Search Results (403)

Background/Objectives: Osteosarcoma (OS) is an aggressive bone malignancy with a complex tumor microenvironment (TME) that influences therapeutic outcomes and resistance. How neoadjuvant chemotherapy (NACT) reshapes the OS TME at single-cell resolution remains largely undefined. This study aimed to characterize cellular heterogeneity in the OS TME after NACT and identify chemokine-mediated intercellular crosstalk driving chemoresistance. Methods: Single-cell RNA sequencing was performed on surgical specimens from 12 OS patients (6 treatment-naive and 6 post-NACT). After quality control, 77,616 cells (36,214 from naive patients, 41,402 from post-NACT samples) were analyzed through the Seurat pipeline. Unsupervised clustering, differential expression analysis, and cell–cell communication network construction were performed, and candidate signaling axes were validated using transwell assays and Western blotting. Results: Cells were classified into 10 major cell types. Osteoblasts, identified as malignant cells, were partitioned into 11 subpopulations with marked transcriptional heterogeneity and differential PI3K/Akt pathway activity. Post-NACT, stromal and vascular components underwent molecular and functional remodeling, shaping an immune-activated microenvironment. Mononuclear phagocytes resolved into three discrete clusters—monocytes, macrophages, and dendritic cells—with differentiation gradients. Endothelial cells maintained robust CXCL2 expression throughout the therapeutic course. Functional validation via transwell assays and Western blotting confirmed that endothelial-derived CXCL2 promoted macrophage chemotaxis via CXCR2, with corresponding CXCR2 upregulation in macrophages. Conclusions: Collectively, these findings suggest the complex cellular and transcriptional heterogeneity of the OS microenvironment and its chemokine-driven molecular remodeling after NACT, indicating that TME dynamics may be a determinant of therapeutic response and chemoresistance. Full article

Today, the use of increasingly ubiquitous synthetic media, or ‘deepfakes’, has become a risk to online trust, information integrity and individual security and is being created by artificial intelligence (AI). The current approaches are mainly based on either spatial features of CNNs or high-level semantic representations of Vision Transformer; both have major drawbacks in effectively leveraging multi-domain forensic cues. This paper presents FAViT (Frequency-Aware Vision Transformer), a hybrid architecture capable of jointly utilizing spatial- and frequency-domain forensic information by the means of a bidirectional cross-attention fusion scheme. We use an 11-channel forensic tensor in each face image (including per-channel Fast Fourier Transform (FFT) magnitude maps, Discrete Wavelet Transform (DWT) sub-bands, channel noise residual maps, Sobel gradient magnitude and channels of Error Level Analysis (ELA)). A Frequency Branch CNN processes this multi-domain tensor and the original RGB image is encoded with a pretrained ViT-B/16 spatial branch. The two streams are combined through the bidirectional cross-attention which allows the model to localize both spatial and spectral manipulation artifacts. We also present an adversarial cleaning simulation pipeline which partitions the training process with five post-processing attack methods, namely GFPGAN neural face restoration, learned autoencoder cleaning, etc., to increase resistance to real-world forensic defenses. Tests of FaceForensics++ C23 (7926 images, consisting of four manipulation types) show that FAViT attains F1-score of 86.22, AUC-ROC of 94.26 and accuracy of 85.55 on the held-out test set. The strength analysis of 21 attack conditions shows that the max degradation in AUC is 30.3, with specific strengths in GFPGAN restoration (AUC = 98.51). Robustness is evaluated based on 21 post-processing attack cases that include JPEG compression, Gaussian blurring, down-sampling, and GFDGAN neural-based restoration; it should be noted that robustness against gradient-based adaptive attacks requires additional attention. Testing on the CIFAKE and Celeb-DF v2 datasets reveals some limitations of domain generalization. Full article

Background: Subvalvular aortic stenosis (SAS) is an important cause of fixed left ventricular outflow tract obstruction and the second most common form of aortic stenosis. Although traditionally considered a congenital lesion, increasing evidence suggests that SAS often behaves as a progressive condition influenced by anatomical and hemodynamic factors. The disease may lead to significant complications, including left ventricular hypertrophy and progressive aortic regurgitation. Materials and Methods: A narrative review of the current literature was performed to summarize contemporary knowledge regarding the epidemiology, pathophysiology, clinical presentation, diagnostic evaluation, and management of subvalvular aortic stenosis. Relevant studies were identified through searches of major medical databases and were critically analyzed to provide an overview of current concepts and clinical practice. Results: Subvalvular aortic stenosis demonstrates considerable anatomical and clinical heterogeneity, ranging from discrete subaortic membranes to fibromuscular tunnel-type obstruction. The condition often progresses over time, with increasing LVOT gradients and a high incidence of associated aortic regurgitation. Echocardiography remains the primary diagnostic modality, while advanced imaging techniques provide additional anatomical detail and assist in surgical planning. Surgical resection remains the cornerstone of treatment when significant obstruction or related complications develop. Conclusions: Subvalvular aortic stenosis is a complex and progressive disease requiring careful imaging assessment and timely surgical management. Although surgical outcomes are generally excellent, recurrence and progressive aortic valve involvement remain important long-term considerations. Ongoing research aimed at improving risk stratification and optimizing surgical strategies may further enhance patient outcomes. Full article

Cut-cell and ghost-cell finite-volume methods require an accurate boundary distance field, available in closed form only for simple shapes and otherwise obtained through costly nearest-neighbor search. Prior work embeds learned distance fields as an auxiliary component inside a larger neural architecture; this study instead tests the numerical consequences of substituting the distance function alone within an otherwise unmodified classical discretization. We propose a compact trained neural network as a drop-in surrogate for this field, evaluated against exact and classical alternatives, including a KD-tree baseline, on four problems of increasing difficulty ending with a three-dimensional mechanical flange. A truncation analysis bounds the surrogate’s intercept error and identifies the training accuracy needed to preserve the scheme’s formal order. The surrogate matches second-order accuracy in every example, with fitted orders of 2.0 to 2.1 in two dimensions and, in three dimensions, a wider but still order-consistent 1.9 to 2.5, the extra width traced to identified, shared mesh- and timestep-resolution effects rather than to the surrogate itself. Query cost overtakes the KD-tree beyond roughly one hundred thousand points, with speedups up to 26.1 times, and gradient fields are substantially smoother than nearest-point construction throughout, tracking geometric severity in two dimensions and feature scale in three. Full article

With the continuous advancement of deep and ultra-deep well drilling technologies, formations with complex pressure windows are becoming increasingly common. During casing running operations, it is necessary to ensure both leak prevention in loss-prone formations and pressure stabilization in high-pressure formations, demanding increasingly higher accuracy in wellbore pressure calculation. This study establishes a wellbore pressure calculation model for managed pressure casing (MPC) running in deep wells, specifically addressing the scenario where a multi-density gradient drilling fluid column exists in the annulus after tripping out. The model’s novelty lies in integrating transient surge pressure calculation with a dynamic fluid column structure model that tracks the displacement of multi-density drilling fluid layers during casing running. The governing equations based on one-dimensional unsteady flow theory are solved using the method of characteristics with adaptive time stepping and a grid independence study confirming the discretization scheme. Quantitative analysis reveals that casing running speed is the dominant factor affecting surge pressure; when the speed increases from 0.5 m/s to 1.5 m/s, the surge pressure increases from approximately 1.2 MPa to 3.5 MPa at a 2000 m depth. Drilling fluid properties also significantly influence surge pressure: increasing the density from 2.0 g/cm³ to 2.22 g/cm³ results in a surge pressure increase of approximately 0.6 MPa; increasing the yield value from 2.85 Pa to 15 Pa leads to an increase of about 1.1 MPa; the surge pressure shows a clear increasing trend with both the consistency coefficient and flow behavior index. Casing running depth affects the buffering effect of the bottomhole flow channel; when the casing is run to 7000 m, the surge pressure is approximately 0.5 MPa higher than at 2000 m. Taking a typical deep well (8578 m) with a negative pressure window of −0.008 g/cm³ as an example, three casing running speed plans were designed and evaluated. Plan 1 was selected with running speeds ranging from 0.16 m/s in the upper section to 0.115 m/s in the lower section, maintaining the equivalent circulating density (ECD) within the safe density window throughout the entire operation. Field application of this plan proceeded smoothly without any occurrences of lost circulation or overflow. This provides a practical basis for MPC running technology in deep wells with narrow or negative pressure windows. Full article

High-speed turbulent boundary layers govern the transport of momentum, mass, and energy in compressible flows and play a central role in determining aerodynamic performance, skin-friction drag, aerodynamic heating, flow stability, and thermal protection requirements of advanced aerospace vehicles. Over the past three decades, direct numerical simulation (DNS) has revolutionized the study of compressible wall-bounded turbulence by resolving all dynamically relevant turbulent scales without turbulence-model assumptions, providing benchmark-quality databases and unprecedented physical insight into flow phenomena that remain difficult or impossible to measure experimentally. Together with complementary high-fidelity approaches, DNS has substantially advanced the understanding of turbulence dynamics across a broad range of supersonic and hypersonic flow conditions. This review presents a critical assessment of advances in the high-fidelity simulation of compressible turbulent boundary layers under non-reacting conditions. Particular emphasis is placed on the flow physics of canonical zero-pressure-gradient boundary layers, shock-wave/turbulent-boundary-layer interactions (SWTBLIs), pressure-gradient-driven flows, streamline-curvature effects, and thermochemical non-equilibrium phenomena. Recent developments in numerical methodologies are also briefly examined, including high-order discretization techniques, turbulence inflow generation methods, hybrid continuum-kinetic formulations, and advances in high-performance computing that have enabled DNS at increasingly high Reynolds and Mach numbers. The review highlights the major physical insights emerging from DNS studies, demonstrating that many fundamental characteristics of compressible wall turbulence remain closely related to their incompressible counterparts when appropriate compressibility transformations are employed. At the same time, DNS has revealed the critical influence of wall temperature, pressure gradients, streamline curvature, shock interactions, and finite-rate thermochemistry on turbulence structure, coherent motions, interscale energy transfer, boundary-layer separation, and aerodynamic heating. Full article

Objectives: Sleep staging from polysomnographic (PSG) recordings is clinically critical for diagnosing sleep-related disorders, yet manual scoring by certified technologists remains time-consuming, costly, and subject to inter-rater variability. Methods: This study presents an automated, explainable, and multimodal framework for five-class sleep stage classification using simultaneously acquired electroencephalography (EEG), electrooculography (EOG), and electromyography (EMG) signals. A total of 1946 annotated 30 s epochs from 30 healthy adult recording sessions (Sleep-EDF Expanded and Sleep Cassette subset) were processed through a 37-dimensional multimodal feature extraction pipeline encompassing temporal amplitude statistics, frequency-domain spectral band powers, nonlinear entropy and complexity measures, and Daubechies-4 discrete wavelet transform (DWT) energy coefficients. Four classical machine learning classifiers -Random Forest (RF), Support Vector Machine with radial basis function kernel (SVM-RBF), Gradient Boosting (GB), and K-Nearest Neighbours (KNN, k = 7) were benchmarked under stratified five-fold cross-validation. Results: SVM-RBF achieved the highest macro-averaged F1-score of 0.7322 (Cohen’s kappa 0.6784, overall accuracy 75.18%). N3 deep slow-wave sleep achieved the highest per-class F1 of 0.879, while N1 light sleep was the most challenging (F1 = 0.668). SHapley Additive exPlanations (SHAP) and RF mean decrease in Gini impurity (MDGI) analysis jointly identified EMG root mean square amplitude (MDGI = 0.0805), gamma band power (0.0784), and permutation entropy (0.0434) as the three most discriminative features. As a novel methodological contribution, sixteen categories of signal sculpting visualisations were developed, translating abstract multivariate features into clinically interpretable graphical representations. Conclusions: The proposed framework achieves substantial kappa agreement approaching the lower bound of expert inter-rater reliability (0.76–0.82) while providing full model transparency, with direct implications for wearable sleep monitoring device design. Full article

Addressing the challenges of low environmental modeling accuracy and inadequate obstacle avoidance precision in complex obstacle scenarios in unmanned aerial vehicle (UAV) 3D path planning, this study proposes a UAV 3D path planning method that integrates the truncated signed distance field (TSDF) with an improved particle swarm optimization algorithm (IPSO). A unified planning space integrating a voxel occupancy grid with a truncated signed distance field is constructed offline: the Euclidean distance to obstacle surfaces is truncated and confined within an effective band, whose extent is coordinated with the UAV safety distance threshold determined by physical dimensions and task requirements, thereby preserving the continuous geometric information needed for safety assessment. On this basis, the continuous distance and gradient information provided by the truncated distance field are utilized to formulate a piecewise continuous, distance-based threat cost function, replacing traditional binary collision detection; the distance and gradient are further embedded into the initialization, fitness evaluation, and velocity update procedures of the particle swarm. Moreover, an adaptive inertia weight and a Lévy escape mechanism are introduced to improve search efficiency and global exploration capability. Experimental results demonstrate that under dense discrete safety verification, the proposed method achieves a 100% success rate in complex unstructured environments and that the safety distance threshold can be flexibly adjusted according to task requirements while consistently satisfying the specified safety requirement. The resulting paths achieve a favorable balance among length, smoothness, and controllable safety margin, validating the effectiveness of the proposed method. Full article

Near-infrared (NIR) spectroscopy-based adulteration detection approaches are still dominated by one-dimensional (1D) spectral analysis, which inherently limits the exploration of complex patterns and nonlinear interactions in spectral data. Therefore, the objective of this study is to use a two-dimensional (2D) NIR spectrogram, combined with Explainable Artificial Intelligence (XAI), to predict the level of adulteration in palmyrah sap. The dataset matrix dimension is 110 × 1101, derived from the sample adulteration level (0–100%) and the NIR wavenumber (4000–12,500 cm⁻¹). Following Kennard–Stone partitioning, the evaluated preprocessing methods were applied using parameters derived exclusively from the training set. For the 2D modeling branch, the resulting training and testing spectra were subsequently transformed separately using the Continuous Wavelet Transform (CWT). A total of six AI algorithms, three from machine learning (PLS, kNN, ANN) and three from deep learning (CNN, AlexNet, ResNet), were applied in this study. The best model AI was interpreted using Shapley Additive Explanations (SHAP) for 1D NIRs and the Gradient-weighted Class Activation Mapping (Grad-CAM) for 2D NIR spectrograms. The four best-performing model configurations can predict the level of palmyrah sap adulteration, with R² values ranging from 0.969 to 0.994 and RMSE ranging from 2.333% to 5.547% in the training. In the testing, the model’s performance is in the R² range of 0.959–0.990, RMSE of 3.093–6.396%, MAE of 2.358–4.252%, RPD of 5.06–10.46 and Bias of 0.03–0.93%. The SHAP and Grad-CAM XAI revealed that the wavenumber associated with this sap counterfeiting is critical to the level of adulteration of palmyrah sap. This approach provides a quantitative method that accounts for advanced dimensions and treats them as essential information to support large-scale data matrices in AI modeling. The application of this method is an alternative that is easy to interpret and implement, and can be applied to long- and short-wavelength data from continuous NIR or discrete multi-wavelength NIR. Full article

Conventional resistive-type superconducting fault-current limiters (RSFCLs) rely exclusively on fault currents and temperature increases to trigger quenching, resulting in delayed fault response and excessive heat buildup under short-circuit conditions. To mitigate these limitations, this paper proposes a self-triggered, magnetic-field-excited superconducting current-limiting reactor (SCLR) integrated with a solenoidal magnet assembly. During the design and simulation phases, a segmented discretization method is employed to quantitatively characterize the gradient distribution of the external perpendicular field within the superconducting tapes and coils. This approach theoretically elucidates the mechanism by which spatially non-uniform magnetic fields influence current-limiting performance. DC short-circuit simulations show that the background magnetic field instantly reduces the critical current, rapidly transitioning the superconducting layer into a nonlinear resistive state. In contrast to the conventional topology, the proposed SCLR achieves two key performance improvements during short-circuit faults: it limits the peak fault current to just 45.9% of the value recorded with the conventional RSFCL scheme, and it reduces the maximum temperature rise by 1.4 K. The findings of this study provide a theoretical foundation and technical references for multi-field coupling modeling and structural optimization of magnetic-field-regulated current-limiting devices in DC grids. Full article

Extracting transferable design knowledge from Discrete Element Method (DEM) simulations remains challenging in granular material handling. We develop an explainable surrogate framework for two solar-panel-recycling subsystems: a silo discharge system (outlet width 42–111 mm; hopper half-angle 30–60°) and an inclined belt conveyor (fin height 20–50 mm; belt velocity 0.109–0.627 m/s). Five algorithms—Response Surface Methodology (RSM), Artificial Neural Network (ANN), Random Forest (RF), Gradient Boosting Machine (GBM), and Gaussian Process Regression (GPR)—were evaluated by leakage-free grouped five-fold cross-validation using 34 design points per system (2² factorial expanded by Latin Hypercube Sampling). The best surrogate is response-specific: ANN was most accurate for silo discharge (R² = 0.964, RMSE = 0.91 kg/s); GPR gave the highest raw belt-MFR accuracy (R² = 0.976 ± 0.023, RMSE = 0.43 kg/s), though the interpretable RSM was near-equivalent and was adopted for optimization; and RSM was near-perfect for belt discharge angle (R² = 0.999 ± 0.001, RMSE = 0.057°). GPR performed the worst for silo discharge (R² = 0.384), confirming the algorithm selection must match the response complexity. SHAP analysis identified outlet width (78.6%) and belt velocity (58.9–76.8%) as dominant predictors. The proposed SHAP Asymmetry Ratio (SAR) offers an exploratory diagnostic for algorithm pre-selection in expensive simulation-driven surrogate workflows. Full article

This study presents a materials-oriented, code-based parametric optimization of the load-bearing steel tubular arch girder in a tied-arch footbridge inspired by the Father Bernatek Footbridge in Krakow. The objective was to reduce structural steel demand by minimizing the arch-girder weight under Eurocode load combinations with ultimate limit state (ULS) and serviceability limit state (SLS) constraints, while accounting for discrete tubular cross-section changes within a realistic finite element model. A semi-automated workflow linked Autodesk Dynamo, Python scripts, and Autodesk Robot Structural Analysis to generate bridge geometry, build the finite element method (FEM) model, apply code-based loads and combinations, and evaluate structural response using a discrete, non-gradient-based search. A preliminary sensitivity screening was performed for the full set of design parameters, while the final optimization was governed mainly by arch rise, hanger number, and ULS-controlled discrete arch cross-section changes. The optimization reduced the arch-girder weight by 10.7% relative to the reference configuration, from 360.3 × 10³ kg to 321.6 × 10³ kg, within the adopted design domain. The optimum solution corresponded to an arch rise of 23.4 m, 31 hangers, and a deck spacing of 6.0 m. Hanger arrangement strongly affected force redistribution in the arch girder, while the final optimum was controlled by code-based utilization thresholds. The results show that an application programming interface (API)-driven parametric workflow can support early-stage optimization of tied-arch footbridges under code-based design constraints. The scientific contribution of the study lies not in automating Eurocode verification alone, but in identifying the structural mechanisms that govern the minimum-weight solution, including the interaction between arch rise, hanger arrangement, force redistribution, ULS utilization, and discrete tubular cross-section changes. Full article

Metaheuristic optimization algorithms are widely used in wrapper-based feature selection, since they can search complex combinatorial spaces without gradient information. The standard sine cosine algorithm (SCA) is simple and easy to implement, but depends on the current population state and a single global optimal guide. This dependence can cause premature convergence and late-stage oscillations in discretized feature-selection tasks. To address these issues, we propose the fractional-order elite-memory sine cosine algorithm (FOSCA). The FOSCA integrates short-memory fractional-order position reconstruction, dynamic elite-center guidance and nonlinear search-factor decay into the SCA search dynamics. These mechanisms improve the trajectory continuity, guidance diversity and convergence stability. Experiments on 14 classification datasets showed that the FOSCA achieved a competitive accuracy, F1-score, precision and recall. The FOSCA also outperformed the original SCA on 12 out of the 14 datasets. Statistical tests, a stability analysis, a performance–sparsity trade-off analysis and ablation studies confirmed the effectiveness of the proposed mechanisms. Overall, the FOSCA improves the SCA search reliability in discrete feature selection and offers a reproducible basis for related optimizer-based methods. Full article

Ransomware detection remains challenging because modern variants exhibit diverse, elusive, and partly benign behaviors and can propagate rapidly across interconnected enterprises and sensor-enabled cyber-physical systems, causing cascading operational failures. These characteristics undermine signature-based and static-detection methods. Although machine learning has improved detection, many approaches still rely on fixed objectives that weight samples uniformly, limiting their adaptation to heterogeneity and overlaps between ransomware and benign activities. To address this challenge, we introduce TL-RL-FusionNet, a reinforcement learning (RL)-guided hybrid framework that combines dual transfer learning (TL) backbones, EfficientNetB0 and InceptionV3, with a lightweight residual multi-Layer perceptron (MLP) classifier. The framework converts sandbox reports into RGB grids, extracts features using frozen CNN backbone networks, and fuses embeddings for classification. Training is guided by a tabular Q-learning sample-weighting agent, formulated as a per-sample bandit over discrete weight actions. To prevent cross-fold information leakage, the Q-table is freshly initialized in each cross-validation fold and updated only using the fold-local training partition, whereas the held-out fold is used for the final evaluation. The framework was evaluated using two datasets. On our dataset, TL-RL-FusionNet achieved the best overall performance on Dataset 1, with 99.20% accuracy, 99.40% recall, and 99.84% AUC. On the public EldeRan benchmark, it achieved 90.36% accuracy using the full dynamic feature space and 92.08% using a Mutual Information-selected compact subset. Paired Wilcoxon tests across five folds were used to assess the RL contribution, while additional grid-order sensitivity analysis showed that the image-based representation remained robust under five random 10 × 10 feature-grid permutations. Interpretability analysis using t-distributed stochastic neighbor embedding (t-SNE) and gradient-weighted class activation mapping feature-grid mapping further showed that the model captured discriminative behavioral patterns. Overall, these results demonstrate that RL-guided sample reweighting improves adaptive ransomware detection while maintaining efficiency and interpretability. The dataset and supporting code are publicly available on GitHub. Full article

Near-field direct illumination from extended luminaires presents a fundamental modeling challenge in automated visual inspection: standard far-field IES (Illuminating Engineering Society) photometric data treat luminaires as point sources, an assumption that breaks down when source-to-target distances are comparable to luminaire dimensions. This paper addresses this breakdown by establishing, both analytically and experimentally, the conditions under which IES-based point-source discretization remains valid and the minimum discretization required when it does not. A solid-angle-based illuminance formulation on triangular meshes is coupled with a controlled virtual-emitter discretization of elongated luminaires, and a reproducible selection workflow is derived that relates emitter count to the source–target distance ratio and a specified error criterion. The dependence of discretization requirements on source–target distance is characterized by an angular-subtense argument, yielding the scaling relation $N\sim (L/H)·C$, where L is the luminaire length, H the working distance, and C a constant determined by the IES angular gradient and the required accuracy. Validation experiments with one and two industrial linear luminaires (1.2 m, Banner Engineering WLS15xDW1200Dx) include single-luminaire heights of $H=20\mathrm{cm}$, $25\mathrm{cm}$, $35\mathrm{cm}$, $50\mathrm{cm}$, and $60\mathrm{cm}$, as well as effective-length reduction cases produced by opaque end masking at $H=35\mathrm{cm}$. The single-luminaire validation set compares $N=1$–7 virtual emitters for all cases, with the $H=20\mathrm{cm}$ case extended to $N=9$. The selected emitter counts are chosen as the best-performing candidate simulations after considering scalar error, smoothness, and profile consistency. The selected uncovered cases are $N=9$ at $H=20\mathrm{cm}$, $N=7$ at $H=25\mathrm{cm}$, $N=5$ at $H=35\mathrm{cm}$, and $N=3$ at $H=50$–$60\mathrm{cm}$. The masked-source experiments confirm that reducing the effective source length at fixed height reduces the required emitter count. The fitted selection relation developed from these cases indicates that the working distance H has the stronger influence within the tested range. The method provides a computationally efficient, practitioner-ready tool for inspection-lighting design when near-field goniophotometry or optical ray tracing is unavailable. Full article