Search Results (4,326)

This study presents a comprehensive review of more than 100 research papers on sign language recognition (SLR) published between 2020 and 2026. The analysis focuses on deep learning approaches applied to video-based SLR, including spatiotemporal feature extraction, temporal modeling, attention mechanisms, motion-based representations, hybrid frameworks, transfer learning methods and other methods. Particular attention is given to how these methods model spatiotemporal dynamics and capture subtle gesture characteristics in sign language communication. The review highlights several recent developments, such as the introduction of specialized datasets, the emergence of real-time recognition systems, and the integration of multimodal fusion strategies. At the same time, persistent challenges remain, including data scarcity in low-resource sign languages, limited linguistic standardization of datasets, and insufficient model interpretability. The findings underline the importance of developing scalable and generalizable models capable of handling diverse datasets and user variability. The distinct contributions of this review are fourfold: (1) a comprehensive synthesis of over 100 studies published between 2020 and 2026, covering the full spectrum of deep learning architectures for video-based SLR; (2) a structured six-category taxonomy enabling systematic cross-architectural comparison; (3) a comprehensive focus on low-resource sign languages, which remain underrepresented in the existing literature; and (4) a critical analysis of the current benchmark landscape for low-resource sign languages, identifying key gaps and outlining strategic directions for future dataset development. These contributions are intended to guide further research toward more robust, inclusive, and universally applicable SLR systems. Full article

Handwritten signatures retain their legal validity and play a central role in the forensic examination of documents. While it is recognized that signatures can be influenced by contextual and human factors, research on the systematic evaluation of signature stability within a structured and quantitative framework under prolonged professional shift conditions remains limited. The aim of this study is to identify the temporal pattern of signature variability over a 24-h shift and to systematically quantify this variability using a structured ordinal rating system. This prospective observational study included 25 emergency department resident physicians working on a 24-h shift system. Handwriting samples were collected from the participants at the 0th, 8th, 16th, and 24th h of their shift. The signatures were evaluated using the Signature Change Scale, which assigns an ordinal score on a 0–5 scale, with each participant’s baseline (0-h) signature serving as the reference. The evaluations were conducted under double-blind conditions by two independent experts, each with over 10 years of experience in forensic document examination. In order to assess reliability, the inter- and intra-observer reliability were calculated. Inter-observer reliability was assessed by comparing the scores assigned by two independent raters for the same signatures. Intra-observer reliability was determined by re-evaluating the same anonymized dataset under the same conditions. Statistical analysis was performed using descriptive statistics, Wilcoxon signed-rank test, and intraclass correlation coefficient (ICC) analysis for reliability assessment. It was observed that signature deviation scores varied across time points. At the 8-h mark, mild deviation was detected in 84% of participants. By the 16th h, the frequency of higher deviation categories had increased; by the 24th h, mild deviation was observed in 60% of participants, mild-to-moderate in 32%, and moderate in 8%. It was noted that average deviation scores increased up to the 16th h, with no further significant increase observed between the 16th and 24th h. The inter- and intra-observer reliability ranged from 0.89 to 0.97 and 0.85 to 0.93, respectively. These ICC values indicate that inter- and intra-observer reliability was at a good to excellent level. A 24-h shift duration was observed to be associated with intra-individual variability in signature characteristics under controlled professional conditions. High inter- and intra-observer reliability observed under blind and repeated evaluations suggests good scoring consistency within the controlled conditions of the present study. The findings suggest that signature characteristics may exhibit context-sensitive variability influenced by individual and environmental factors. The Signature Change Scale should be regarded as an exploratory methodological framework only, and further studies with larger samples and comprehensive content, construct, criterion, and external validity analyses are required before any forensic or practical application. Full article

Background: Radio Electric Asymmetric Conveyer (REAC) neurobiological modulation is proposed as an approach designed to interact with endogenous bioelectrical processes involved in cortical regulation. However, its electrophysiological correlates in physiologically preserved neural systems remain insufficiently characterized. The present study explored whether a standardized REAC Brain Wave Optimization Gamma (BWO-G_B) protocol is associated with measurable changes in resting-state EEG activity in healthy adults. Methods: Nine neurologically healthy participants completed a standardized REAC BWO-G_B protocol consisting of 18 sessions administered over six consecutive days. Resting-state EEG recordings were obtained before and after the intervention. Spectral power was analyzed across the 1–100 Hz range. Multivariate organization of cortical activity was explored using Principal Component Analysis (PCA) and Canonical Discriminant Analysis (CDA), with CDA used only as a descriptive visualization of within-dataset multivariate organization. Cross-correlation analysis was applied to evaluate changes in inter-regional temporal synchronization. Individual-level non-parametric testing (Wilcoxon signed-rank test) was conducted only to characterize within-subject directional spectral modulation across the recorded montage. Results: Post-intervention EEG recordings showed a consistent redistribution of spectral power across cortical regions, predominantly within frequencies below approximately 20 Hz. This pattern was observed across subjects at the individual level. Multivariate analysis revealed a dissociation between PCA, which showed partial overlap between conditions, and CDA, which descriptively showed within-dataset separability between baseline and post-intervention cortical states. Cross-correlation analysis indicated a spatially differentiated redistribution of temporal synchronization across cortical regions. At the individual level, descriptive Wilcoxon analyses indicated broadband spectral differences in seven of nine participants (p < 0.05), with consistent directional trends across all subjects; these p-values should not be interpreted as confirmatory statistical evidence. Conclusions: The findings indicate the presence of a reproducible electrophysiological pattern observed after completion of the REAC BWO-G_B protocol in healthy adults. The observed combination of spectral redistribution, descriptive multivariate organization, and changes in temporal synchronization is consistent with a structured post-intervention modification of cortical activity organization within the present dataset. However, given the exploratory design, small sample size, absence of a control condition, and absence of objective vigilance monitoring, these results should be interpreted cautiously and should not be considered as evidence of intervention-specific effects. Further controlled studies are required to determine specificity, underlying mechanisms, and potential functional relevance. Full article

Artificial intelligence (AI) systems increasingly depend on complex, multi-stage supply chains that incorporate pre-trained models, third-party datasets, open-source libraries, and automated training pipelines. This dependency creates a rapidly expanding attack surface in which model poisoning, dependency compromise, and provenance manipulation can undermine system integrity long before deployment. Existing AI governance frameworks—including the NIST AI Risk Management Framework and NIST’s Secure Software Development Framework—acknowledge supply chain risks but do not define a verifiable model provenance structure or cryptographically durable integrity guarantees. Simultaneously, the transition to post-quantum cryptography (PQC) introduces new requirements for long-lived AI artifacts: classical digital signatures used to verify model lineage, dataset integrity, and pipeline attestation will become vulnerable to quantum-enabled forgery within the expected operational lifetime of many AI systems. This paper synthesizes evidence from policy, standards, and benchmark sources to characterize the emerging AI supply chain threat landscape and identify cryptographic dependencies that the PQC transition disrupts. We propose a formal Model Bill of Materials with PQC-safe extensions (MBOM-PQC), a unified signing and attestation pipeline integrating ML-DSA and hybrid signature modes, and a five-level Supply Chain Assurance Maturity Model (SCAMM) supporting repeatable organizational evaluation. Together, these contributions aim to provide a structured foundation for AI supply chain integrity, supporting verifiable model lineage, authenticity, and trustworthiness through the PQC transition and beyond. The framework is presented as a design-science contribution comprising three integrated artifacts and is extended with operational guidance for continuous-learning pipelines (§6.5), a formal scoring methodology for organizational assessment (§7.3.5), and a hardware-root-of-trust migration cost matrix (§8.3.6). Full article

In-flight propulsion system diagnosis in multirotor unmanned aerial vehicles (UAVs) remains a challenging problem due to closed-loop control interactions, strong environmental disturbances, and common-mode effects that obscure rotor-specific anomalies. Conventional passive monitoring approaches based solely on electrical or mechanical measurements are often insufficient for reliable fault localization and for distinguishing global degradations from nominal operation. This paper proposes an active diagnostic framework that exploits low-amplitude sinusoidal excitation injected into the control null space during hover operation. By employing lock-in detection, rotor responses are selectively extracted at the excitation frequency, enabling the derivation of robust amplitude-based sensitivity indicators from rotational speed, current, and electrical power signals. A pairwise signed diagnostic metric is formulated to achieve reliable localization of asymmetric rotor faults. In addition, an absolute indicator referenced to a baseline condition is introduced to capture symmetric degradations affecting all rotors through the combined use of current- and power-based sensitivities. The proposed method is validated in a high-fidelity quadrotor simulation environment incorporating viscous-friction and thrust-coefficient degradation faults. Extensive Monte Carlo analyses demonstrate robust fault-detection and localization performance, including scenarios that are indistinguishable using conventional pairwise normalization techniques. Full article

Background/Objectives: Childhood interstitial lung disease (chILD) represents a heterogeneous group of rare pulmonary disorders. Practical diagnostic approaches tested for feasibility and impact in comprehensive cohorts are lacking. We aimed to assess a simple etiologically focused classification approach, clarify the role of genetic testing and quantify the impact of non-pulmonary organ manifestations. Methods: We hypothesized that chILD can be classified in a clinically meaningful and versatile way by answering three questions: Which children have an etiological chILD diagnosis due to (1) identified (exposure-related) cause/lung injury, or (2) systemic disease? (3) In how many children without an etiological diagnosis can a genetic cause be identified? We also calculated the predictive value of non-pulmonary organ involvement for underlying systemic conditions. Results: Among 1693 patients, 24.7% were grouped as ILD related to exposure, 22.7% as ILD with systemic condition, 16.6% as ILD with genetic diagnosis of systemic disease, 10.0% as ILD with genetic diagnosis affecting the lungs only, and 25.8% as ILD without genetic diagnosis. The average genetic diagnostic yield was 50.8%, with higher rates in interstitial pneumonia (61.4%) or pulmonary alveolar proteinosis (87.1%). The presence of ≥two non-pulmonary organ manifestations increased the likelihood of an underlying systemic disease by three to five-fold. Conclusions: An etiological diagnostic strategy effectively classifies chILD and guides genetic testing. Exome or genome sequencing should be considered if ≥two non-pulmonary organs are involved or if the initial diagnosis becomes uncertain due to an unusual disease course or signs of a second underlying condition. Full article

Acute ischemic stroke (AIS) is associated with redox imbalance; however, the early temporal changes in enzymatic and non-enzymatic antioxidant responses are poorly understood. The aim of this study was to evaluate changes in selected oxidative stress markers during the early phase of AIS. The study was designed as a longitudinal within-subject analysis, with each patient serving as their own reference between day 1 and day 8. A total of 48 patients (mean age 69.31 ± 1.59 years; 56.3% male; mean body mass index (BMI) 27.05 ± 0.61 kg/m²), predominantly presenting with mild to moderate stroke severity, were enrolled in a prospective observational study. The cohort was characterized by a high prevalence of hypertension (87.5%), dyslipidemia (45.8%), and diabetes or prediabetes (45.9%). Blood samples were collected on day 1 and day 8 after stroke onset. Depending on the distribution of paired differences, either the paired Student’s t-test or Wilcoxon signed-rank test was applied. A significant increase in superoxide dismutase (SOD) activity was observed (1932.73 vs. 2086.55 U/g Hb, p = 0.032), whereas catalase (CAT; 403.19 vs. 415.30 × 10³ U/g Hb, p = 0.444) and glutathione peroxidase (GPx; 24.70 vs. 24.40 U/g Hb, p = 0.477) showed no significant changes. Similarly, malondialdehyde (MDA) levels remained stable in both erythrocytes (182.96 vs. 187.15 nmol/g Hb, p = 0.838) and plasma (0.41 vs. 0.41 nmol/mL, p = 0.922). In contrast, melatonin (59.65 vs. 55.49 pg/mL, p = 0.042) and 25-hydroxyvitamin D (25(OH)D; 19.31 vs. 16.52 ng/mL, p < 0.001) concentrations significantly decreased. These findings suggest that the early phase of AIS may be associated with a selective and potentially maladaptive antioxidant response, involving increased SOD activity alongside depletion of systemic modulators, which may contribute to persistent redox imbalance. Full article

Juvenile idiopathic arthritis (JIA) is the most common systemic chronic inflammatory connective tissue disease in children, characterized by joint inflammation lasting at least six months. Temporomandibular joint (TMJ) involvement can occur in conjunction with other joints and may often be asymptomatic in its early stages. Objective: This study aims to evaluate the relationship between clinical symptoms of the stomatognathic system and radiologically confirmed cone beam computed tomography (CBCT)-detected structural TMJ changes in children with JIA. The research hypothesis posits that specific clinical symptoms are more prevalent in patients with CBCT-confirmed structural TMJ changes. Methods: A cohort of children diagnosed with JIA was examined. Clinical symptoms, including facial asymmetry, limited mandibular movement, and joint and masticatory muscle pain upon palpation, were assessed. CBCT imaging was performed to assess osseous TMJ structural changes. Results: The frequency of orofacial clinical symptoms was assessed and compared between patients with and without radiological evidence of TMJ involvement. Children with CBCT-confirmed TMJ changes demonstrated significantly higher rates of facial asymmetry, reduced maximum mouth opening, mandibular deviation during opening, and limitations in lateral or protrusive movements compared with those without TMJ involvement. Pain-related symptoms (TMJ pain, muscle tenderness, and pain during movement) and joint sounds occurred at similar frequencies in both groups. Conclusions: Facial asymmetry, mandibular deviation during opening and reduced mandibular mobility are the clinical signs most strongly associated with structural TMJ involvement in JIA and should prompt targeted imaging. Pain-related symptoms show limited diagnostic value, highlighting the need for focused clinical assessment and future studies integrating CBCT and MRI to refine early screening protocols. Full article

As unmanned aerial vehicles (UAVs) are increasingly deployed in mission-critical domains such as military operations, infrastructure inspection, and disaster response, the threat of GPS and network spoofing attacks has emerged as a fundamental challenge to operational continuity. Existing intrusion detection systems based on threshold rules or shallow machine learning models are inherently limited in their ability to identify the latent onset of sophisticated, gradually intensifying spoofing campaigns—a gap that motivates the present work. This study proposes a deep learning-based early detection and network resilience prediction framework that employs Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) architectures operating on three spatio-temporal network features—Hop Count Spike Rate (HCS), Packet Drop Volatility (PDV), and Spatial Disconnect Density (SDD)—proposed in this study. To reflect realistic adversarial conditions, we design a Gradual Adaptive Attacker model in which the spoofing intensity escalates progressively across six operational phases, including a second-stage adaptive attack that modulates its gradient upon detecting initial countermeasures. Both models are trained on 1000 simulated runs using sliding-window time-series sequences and evaluated across 500 independent test runs with paired statistical testing. The GRU model achieves a mean ROC-AUC of 0.9915 (±0.0091) and a mean F1-Score of 0.9099 (±0.0462), outperforming LSTM across all metrics with statistical significance at p < 0.001 under both the paired t-test and the Wilcoxon signed-rank test. Critically, GRU detects spoofing onset with an average latency of 0.503 time steps—approximately 29.4% faster than LSTM (0.712 steps)—a difference confirmed as statistically significant (p < 0.001, Cohen’s d = 0.41). Network resilience simulations further demonstrate that integrating GRU-based autonomous evasion maintains a Connectivity Ratio (CR) above 80% even under severe attack phases, whereas passive networks experience total connectivity collapse (CR = 0%). These findings establish GRU as the superior architecture for real-time UAV edge deployment and affirm that the proposed pipeline extends beyond threat alerting to actively preserving mission continuity under adversarial spoofing conditions. Full article

Predicting student performance has emerged as an essential element of contemporary learning assessment, allowing educational organizations to determine problematic students and offer early intellectual assistance. Many machine learning (ML) methodologies prioritize predicted accuracy at the expense of interpretability and practical insights. This paper provides a framework for predicting student performance that is both risk aware and explainable utilizing a chaotic educational competition optimizer (ECO) in conjunction with a support vector classifier (SVC) to overcome existing challenges. The ECO serves as a metaheuristic feature selection technique for selecting the most significant features from a multivariate educational dataset consisting of 1195 students and 29 behavioral, demographic, and academic characteristics. Experimental findings demonstrate that ECO effectively condenses the feature space to 11 essential indications and improves generalization of model while maintaining classification robustness. Utilizing the chosen features, the ECO–SVC model attains a complete classification accuracy of 87.03%, with F1-scores of 0.92, 0.69, and 0.82 for high-, medium-, and low-performance student categories, respectively, surpassing other benchmark ML methods. The proposed framework incorporates explainable artificial intelligence (XAI) to improve transparency by utilizing local explanations and permutation-driven feature significance. The XAI research verifies that institutional support, learner engagement, and previous academic success are the most important contributing factors to predictive results. Notably the ECO functions as a classifier-independent feature selection mechanism; however, the support vector classifier (SVC) is adopted in this study due to its strong generalization capability and effectiveness in exploiting the optimized feature space. The findings are analyzed using a semiotic-linguistic framework, wherein certain qualities are correlated with symbolic, indexical, and temporal educational signs, converting numerical significance into substantive pedagogical insights. Furthermore, an initial academic risk profile strategy is established by utilizing SVC decision confidence and elucidating feature contributors. The consequent risk ratings accurately categorize students into low-, medium-, and high-risk categories, facilitating the detection of at-risk learners beyond mere final score assessment. The proposed risk-aware and explainable ECO–SVC framework enhances learning outcomes assessment by integrating interpretability, high accuracy, and proactive academic reasoning, rendering it suitable for real-life educational decision-support systems. Full article

This study presents a unified semantic segmentation framework for UAV-based multispectral crop stress mapping, focusing on the integration of water stress and rust disease conditions within a common label space. Unlike conventional approaches that address individual stress factors independently, the proposed framework harmonizes heterogeneous datasets with different annotation schemes into a single multi-class segmentation problem. To achieve this, UAV multispectral orthomosaics are processed using a patch-based strategy and a multi-head UNet++ architecture incorporating segmentation, edge-aware, and Signed Distance Transform (SDT) branches. In addition, a physics-informed output-space refinement module based on fractional partial differential equations (FPDE) is introduced to enhance spatial coherence and boundary preservation in the predicted maps. Experimental results demonstrate the effectiveness of the proposed framework within the evaluated dataset setting, particularly in terms of boundary delineation, spatial consistency, and minority-class detection. The study highlights the feasibility of integrating heterogeneous stress conditions into a unified segmentation framework and provides a foundation for future research on scalable multi-source agricultural monitoring systems. Full article

Retrieval-Augmented Generation (RAG) systems depend critically on the quality of document preprocessing, yet no prior study has evaluated PDF processing frameworks by their impact on downstream question-answering accuracy. We address this gap through a systematic comparison of four open-source PDF-to-Markdown conversion frameworks, Docling, MinerU, Marker, and DeepSeek OCR, across 21 pipeline configurations, varying the conversion tool, cleaning transformations, splitting strategy, and metadata enrichment. Evaluation was performed using a 50-question benchmark over a corpus of 36 Portuguese administrative documents (1706 pages, ~492K words), with LLM-as-judge scoring over 50 independent runs per configuration. Statistical significance was assessed via Wilcoxon signed-rank tests with Cohen’s d effect sizes. Two baselines bounded the results: naïve PDFLoader (86.2%) and manually curated Markdown (91.3%). Docling with hierarchical splitting and image descriptions achieved the highest automated accuracy (94.1 ± 1.6%), surpassing even manual curation. A per-question-type analysis revealed that table-dependent questions drive the largest accuracy differences, with a 33-percentage-point gap between basic and hierarchical splitting. Metadata enrichment and hierarchy-aware chunking contributed more to accuracy than the conversion framework alone. An exploratory GraphRAG implementation underperformed basic RAG (82% vs. 94.1%). These findings demonstrate that data preparation quality is the dominant factor in RAG system performance. Full article

Background/Objectives: Erythroderma is a rare but potentially life-threatening dermatological emergency characterised by generalised erythema and scaling involving more than 80% of the total body surface area. Erythroderma is associated with significant morbidity and mortality due to systemic complications and diverse underlying aetiologies. Methods: In this narrative review, PubMed was searched up to February 2026. Studies were screened for relevance to emergency physicians, with emphasis on epidemiology, diagnostic approach, and acute management. Non-English publications and conference abstracts were excluded. A total of 122 sources were included in the final synthesis. Results: Erythroderma most commonly results from exacerbation of pre-existing inflammatory dermatoses, drug reactions, infections, or cutaneous T-cell lymphoma. Clinical presentation includes diffuse erythema and scaling affecting ≥80–90% of body surface area, often accompanied by pruritus, systemic symptoms, and signs of organ dysfunction. Systemic complications arise from cutaneous barrier failure and include fluid imbalance, thermoregulatory dysfunction, cardiovascular strain, protein loss, and secondary infection. Initial emergency department management prioritises supportive care, fluid and nutritional optimisation, restoration of skin barrier function, and assessment for organ dysfunction. While a definitive aetiological diagnosis is not always immediately required, certain conditions—particularly severe drug reactions and infectious causes such as staphylococcal scalded skin syndrome—necessitate urgent targeted intervention. Conclusions: Erythroderma represents a syndromic emergency requiring systematic evaluation and early supportive management. Prompt recognition of high-risk aetiologies and timely dermatology referral are essential to optimise outcomes and reduce morbidity and mortality. Full article

Objectives. This study evaluated the performance of three intraoral scanners with different acquisition technologies in detecting early signs of tooth wear, using micro-computed tomography (micro-CT) as the reference standard. Methods and Materials. Three IOS were examined, including an active triangulation scanner, a structured-light triangulation scanner, and a parallel confocal technology scanner. Ten extracted unrestored and caries-free premolars were placed in the maxillary left second premolar position of a dental mannequin and scanned at baseline, generating quadrant digital models. Micro-CT scans were also obtained at baseline. Wear was simulated by immersion in a 1% citric acid solution followed by brushing of the buccal surfaces. All specimens were rescanned with IOS and micro-CT. Micro-CT datasets were reconstructed into stereolithography models and compared with IOS models using mesh analysis software. Statistical analysis was performed in R using linear mixed-effects models to account for repeated measurements across teeth. RMS values and absolute errors relative to the micro-CT reference were analysed with device as a fixed effect and tooth as a random effect, with Tukey-adjusted pairwise comparisons. Repeatability was additionally assessed from the repeated scans using within-tooth variability. Results. Significant differences were observed among the evaluated systems in the detection of changes related to tooth wear (p < 0.001). The micro-CT reference showed the lowest RMS value, followed by Trios 3, Primescan, and Omnicam. Model-based analyses confirmed significant differences among the evaluated systems, while the magnitude and statistical support of pairwise contrasts depended on the specific outcome considered. Repeatability analysis showed that Trios 3 had the lowest within-tooth standard deviation and repeatability coefficient (0.0215 mm and 0.0595 mm, respectively), followed by Primescan (0.0290 mm and 0.0802 mm), whereas Omnicam showed the highest within-tooth variability and repeatability coefficient (0.0624 mm and 0.173 mm). Conclusions. The parallel confocal and structured-light triangulation intraoral scanners produced RMS values numerically closer to the micro-CT reference than the active triangulation scanner. However, none of the evaluated intraoral scanners demonstrated quantitative agreement sufficient to be considered interchangeable with the reference standard. Full article

The growth of Internet of Things (IoT) environments has expanded the attack surface of modern systems. Trojan attacks are a major challenge as they evade conventional detection mechanisms and operate silently within legitimate processes. This paper presents an automated Trojan detection framework for Windows-based IoT gateways. The framework combines custom dataset generation informative feature engineering and deep learning-driven analysis. A dataset of 3000 real world executable samples was created through controlled sandbox execution and forensic monitoring. The process captured behavioral static and network-level characteristics. An initial set contained 146 extracted features. A multi-stage feature selection process identified 33 informative attributes. This step allowed efficient learning and preserved discriminative power. A custom deep neural network model named TrDNN was developed using these features. The model captures complex nonlinear patterns linked to Trojan activity. The framework was evaluated against five classical machine learning models. It was also compared with five deep learning baselines. Results show that TrDNN achieves strong detection performance. The accuracy is 0.975. The precision is 0.972. The recall is 0.969. The F1 score is 0.970. The study also examines inference time and energy consumption. The model shows a balance between detection effectiveness, computational cost and energy efficiency. This makes it suitable for resource-constrained IoT gateway deployment. The detection model was integrated into an automated real-time monitoring pipeline. The system enables continuous process surveillance through Windows command line automation with minimal operational overhead. Statistical validation used paired t tests, Wilcoxon signed rank tests and McNemar chi-square test. The performance gains are statistically significant and do not indicate overfitting. The framework provides a reliable, efficient and deployable solution for Trojan detection in modern IoT systems. Full article