosoma intermediate host in Kimpese, supporting targeted malacological surveillance and integrated control strategies.

Abstract: Connexin-43 (CX43) lateralization in ventricular myocardium has been associated with abnormal impulse propagation and increased arrhythmia susceptibility. Its quantitative assessment in histological sections remains challenging because of the difficulty of segmenting individual cardiomyocytes and the reliance of previous methods on geometric rules applied to segmented cell profiles. Here, we present CLARISA, a deep learning framework for classifying CX43-positive regions as either terminal or lateralized directly from fluorescence images, without requiring cardiomyocyte segmentation. An expert-annotated dataset was generated from left-ventricular cryosections of Wistar rat hearts, in which CX43-positive regions were labeled according to their distribution pattern. A dual-stream convolutional classifier based on EfficientNetV2-S was trained to capture both the local and contextual morphology of each region. In addition, an inference module applicable to whole tissue sections was developed to generate spatial lateralization probability maps and global percent lateralization estimates consistent with expert annotation. On the test set, CLARISA achieved a ROC-AUC of 0.905 and a PR-AUC of 0.810. These results support the feasibility of automated assessment of CX43 distribution patterns without explicit cardiomyocyte segmentation. The complete codebase is publicly available, together with access to the pretrained model and the image data used in this study. The Hugging Face model card reports the same held-out test metrics and states that the checkpoint is intended to be used with the main repository.

Abstract: Background: The char regions of Bangladesh — temporary riverine islands — experience compound climate vulnerability intensified by chronic structural poverty, yet sustainable financing models for community-based adaptation remain underdeveloped. Aim: This study diagnoses the capacity–commitment gap between households’ expressed willingness to support climate adaptation and their actual financial capacity, and proposes an evidence-based blended finance instrument. Methods: Using the Contingent Valuation Method (CVM) with a payment-card format and an open-ended follow-up, we surveyed 400 households across two char sites (Bahadurpur and Vasarpara). Probit models estimate the binary decision to contribute; Tobit models estimate the determinants of the contribution amount conditional on willingness. Results: Willingness to pay is high (65% of households), but capacity is sharply constrained: 90% of willing households pledge ≤ 400 BDT/month (mean = 244.5 BDT, median = 220 BDT). Probit and Tobit estimates show that education (β = 1.46, p < 0.001; β = 101.39, p < 0.001) and direct disaster experience (β = 1.49, p < 0.001; β = 153.85, p < 0.001) are three-to-eight times more influential than income (β = 0.49, p < 0.001; β = 19.33, p = 0.034). An institutional-trust paradox emerges: lower trust in government effectiveness is weakly associated with higher stated contributions (Tobit β = −17.88, p = 0.066), consistent with compensatory self-reliance. Near-universal clustering of WTP in the lowest payment class across seven adaptation strategies (89.7–100%) indicates a binding affordability ceiling rather than strategy-specific variation in valuation. Conclusions: These findings invalidate user-pays models for char populations and reframe household WTP as a signal of prioritised demand under a structural affordability ceiling. We translate this diagnostic into the Char Resilience Bond — a blended-finance instrument that securitises formalised in-kind community co-investments (labour, local knowledge, materials) to credit-enhance and leverage external capital, offering a replicable template for financing adaptation public goods in subsistence economies.

Abstract: Background/Objectives: Fibrodysplasia ossificans progressiva (FOP) is an ultra rare genetic disorder causing progressive heterotopic ossification. The dental phenotype has never been systematically characterised. We quantified dental pathologies and oral health related quality of life across three age groups of genetically confirmed FOP patients and compared them with 156 matched healthy controls (2022–2025). Methods: 52 FOP patients (Group I: 1–5 y, n=14; Group II: 6–17 y, n=21; Group III: 18–35 y, n=17) underwent standardised dental examination (Decayed, Missing, and Filled Teeth index (DMFT), Oral Hygiene Index Simplified (OHI S), Angle classification, temporomandibular joint (TMJ) assessment), computed tomography (CT) densitometry, sialometry, salivary crystal analysis, and Oral Health Impact Profile 14 (OHIP 14). Statistical analysis used Kruskal Wallis, Mann Whitney U, Benjamini Hochberg false discovery rate (FDR) correction, and effect sizes. Results: Caries (DMFT≥4) was highly prevalent across all FOP groups (82–86%) and significantly higher than in controls (84.6% vs. 38.5%, p<0.001). Chronic stomatitis increased steeply with age (7.1% in Group I → 100% in Group III, p<0.001); it was universal in FOP adults vs. 6.4% in controls. Enamel hypoplasia (21.4% → 58.8%) and Angle class II malocclusion (0% → 47.1%) also showed large age group differences. Total TMJ disorders were observed in 7.1% of Group I and 100% of Group III (p<0.001); maximal mouth opening was lower by 17.4 mm in Group III (Cohen’s d=2.1). Salivary flow rate was 20% lower in adults (0.35→0.28 ml/min, p=0.01). Calcium phosphate crystals were detected in 17.6% of adults and correlated with CT calcification grade (ρ=0.67, p=0.003). OHIP 14 total score was higher (worse) in Group III (48.9 vs. 12.4 in Group I, Cohen’s d=1.95). Conclusions: This first systematic characterisation of the dental phenotype in FOP shows that chronic stomatitis and TMJ dysfunction become nearly universal by early adulthood, severely impairing quality of life. The correlation between salivary calcium phosphate crystals and CT calcification generates the hypothesis of a non invasive biomarker, requiring prospective validation. The proposed clinical phenotype and minimally invasive recommendations provide a framework for safer dental management of FOP patients.

Abstract: In this work, a numerical investigation of an organic light-emitting diode (OLED) based on a bilayer architecture is presented, with particular emphasis on the influence of ZnO nanoparticles (ZNPs) concentration on charge transport, recombination dynamics, exci-ton formation, and luminescence performance. The studied device consists of a hole injec-tion layer combined with an electron transport and emissive layer based on Alq₃ doped with ZNPs. The impact of ZNPs concentration has been explicitly introduced into carrier mobility, dielectric permittivity, Langevin recombination rate, and radiative exciton decay. The simulation results show that increasing ZNPs concentration enhances charge bal-ance, recombination efficiency, exciton density, and luminescence power. Furthermore, the variation of ZNPs concentration from 0% to 10% in Alq₃ polymer layer increases the electron charge density from 0.65 x 10²¹cm^-3 to 1.4 x 10²¹cm^-3, the recombination rate from 1.25 x10²⁵ cm^-3 s^-1 to 12.5 x10²⁵ cm^-3 s^-1, the exciton density from 0.05 x 10¹⁵cm^-3 to 0.75 x 10¹⁵cm^-3 and the power of luminescence from 0.015W/μm² to 0.75W/μm². Since, the per-formance of Alq3-ZNPs-OLED is tenfold higher than of Alq3-OLED pure. These findings demonstrate that the incorporation of ZNPs is a key parameter for ameliorate and opti-mizing OLED performance which can serve many optoelectronic designs.

Abstract: Background: Perception-based evaluation using Likert-scale survey data is widely applied in tourism and transport research, yet conventional point-valued encoding imposes artificial precision and overlooks ambiguity between adjacent ordinal categories. This limitation is particularly relevant in experiential contexts, where subjective judgments often involve transitional evaluations. Methods: This study develops a parameterized fuzzy–entropy exploratory factor analysis (FE-EFA) framework for uncertainty-aware analysis of ordinal perception data. The approach transforms ordinal responses into fuzzy membership distributions, constructs a correlation structure in membership space, and incorporates Shannon entropy and Jensen–Shannon divergence to characterize distributional dispersion and representation differences. The framework is applied to survey data from Chengdu Tramway Line 2 (N = 1242; 32 indicators). Results: Under the Kaiser criterion (eigenvalues > 1), conventional EFA yields a seven-factor structure, whereas FE-EFA identifies an additional eighth factor located near the retention boundary. Under a unified factor specification, both approaches preserve a consistent high-level structure, while FE-EFA shows clearer factor separation, fewer cross-loadings, and more coherent indicator clustering. From an information-theoretic perspective, FE-EFA produces higher entropy (average = 0.8688) and moderate Jensen–Shannon divergence (average = 0.0133), indicating a controlled redistribution of ordinal information rather than structural distortion. Entropy-informed weighting further reveals systematic shifts in indicator importance across key dimensions. Conclusions: The FE-EFA framework extends conventional Likert-scale analysis by introducing an uncertainty-aware representation layer prior to factor extraction. It preserves overall structural stability while improving the resolution of latent constructs and the sensitivity of indicator representation. The proposed approach provides a practical and theoretically grounded basis for perception-based evaluation and decision support in tramway cultural-tourism development and related contexts.

Abstract: The objective of this research was to develop an intelligent stress mapping and a smart control platform, utilizing Artificial Intelligence (AI), to increase the fatigue life of a hydraulic crane. The crane's boom was modeled and co-simulated using ANSYS, ADAMS, and MATLAB. A flexible model of the boom was created in ANSYS and then exported to ADAMS. Stress analysis was performed using the maximum principal hotspot method and the von Mises yield criterion. Stress optimization was conducted using a Neural Network (NN) algorithm, which is a key implementation of AI in this study. Two control platforms, one based on Neural Networks and another on Fuzzy Logic, were designed to apply AI in controlling the crane's movements. The Neural Network algorithm optimized the crane's movement by adjusting velocity at critical positions where structural stress was high, while the fuzzy logic-based control algorithm utilized stress feedback from the crane's structure. Both AI-driven control algorithms were integrated into the physical crane in the lab, and extensive testing demonstrated a significant increase in the crane's fatigue life, along with effective damping of crane vibrations. This paper introduces a novel AI-driven approach combining Neural Networks and Fuzzy Logic for intelligent stress mapping and control, specifically tailored for hydraulic cranes. Unlike previous works, this research integrates real-time stress feedback into the control process and validates the algorithms through experimental implementation on a prototype crane, significantly improving its fatigue life.

Abstract: This paper was about creating and testing quality, microbial safety, chemical stability, and shelf life of CBD infused bottled water. Regular water does not mix well with lipophilic cannabidiol, which results in dose inconsistency, degradation, microbial contamination, and limited stability. To counteract these problems, a controlled CBD incorporation method was combined with clean, room bottling and systematic quality control protocols. The bottled water was subjected to various tests after being stored for 28 days, including cannabidiol concentration, degradation products, physicochemical parameters (pH, total solids, water activity) and microbial safety, total plate counts, yeast, mold, and pathogenic bacteria. CBD concentration was maintained with negligible degradation and microbial analyses revealed that total counts were low and no pathogens were detected. This proves that aseptic processing is very effective. Physicochemical parameters did not change, which means that the beverage matrix was not affected by either the addition of CBD or the storage. These results guarantee consistent potency, chemical integrity, microbial safety and product stability effectively solving the problem of producing CBD beverages. The paper demonstrates a reliable method of making safe and high, quality CBD functional beverages with a good shelf life. The results are relevant for manufacturing operations of different scales and supply insight on standardized production, quality monitoring, and storage practices. This research is in line with regulatory compliance and consumer safety and consistent product performance, providing a foundation for the safe commercialization of CBD-infused bottled water.

Abstract: Background: The Dietary Inflammatory Index (DII®) is a commonly used tool to assess diet-related inflammation. Higher DII scores are associated with increased cardiovascular disease risk in large observational cohorts yet, controlled-trial evidence evaluating cardiovascular outcomes across DII levels is scarce. This secondary analysis examined cross-sectional differences and longitudinal associations between dietary inflammatory potential and cardiovascular outcomes in healthy Australian adults. Methods: In a double-blind randomised crossover trial, 50 participants consumed 60 mL/day of either high-phenolic (320 mg/kg) or low-phenolic (86 mg/kg) olive oil for two 3-week intervention periods, separated by a 2-week washout. Anthropometry (weight, height, waist circumference, BMI) and cardiovascular outcomes (i.e., blood pressure, lipids, oxidised LDL, and HDL-cholesterol efflux capacity) were measured at four timepoints. DII and energy-adjusted DII (E-DIITM) scores were calculated from 3-day food diaries at baseline and follow-up of each 3-week intervention phase. Linear mixed-effects models compared cardiovascular outcomes across DII tertiles (low, medium, high) adjusting for intervention, period, sequence, age, sex, and waist circumference. Results: Forty-three participants completed the study. At baseline, BMI, waist circumference, systolic blood pressure, total cholesterol, and LDL differed significantly across DII tertiles (p<0.05). Across the study period, cardiovascular outcomes did not differ between medium or high versus low DII tertiles, and no significant time-by-tertile interactions were observed (all p>0.05). DII values remained stable across timepoints, while E-DII decreased modestly within individuals in both intervention periods. Conclusions: In this healthy cohort, DII was not associated with adverse short-term changes in cardiovascular outcomes. Longer-duration studies with greater contrast in dietary inflammatory potential are warranted to clarify the relationship between DII and cardiovascular health.

Abstract: Building upon foundational Item Response Theory (IRT) research conducted at Tecnologico de Monterrey with University of São Paulo (USP), this study presents CONF.i, a framework integrating Canvas LMS with a three-variable IRT model (Grade-Confidence-Performance) and Google's Gemini AI. Using design-based research methodology, an external Google Apps Script application was developed, interfacing with Canvas LTI standards, implementing IRT-based assessment with student confidence ratings and AI-generated personalized feedback and learning resource recommendations. Pilot testing with twenty-three undergraduate students at Tecnologico de Monterrey, Mexico, with theoretical validation from USP collaborators, demonstrated technical feasibility and pedagogical value. Results revealed that 82% of students rated the interface positively, 87% understood the confidence rating mechanism, and 91% would recommend the approach. The three-variable model revealed four learning patterns within the pilot sample that would be invisible to traditional scoring: aligned mastery (34.8%), underconfident competence (21.7%), overconfident struggle (26.1%), and aligned struggle (17.4%). These observed patterns suggest potential for enabling targeted instructional interventions, warranting further investigation with larger samples. This Brazil-Mexico collaboration demonstrates that sophisticated educational technologies can be integrated within existing institutional infrastructure without commercial licensing costs, contributing to Sustainable Development Goal #4 (Quality Education) by making adaptive learning technologies more accessible through mainstream platforms.

Abstract: Selection bias in Large Language Models has emerged as a fundamental obstacle to reliability, fairness, and robustness. Defined operationally as systematic decision changes under equivalence-preserving input perturbations, including option permutation, label renaming, candidate-order swapping, and evidence relocation, the phenomenon is examined across four representative task families: multiple-choice question answering, in-context classification, LLM-as-a-Judge evaluation, and long-context or retrieval-augmented generation. Selection bias is first analyzed through a causal chain that links biased behavior to training-data priors, architectural asymmetries, and post-training amplification. Existing mitigation methods are then synthesized through an intervention-level taxonomy spanning inference-time calibration and prompt optimization, architecture-level modification, and training-level debiasing. The evaluation landscape is unified by summarizing commonly used metrics, benchmark families, and application settings, with the lack of standardized and cross-task-comparable protocols identified as a central bottleneck. Selection bias is best understood as a failure of invariance under non-semantic reformatting, and mitigating it is essential for trustworthy, robust, and selection-invariant language models.

Abstract: The paper introduces a fundamental shift in the representation of physical reality, moving from a particle-based paradigm to a Recursive Complex Representation of 5 scaling levels (RCR) of a “hypophenomenal” geometric model. A unique scaling base ξ is defined that is deduced from the Planck constant and the gravitational constant G. The model posits that space is not a static container but a Plenum (David Bohm’s name for vacuum) of Planck contiguous cells (P-cells) whose vibrations constitute the fundamental energy of the universe. Masses are “trapped light”, viewed as localized vibrational resonances of the signal c that maintain a portion of the signal within contiguous groups of P-cells, and can propagate along the plenum by keeping their internal vibrational configuration. The fine-structure constant α acts within a universal renormalizing factor, strictly related to the scaling factor ξ = (GL_P)^1/5. Experimental masses, across 60 orders of magnitude (from the neutrino to the Sun) are retrieved from their ξ logarithmic localization with respect to Planck’s mass. The fundamental equations of Planck, Einstein, and de Broglie are not independent postulates, but natural geometric emergences of the signal’s vibrational dynamics. Gravitational force and constant G are formulated in terms of matter aggregation and dynamic curvature of the signal in the Plenum.

Abstract: Background/Objectives: Obstructive sleep apnea (OSA) is a common and serious sleep-related disorder that causes repeated interruptions in breathing during sleep. Traditional diagnostic methods, such as polysomnography, are accurate but costly, time-consuming, and unsuitable for large-scale screening. This study proposes and evaluates a lightweight diagnostic framework based on an Extreme Learning Machine (ELM) optimized by a set of basic and advanced metaheuristic optimizers (GA, RUN, MEO, CL-PSO, HI-WOA, GWO, HGS, HHO, SeaHO, MGO, and the hybrid GWO--WOA). The model aims to improve early detection of OSA using demographic and clinical data. Methods: Two real datasets were employed to train and evaluate the proposed framework: (i) a clinical OSA dataset with 274 subjects and 31 demographic/anthropometric and sleep-related predictors, and (ii) a public strongly imbalanced Sleep-Disordered Breathing (SDB) dataset with 500 subjects and 10 structured predictors. Metaheuristic algorithms are used to optimize ELM weights and biases, addressing the instability of random initialization and improving model generalization. The optimized models are evaluated against eight baseline classifiers, including Logistic Regression (LR), k-nearest neighbours (KNN), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Multilayer Perceptron (MLP), XGBoost (XGB), and a standard ELM classifier. Results: Results show that metaheuristic optimization improves ELM on the OSA dataset, increasing ROC-AUC from 0.6527 to about 0.73 and accuracy from 0.6573 to about 0.69–0.70, while on the highly imbalanced SDB dataset, it yields modest ROC-AUC gains (from 0.5132 to about 0.544–0.548) with small decreases in accuracy and F1-score. Conclusions: The proposed framework provides a fast, lightweight, and cost-effective screening tool for large-scale, resource-limited healthcare settings, enabling early OSA detection and preventive intervention.