Search Results (6,584)

Background: Alzheimer’s disease (AD) is a progressive neurodegenerative condition marked by cognitive decline and memory loss. Despite advancements in AI-driven neuroimaging analysis for AD detection, clinical deployment remains limited due to challenges in model interpretability and usability. Explainable AI (XAI) frameworks such as XRAI offer potential to bridge this gap by providing clinically meaningful visualizations of model decision-making. Methods: This study developed a comprehensive, clinically deployable AI system for AD severity classification using 2D brain MRI data. Three deep learning architectures MobileNet-V3 Large, EfficientNet-B4, and ResNet-50 were trained on an augmented Kaggle dataset (33,984 images across four AD severity classes). The models were evaluated on both augmented and original datasets, with integrated XRAI explainability providing region-based attribution maps. A web-based clinical interface was built using Gradio to deliver real-time predictions and visual explanations. Results: MobileNet-V3 achieved the highest accuracy (99.18% on the augmented test set; 99.47% on the original dataset), while using the fewest parameters (4.2 M), confirming its efficiency and suitability for clinical use. XRAI visualizations aligned with known neuroanatomical patterns of AD progression, enhancing clinical interpretability. The web interface delivered sub-20 s inference with high classification confidence across all AD severity levels, successfully supporting real-world diagnostic workflows. Conclusions: This research presents the first systematic integration of XRAI into AD severity classification using MRI and deep learning. The MobileNet-V3-based system offers high accuracy, computational efficiency, and interpretability through a user-friendly clinical interface. These contributions demonstrate a practical pathway toward real-world adoption of explainable AI for early and accurate Alzheimer’s disease detection. Full article

This study proposes a Personalized Deep Learning-Based Sleep Recommendation System Using Lifelog Data (PDSRS-LD). Traditional sleep research primarily relies on bio signals such as EEG and ECG recorded during sleep but often fails to sufficiently reflect the influence of daily activities on sleep quality. To address this limitation, we collect lifelog data such as stress levels, fatigue, and sleep satisfaction via wearable devices and use them to construct individual user profiles. Subsequently, real sleep data obtained from an AI-powered motion bed are incorporated for secondary training to enhance recommendation performance. PDSRS-LD considers comprehensive user data, including gender, age, and physical activity, to analyze the relationships among sleep quality, stress, and fatigue. Based on this analysis, the system provides personalized sleep improvement strategies. Experimental results demonstrate that the proposed system outperforms existing models in terms of F1 score and Average Precision (mAP). These results suggest that PDSRS-LD is effective for real-time, user-centric sleep management and holds significant potential for integration into future smart healthcare systems. Full article

Hydrogen-assisted fatigue cracking presents a critical challenge to the structural integrity of legacy carbon steel natural gas pipelines being repurposed for hydrogen transport, posing a major barrier to the deployment of hydrogen infrastructure. This study systematically evaluates the fatigue crack growth (FCG) behavior of API 5L X60 pipeline steel under varying hydrogen–natural gas (H₂–NG) blending conditions to assess its suitability for long-term hydrogen service. Experiments are conducted using a custom-designed autoclave to replicate field-relevant environmental conditions. Gas mixtures range from 0% to 100% hydrogen by volume, with tests performed at a constant pressure of 6.9 MPa and a temperature of 25 °C. A fixed loading frequency of 8.8 Hz and load ratio (R) of 0.60 ± 0.1 are applied to simulate operational fatigue loading. The test matrix is designed to capture FCG behavior across a broad range of stress intensity factor values (ΔK), spanning from near-threshold to moderate levels consistent with real-world pipeline pressure fluctuations. The results demonstrate a clear correlation between increasing hydrogen concentration and elevated FCG rates. Notably, at 100% hydrogen, API X60 specimens exhibit crack propagation rates up to two orders of magnitude higher than those in 0% hydrogen (natural gas) conditions, particularly within the Paris regime. In the lower threshold region (ΔK ≈ 10 MPa·√m), the FCG rate (da/dN) increased nonlinearly with hydrogen concentration, indicating early crack activation and reduced crack initiation resistance. In the upper Paris regime (ΔK ≈ 20 MPa·√m), da/dNs remained significantly elevated but exhibited signs of saturation, suggesting a potential limiting effect of hydrogen concentration on crack propagation kinetics. Fatigue life declined substantially with hydrogen addition, decreasing by ~33% at 50% H₂ and more than 55% in pure hydrogen. To complement the experimental investigation and enable predictive capability, a modular machine learning (ML) framework was developed and validated. The framework integrates sequential models for predicting hydrogen-induced reduction of area (RA), fracture toughness (FT), and FCG rate (da/dN), using CatBoost regression algorithms. This approach allows upstream degradation effects to be propagated through nested model layers, enhancing predictive accuracy. The ML models accurately captured nonlinear trends in fatigue behavior across varying hydrogen concentrations and environmental conditions, offering a transferable tool for integrity assessment of hydrogen-compatible pipeline steels. These findings confirm that even low-to-moderate hydrogen blends significantly reduce fatigue resistance, underscoring the importance of data-driven approaches in guiding material selection and infrastructure retrofitting for future hydrogen energy systems. Full article

Background: As the United States population becomes increasingly diverse, the representation of minorities in health professions is critical to addressing health disparities. Few investigations have been conducted among students enrolled in the first semester of the nursing program, a vulnerable and adjustment period for most nursing majors. Thus, this study examined the association between reading comprehension and science aptitude on student retention and standardized test scores. Method: A cross-sectional repeated measures study was conducted to investigate the outcomes from a compendium of programmatic interventions implemented among n = 80 nursing students enrolled in the first semester of a pre-licensure baccalaureate nursing program in one Hispanic-serving institution. These interventions included the Weaver™ reading online program, case studies, NCLEX-type practice tests, test-taking skills, and peer-mentoring. Data collection was conducted in Spring 2024. Multivariate statistical analysis was used to determine predictors associated with student retention and standardized test scores. An independent t-test was used to examine any significant difference in the reading comprehension level among the cohort’s participants. A qualitative investigation using thematic analysis was conducted to understand students’ experiences with the programmatic interventions. Results: Students’ baseline reding comprehension level was significantly associated with failure in the first semester of the nursing program (β = −0.815; SE = 0.349; Wald = 5.444; p < 0.05). End-of-term reading comprehension level was significantly associated with end-of-course HESI score in the Foundations in Nursing course (β = 26.768; SE = 10.049; Beta = 0.445; p < 0.05) while science GPA was significantly associated with end-of-course HESI score for Health Assessment (β = 3.022; SE = 1.315; Beta = 0.434; p < 0.05. Cohort retention was 75%. The independent t-test result indicated a significant difference in reading level was found between those who dropped out from the cohort (M = 4.23, SE = 0.173 and those who did not (M = 5.15, SE = 0.188), t (68) = −3.037, p < 0.01. A reading level of grade 10 and above was associated with student progression to the next semester (M = 10.16, SE = 0.375, t (70) = −0.560, p < 0.05. Although the participants found the reading comprehension modules tedious, test-taking strategies, applying the nursing process in case studies, and the expertise of a nurse educator, who understood the learning needs of first-semester students, were perceived as critical to academic success. Conclusions: Reading comprehension and science aptitude are essential to students’ early success in the nursing program. Addressing gaps in reading comprehension and science aptitude before admission to a nursing program would increase chances of success in the early stages of a nursing major. Full article

This paper introduces Contextual Object Grouping (COG), a specific computer vision framework that enables automatic interpretation of technical security diagrams through dynamic legend learning for intelligent sensing applications. Unlike traditional object detection approaches that rely on post-processing heuristics to establish relationships between the detected elements, COG embeds contextual understanding directly into the detection process by treating spatially and functionally related objects as unified semantic entities. We demonstrate this approach in the context of Cyber-Physical Security Systems (CPPS) assessment, where the same symbol may represent different security devices across different designers and projects. Our proof-of-concept implementation using YOLOv8 achieves robust detection of legend components (mAP50 ≈ 0.99, mAP50–95 ≈ 0.81) and successfully establishes symbol–label relationships for automated security asset identification. The framework introduces a new ontological class—the contextual COG class that bridges atomic object detection and semantic interpretation, enabling intelligent sensing systems to perceive context rather than infer it through post-processing reasoning. This proof-of-concept appears to validate the COG hypothesis and suggests new research directions for structured visual understanding in smart sensing environments, with applications potentially extending to building automation and cyber-physical security assessment. Full article

The rapid evolution of 5G New Radio networks has introduced a wide range of services with diverse requirements, complicating their coexistence within the shared radio spectrum and posing challenges in traffic scheduling and resource allocation. This study aims to analyze and categorize the methods, approaches, and techniques proposed to ensure efficient joint and dynamic packet scheduling and resource allocation among heterogeneous services—namely eMBB, URLLC, and mMTC—in 5G and beyond, with a focus on Quality of Service and user satisfaction. This scoping review draws from publications indexed in IEEE Xplore and Scopus and synthesizes the most relevant evidence related to packet scheduling across heterogeneous services, highlighting key approaches, core performance metrics, and emerging trends. Following the PRISMA-ScR methodology, 48 out of an initial 140 articles were included for explicitly addressing coexistence, scheduling, and resource allocation. The findings reveal a research emphasis on eMBB and URLLC coexistence, while integration with mMTC remains underexplored. Moreover, the evidence suggests that hybrid and deep learning-based approaches are particularly promising for tackling coexistence and resource management challenges in future mobile networks. Full article

Frequency-Modulated Continuous-Wave (FMCW) Laser Detection and Ranging (LiDAR) systems are widely used due to their high accuracy and resolution. Nevertheless, conventional distance extraction methods often lack robustness in noisy and complex environments. To address this limitation, we propose a deep learning-based signal extraction framework that integrates a Dual Convolutional Neural Network (DCNN) with a Transformer model. The DCNN extracts multi-scale spatial features through multi-layer and pointwise convolutions, while the Transformer employs a self-attention mechanism to capture global temporal dependencies of the beat-frequency signals. The proposed DCNN–Transformer network is evaluated through beat-frequency signal inversion experiments across distances ranging from 3 m to 40 m. The experimental results show that the method achieves a mean absolute error (MAE) of 4.1 mm and a root-mean-square error (RMSE) of 3.08 mm. These results demonstrate that the proposed approach provides stable and accurate predictions, with strong generalization ability and robustness for FMCW LiDAR systems. Full article

Three-dimensional (3D) ocean physical fields are essential for understanding ocean dynamics, but reconstructing them solely from sea-surface remote sensing remains challenging. We present 3DV-Unet, an end-to-end deep learning framework that reconstructs eddy-resolving three-dimensional essential ocean variables (temperature, salinity, and currents) from multi-source satellite data. The model employs a 3D Vision Transformer bottleneck to capture cross-depth and cross-variable dependencies, ensuring physically consistent reconstruction. Trained on 2011–2019 reanalysis and satellite data, 3DV-Unet achieves RMSEs of ~0.30 °C for temperature, 0.11 psu for salinity, and 0.05 m/s for currents, with all R² values above 0.93. Error analyses further indicate higher reconstruction errors in dynamically complex regions such as the Kuroshio Extension, while spectral analysis indicates good agreement at 100 km+ but systematic deviation in the 20–100 km band. Independent validation against 6113 Argo profiles confirms its ability to reproduce realistic vertical thermohaline structures. Moreover, the reconstructed 3D fields capture mesoscale eddy structures and their life cycle, offering a valuable basis for investigating ocean circulation, energy transport, and regional variability. These results demonstrate the potential of end-to-end volumetric deep learning for advancing high-resolution 3D ocean reconstruction and supporting physical oceanography and climate studies. Full article

Accurate assessment of coalbed methane (CBM) content is essential for characterizing subsurface reservoir distribution, guiding well placement, and estimating reserves. Current methods for determining coal seam gas content mainly rely on direct laboratory measurements of core samples or indirect interpretations derived from well log data. However, conventional coring is costly, while log-based approaches often depend on linear empirical formulas and are restricted to near-wellbore regions. In practice, the relationships between elastic properties and gas content are highly complex and nonlinear, leading conventional linear models to produce substantial prediction errors and inadequate performance. This study introduces a novel method for predicting gas content in deep coal seams using a Conditional Generative Adversarial Network (CGAN). First, elastic parameters are obtained through pre-stack inversion. Next, sensitivity analysis and attribute optimization are applied to identify elastic attributes that are most sensitive to gas content. A CGAN is then employed to learn the nonlinear mapping between multiple fluid-sensitive seismic attributes and gas content distribution. By integrating multiple constraints to refine the discriminator and guide generator training, the model achieves accurate gas content prediction directly from seismic data. Applied to a real dataset from a CBM block in the Ordos Basin, China, the proposed CGAN-based method produces predictions that align closely with measured gas content trends at well locations. Validation at blind wells shows an average prediction error of 1.6 m³/t, with 83% of samples exhibiting errors less than 3 m³/t. This research presents an effective and innovative deep learning approach for predicting coalbed methane content. Full article

Automated Road defect detection using Unmanned Aerial Vehicles (UAVs) has emerged as an efficient and safe solution for large-scale infrastructure inspection. However, object detection in aerial imagery poses unique challenges, including the prevalence of extremely small targets, complex backgrounds, and significant scale variations. Mainstream deep learning-based detection models often struggle with these issues, exhibiting limitations in detecting small cracks, high computational demands, and insufficient generalization ability for UAV perspectives. To address these challenges, this paper proposes a novel comprehensive network, RoadNet, specifically designed for high-precision road defect detection in UAV-captured imagery. RoadNet innovatively integrates Transformer modules with a convolutional neural network backbone and detection head. This design not only significantly enhances the global feature modeling capability crucial for understanding complex aerial contexts but also maintains the computational efficiency necessary for potential real-time applications. The model was trained and evaluated on a self-collected UAV road defect dataset (UAV-RDD). In comparative experiments, RoadNet achieved an outstanding mAP@0.5 score of 0.9128 while maintaining a fast-processing speed of 210.01 ms per image, outperforming other state-of-the-art models. The experimental results demonstrate that RoadNet possesses superior detection performance for road defects in complex aerial scenarios captured by drones. Full article

Ensuring secure and reliable communication in Vehicular Ad hoc Networks (VANETs) is critical for safe transportation systems. This paper presents Vehicular Misbehavior Network (VeMisNet), a deep learning framework for detecting misbehaving vehicles, with primary contributions in systematic feature engineering and scalability analysis. VeMisNet introduces domain-informed spatiotemporal features—including DSRC neighborhood density, inter-message timing patterns, and communication frequency analysis—derived from the publicly available VeReMi Extension Dataset. The framework evaluates Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional LSTM architectures across dataset scales from 100 K to 2 M samples, encompassing all 20 attack categories. To address severe class imbalance (59.6% legitimate vehicles), VeMisNet applies SMOTE post train–test split, preventing data leakage while enabling balanced evaluation. Bidirectional LSTM with engineered features achieves 99.81% accuracy and F1-score on 500 K samples, with remarkable scalability maintaining >99.5% accuracy at 2 M samples. Critical metrics include 0.19% missed attack rates, under 0.05% false alarms, and 41.76 ms inference latency. The study acknowledges important limitations, including reliance on simulated data, single-split evaluation, and potential adversarial vulnerability. Domain-informed feature engineering provides 27.5% relative improvement over dimensionality reduction and 22-fold better scalability than basic features. These results establish new VANET misbehavior detection benchmarks while providing honest assessment of deployment readiness and research constraints. Full article

Open-vocabulary object detection in remote sensing aims to detect novel categories not seen during training, which is crucial for practical aerial image analysis applications. While some approaches accomplish this task through large-scale data construction, such methods incur substantial annotation and computational costs. In contrast, we focus on efficient utilization of limited datasets. However, existing methods such as CastDet struggle with inefficient data utilization and class imbalance issues in pseudo-label generation for novel categories. We propose an enhanced open-vocabulary detection framework that addresses these limitations through two key innovations. First, we introduce a selective masking strategy that enables direct utilization of partially annotated images by masking base category regions in teacher model inputs. This approach eliminates the need for strict data separation and significantly improves data efficiency. Second, we develop a dynamic frequency-based class weighting that automatically adjusts category weights based on real-time pseudo-label statistics to mitigate class imbalance issues. Our approach integrates these components into a student–teacher learning framework with RemoteCLIP for novel category classification. Comprehensive experiments demonstrate significant improvements on both datasets: on VisDroneZSD, we achieve 42.7% overall mAP and 41.4% harmonic mean, substantially outperforming existing methods. On DIOR dataset, our method achieves 63.7% overall mAP with 49.5% harmonic mean. Our framework achieves more balanced performance between base and novel categories, providing a practical and data-efficient solution for open-vocabulary aerial object detection. Full article

The Technology-Enhanced Learning (TEL) Marketplace was a joint initiative by the vice rectorate for academic affairs and the vice rectorate for digitization and change management at Graz University of Technology to modernize lectures. As part of this initiative, an exercise course on chemical thermodynamics was redesigned as a learner-centered course and enriched with interactive learning materials designed to promote self-directed learning. The core of the method used to implement this redesign is interactive notebooks created in Wolfram Mathematica to enable students to work through the examples independently, in depth, and irrespective of time, with the required theoretical background integrated into the notebooks. In this paper, we ask the following questions: RQ1: How did students use and accept the interactive notebooks? RQ2: What was the impact of the interactive notebooks and the corresponding course design as perceived by the students? To answer these questions, we conducted a questionnaire-based survey with 45 course students and statistically analyzed the results. Key results for RQ1 show that $93.33\%$ of the participating students reported using the interactive notebooks, and technology acceptance (1 = low TA, 5 = high TA) was high in both the dimensions of perceived usefulness ($m=3.88$) and attitude ($m=4.24$). Regarding RQ2, our key results show that students perceived the notebooks to have a positive impact on their learning experience, especially regarding their self-directed learning. The results of this work are in alignment with observations by lecturers, which showed that this more student-centric course design and the integration of the interactive learning materials made it possible to clarify detailed questions during the independent learning phase, allowing the interactive part of the course to focus on the tactical approaches, solutions, and problems that arose during the calculations, which raised the overall level of content teaching. Full article

Student exam pass prediction (EPP) is a key task in educational assessment and can help teachers identify students’ learning obstacles in a timely manner and optimize teaching strategies. However, existing EPP models, although capable of providing quantitative analysis, suffer from issues such as complex algorithms, poor interpretability, and unstable accuracy. Moreover, the evaluation process is opaque, making it difficult for teachers to understand the basis for scoring. To address this, this paper proposes an approximate belief rule base (ABRB-a) student examination passing prediction method based on adaptive reference point selection using symmetry. Firstly, a random forest method based on cross-validation is adopted, introducing intelligent preprocessing and adaptive tuning to achieve precise screening of multi-attribute features. Secondly, reference points are automatically generated through hierarchical clustering algorithms, overcoming the limitations of traditional methods that rely on prior expert knowledge. By organically combining IF-THEN rules with evidential reasoning (ER), a traceable decision-making chain is constructed. Finally, a projection covariance matrix adaptive evolution strategy (P-CMA-ES-M) with Mahalanobis distance constraints is introduced, significantly improving the stability and accuracy of parameter optimization. Through experimental analysis, the ABRB-a model demonstrates significant advantages over existing models in terms of accuracy and interpretability. Full article

Deep learning models for atmospheric pattern recognition require spatially consistent training labels that align precisely with input meteorological fields. This study introduces an automatic cold front detection method using the ERA5 reanalysis dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF) at 850 hPa, specifically designed to generate physically consistent labels for machine learning applications. The approach combines the Thermal Front Parameter (TFP) with temperature advection (AdvT), applying optimized thresholds (TFP < 5 × 10⁻¹¹ K m⁻²; AdvT < −1 × 10⁻⁴ K s⁻¹), morphological filtering, and polynomial smoothing. Comparison against 1426 manual charts from 2009 revealed systematic spatial displacement, with mean offsets of ~502 km. Although pixel-level overlap was low, with Intersection over Union (IoU) = 0.013 and Dice coefficient (Dice) = 0.034, spatial concordance exceeded 99%, confirming both methods identify the same synoptic systems. The automatic method detects 58% more fronts over the South Atlantic and 44% fewer over the Andes compared to manual charts. Seasonal variability shows maximum activity in austral winter (31.3%) and minimum in summer (20.1%). This is the first automatic front detection system calibrated for South America that maintains direct correspondence between training labels and reanalysis input fields, addressing the spatial misalignment problem that limits deep learning applications in atmospheric sciences. Full article