Search Results (2,681)

Spatially explicit water-quality information is critical for precision management in pond aquaculture but point sampling alone cannot capture pond-to-pond heterogeneity in multi-unit farms. This single-date, single-farm study re-evaluated the potential of UAV hyperspectral imagery for water-quality screening in inland aquaculture ponds in Shanghai, China, using site-matched extraction from a 138-band orthomosaic (450–998 nm, Cubert S185) acquired during a single UAV survey on 24 August 2023 and matched with 23 GPS-registered sampling sites. Eight water-quality parameters were analyzed: chemical oxygen demand (COD), total phosphorus (TP), total nitrogen (TN), ammonium (${\mathrm{NH}}_4^{+}$ ), nitrite (${\mathrm{NO}}_2^{-}$), nephelometric turbidity unit (NTU), chlorophyll-a (Chla), and total suspended solids (TSS). Raw single-band correlations were modest ($\left|\mathrm{r}\right|$= 0.236–0.417), but two-band difference spectral indices (DSI), normalized spectral indices (NSI), and ratio spectral indices (RSI) substantially improved sensitivity, with $\left|\mathrm{r}\right|$ reaching 0.558–0.928. Quadratic inversion models were calibrated on the full dataset and assessed using three validation layers: two-fold cross-validation, nested leave-one-pond-out (LOPO) validation with within-fold predictor reselection, and extraction-window sensitivity tests. Bootstrap 95% confidence intervals for calibration (Cal) R² characterize small-sample uncertainty (n = 23). Three parameters satisfied all three defensibility criteria (Cal R² > 0.5, CV R² > 0.2, and LOPO R² > 0.2): ${\mathrm{NH}}_4^{+}$ (Cal R² = 0.836 [0.61, 0.94]; LOPO R² = 0.420), COD (0.607 [0.34, 0.82]; 0.328), and NTU (0.862 [0.77, 0.96]; 0.204). TP, TN, ${\mathrm{NO}}_2^{-}$, TSS, and Chla showed overfit behavior under nested holdout and were demoted to exploratory products. A TreeSHAP analysis confirmed that band-to-band contrast carried more explanatory power than raw reflectance magnitude. Extraction-sensitivity tests further demonstrated that positional uncertainty (±2-pixel offset: ΔCV $R^2$= 0.23–0.41) exceeded averaging-window sensitivity (3 × 3→10 × 10: ΔCV $R^2$ ≤ 0.11), identifying geolocation control as the dominant robustness constraint. This single-date, single-farm reanalysis suggests that UAV hyperspectral imagery may support exploratory pond-scale screening of ${\mathrm{NH}}_4^{+}$, COD, and NTU. However, robust quantitative inversion and broader transferability remain unverified and will require denser sampling, improved geolocation control, pond-edge masking, multi-site observations, and multi-temporal calibration. Full article

Arthropathies are a major global health challenge because of their high prevalence, chronic progression, and significant impact on quality of life and health systems. Therefore, prompt and accurate diagnosis is critical for slowing disease progression and improving outcomes. Traditional imaging modalities, such as ultrasound and magnetic resonance imaging, suffer from significant limitations, including operator dependence, limited accessibility, high cost, and limited reproducibility. Infrared thermography has become a promising non-invasive imaging technique for identifying thermal variations linked to inflammatory and metabolic processes. Advances in quantitative thermography, automated segmentation, and artificial intelligence have greatly enhanced its clinical applicability. This review summarizes recent advances in thermography-based biomarkers, including region-of-interest-derived metrics, asymmetry indices, hotspot burden, spatial and texture descriptors, and composite thermographic scores. It discusses the role of machine learning and deep learning in prediction, phenotyping, and multimodal integration with clinical, laboratory, and imaging data. Heterogeneity of protocols, variability in measurements, domain shift, validation design, overfitting, and reporting quality are also addressed. Overall, thermography combined with AI is highly promising as an adjunct to early diagnosis, assessment of disease activity, and follow-up in arthropathies. However, clinical application at a large scale requires strict standardization, external validation, transparent reporting, and well-elucidated, reproducible analytical processes. Full article

Automated detection of archaeological stone cairns using high-resolution satellite imagery offers a scalable approach for documenting vulnerable heritage landscapes in the Ennedi Massif, where extensive and remote terrain limits traditional field survey, and rapid documentation is required. This study presents a GIS and deep learning framework based on the YOLOv8 model to identify and map stone cairns using Google Satellite RGB imagery at 28.5 cm spatial resolution. Ground-truth data collected via GPS field survey were used to train and validate YOLOv8n. The study area was divided into two regions with contrasting terrain and illumination conditions to evaluate model transferability. The training region included 149 verified cairns, while the independent test region included 103 cairns. Early stopping reduced overfitting, reaching mAP50 of 99.5% and mAP50–95 of 94.3%. A density-based spatial clustering algorithm was applied to merge overlapping detections and generate circular cairn representations. On the test set, the model achieved 83.5% precision, recall, and F1-score, indicating stable performance under the selected operational configuration. Comparison with YOLOv5n showed slightly higher localization accuracy for YOLOv8n, while YOLOv5n yielded marginally higher precision and F1-score. Overall, the framework provides a non-invasive tool for large-scale archaeological prospection and heritage monitoring in remote desert environments. Full article

The volumetric mass transfer coefficient ($k_{L}a$) governs the design, operation, and scale-up of aerobic bioprocesses, yet its dependence on reactor geometry, impeller design, operating conditions, and fluid properties limits prediction by empirical correlations. Machine learning (ML) improves accuracy but faces two barriers in bioprocess practice: selecting the best model among many candidates requires expertise, and small, highly multicollinear data make models chosen based on test error alone prone to overfitting. Using a browser-based, no-code platform, we trained 14 regression algorithms under an identical pipeline on a published $k_{L}a$ dataset, and introduced a composite objective, the generalization-penalized error (GPE), which is the test RMSE plus the absolute train–test RMSE gap. Minimizing GPE rather than test RMSE expanded the top statistically equivalent group to include not only boosting ensembles but also simpler, interpretable models, indicating that black-box models hold no clear advantage once train–test consistency is assessed. Sensitivity analysis showed that tree models produce discontinuous responses, whereas algebraic learning via elastic net (ALVEN) yields smooth surfaces. Shapley additive explanations (SHAP) and an ontology graph, interpreted by a retrieval-augmented language-model agent, identified rotational speed and gas flow rate as dominant, reproducing the established mass transfer mechanism. The framework offers a reproducible, interpretable, expertise-light route to bioprocess model selection. Full article

Impaction of the mandibular third molar in proximity to the mandibular canal increases the risk of inferior alveolar nerve injury. Panoramic radiography is routinely used to assess this relationship. Automated classification of molar–canal overlap could support clinical triage and reduce unnecessary CBCT referrals, while Federated Learning (FL) enables multi-center collaboration without sharing patient data. We compared Local Learning (LL), FL, and Centralized Learning (CL) for binary overlap/no-overlap classification on cropped panoramic radiographs partitioned across eight independent labelers in a simulated heterogeneous multi-center setting. A pretrained ResNet-34 was trained under each paradigm and evaluated using per-client metrics with locally optimized thresholds and pooled test performance with a global threshold. Performance was assessed using area under the receiver operating characteristic curve (AUC) and threshold-based metrics, alongside training dynamics, Grad-CAM visualizations, and server-side aggregate monitoring signals. On the test set, CL achieved the highest performance (AUC 0.831; accuracy ≈ 0.782), FL showed intermediate performance (AUC 0.757; accuracy ≈ 0.703), and LL generalized poorly across clients (AUC range ≈ 0.619–0.734; mean ≈ 0.672). Training curves suggested overfitting, particularly in LL models, and Grad-CAM indicated more anatomically focused attention in CL and FL. Overall, centralized training provided the strongest performance, while FL offers a privacy-preserving alternative that outperforms LL. Full article

Systematic trend-following strategies applied to equity markets are widely studied, yet most reported performance statistics are non-reproducible in live trading. This paper makes three contributions. First, we introduce a formal taxonomy of look-ahead bias organised around point-in-time correctness: a strategy is point-in-time correct if, for every decision time t, its information set lies in the natural filtration ${\mathcal{F}}_t$. Three bias classes—universe-membership contamination, price-data forward leakage, and stop-exit sequencing violations—are characterised as filtration breaches. Second, we formalise the average true range (ATR) trailing stop as a stochastic recurrence and codify its monotonic non-decreasing ratcheting property (Lemma 1), providing a structural per-trade loss bound. Third, we exhibit a closed-form construction (Theorem 1) of two return sequences with identical Sharpe ratios but arbitrarily divergent maximum consecutive negative-year runs, establishing investor-experience metrics as independent optimisation objectives. We complement these contributions with an 18-year empirical study (2008–2025) on the NASDAQ-100 with reconstructed point-in-time index constituency (Class I compliant) and measured residual Class II exposure, applying combinatorially symmetric cross-validation (CSCV) to a 14-configuration ATR-multiplier grid. The grid exhibits a stop-multiplier-insensitive, CAGR-flat region across $k\in [3.5,7.0]$ (CAGR 10.28–10.39%, net of Dutch progressive tax) and a uniform maximum consecutive negative-year run of 1 across all 14 configurations. The correlation-matrix eigenvalue spectrum of the grid is dominated by a single mode (${\lambda }_1=13.91$ of 14), yielding an effective independent-test count of $M_{\mathrm{eff}}=1.09$. This near-degeneracy persists in a parallel grid with the regime classifier disabled, establishing the ATR multiplier as a structurally near-redundant parameter for this strategy class. The associated PBO value of $=0.9351$ co-occurs with this near-degeneracy under the CSCV maximum-selection rule. The plateau-level performance survives Bonferroni correction for both $M=14$ and $M_{\mathrm{eff}}$. The combined evidence supports a region-based interpretation of robust strategy parameters rather than single-point optimisation. Full article

This study presents the first comprehensive machine learning framework for predicting the performance of an air-breathing polymer electrolyte membrane fuel cell, based on high-fidelity multiphysics data and validated under realistic conditions. Using data generated from a validated multiphysics model, four machine learning models are trained: MLR, RFR, ANN, and SVR. The models aim to capture the effects of geometric, material, and operating parameters on cell performance to support the development of more efficient and sustainable clean energy systems. Evaluation with standard error metrics shows that MLR exhibits large deviations from actual values, highlighting the limitations of linear models and underscoring the need for more complex approaches. ANN and SVR provide high predictive accuracy and generalize well to unseen data, while RFR tends to overfit. Robustness analysis using white Gaussian noise and four-fold cross-validation further confirms the reliability of top-performing models. ANN and SVR models generate polarization curves 4000 and 40,000 times faster, respectively, than the multiphysics model, enabling real-time applications. Both models achieved excellent predictive performance, with R² values exceeding 0.999 under normal operating conditions and remaining above 0.98 even in the presence of noisy inputs. Full article

Because injuries are often vague and easily unnoticed, missed diagnosis of scaphoid fractures on emergency radiographs reveals a critical limitation of acute care imaging. In addition, owing to unremarkable radiographic features, scaphoid fractures are particularly challenging. Therefore, a deep learning-based scaphoid fracture detection (DLSFD) framework is developed in this study for predicting scaphoid fractures on anteroposterior wrist radiographs. A ten-year retrospective cohort of wrist radiographs including both fractures and non-fractures were collected and analyzed in the study. Furthermore, data augmentation and labeling were used to improve model performance. The proposed deep learning-based scaphoid fracture detection framework first applies the YOLOv8 algorithm to localize and segment the scaphoid region in anteroposterior wrist radiographs. Then, a U-Net-based classifier is employed to predict the fracture or non-fracture with 5-fold cross-validation to prevent overfitting. Instead of using heat maps to represent the regions of scaphoid fractures, this study carries out pixel-level segmentation and generates pixel-wise masks to clearly locate scaphoid fracture area. Numerical results indicate that the proposed DLSFD framework is a feasible and promising alternative in predicting scaphoid fractures in terms of classification performance. Moreover, overlay segmentation masks generated by the developed DLSFD framework provide visual assistance for clinical interpretation. Thus, the designed DLSFD framework is able to successfully identify scaphoid fractures and may be useful in clinical practice for assisting clinical assessment. Full article

Human sitting posture recognition is critical for smart seating, ergonomic monitoring, and healthcare systems. However, existing deep learning approaches typically rely on highly complex network architectures that are computationally expensive, hindering their lightweight deployment on edge devices. Furthermore, current methods frequently struggle with indistinct boundaries among multi-class postures and are highly prone to overfitting when constrained by small-sample pressure sensor datasets. To bridge this gap, this paper proposes a novel, lightweight posture recognition framework specifically tailored for pressure distribution maps. First, sitting pressure data is collected using a thin-film pressure array sensor and uniformly mapped into an [M × N] image representation, establishing an effective sample format for Convolutional Neural Network (CNN) inputs. Second, as our primary architectural contribution, we fundamentally optimize the classic LeNet-5 network to enhance complex feature representation without inflating model complexity. Specifically, the depth of the convolutional layers is increased with a progressively increasing channel configuration. Batch Normalization (BN) is introduced to accelerate convergence and ensure training stability, while a Dropout mechanism is embedded within the fully connected layers to strictly penalize overfitting under small-sample constraints. These architectural improvements are synergistically combined with targeted data augmentation strategies—including random translation, rotation, and intensity perturbation—to further strengthen the model’s generalization capability. Experimental results demonstrate that the proposed method achieves a classification accuracy of 95.5% in a five-class sitting posture recognition task, significantly outperforming baseline models such as the traditional LeNet-5, AlexNet-Lite, and VGG-Small. The findings indicate that this approach achieves an optimal balance among recognition accuracy, training stability, and low model complexity, providing a robust algorithmic baseline and proof-of-concept for smart healthcare perception systems, paving the way for future large-scale subject-independent validation. Full article

Urban canopy height is an essential indicator for characterizing vegetation structure and carbon sequestration, yet satellite LiDAR often lacks sufficient spatial resolution, airborne LiDAR is costly, and SAR has limited sensitivity to vegetation structure. This study proposes a canopy height inversion framework using high-resolution stereo pairs from the Gaofen-7 (GF-7) satellite. A 0.65 m Digital Surface Model (DSM) was generated from GF-7 data, and a relative surface height was derived by differencing the GF-7 DSM from a coarse 30 m DSM reference. Key features were selected via Boruta and Random Forest Recursive Feature Elimination (RF-RFE), and six models—linear, polynomial, support vector machine, backpropagation neural network, XGBoost, and RF—were compared. The results showed that the Boruta feature set improved average R² by 8.2%. Among all models, RF performed best (test set R² = 0.71, RMSE = 1.70 m) and exhibited the strongest resistance to overfitting. Canopy heights within Beijing’s Fifth Ring Road showed an “outer-high, inner-low” pattern: large parks exceeded 30 m, while the Central Business District remained below 3 m. GF-7 stereo pairs enable efficient and cost-effective retrieval of canopy height in fragmented urban green spaces, supporting ecological parameter quantification and urban green-space management. Full article

Lung cancer is the deadliest cause of cancer-related deaths worldwide, and early and accurate detection is key to improving patient outcomes. IQ-OTH/NCCD CT scan images are used in this study to present an optimized computer-aided diagnosis (CAD) framework for lung cancer detection. In order to extract deep features and improve diagnostic accuracy, a weighted geometric mean (WGM) ensemble of pretrained convolutional neural networks (CNNs) called the LCD-VRD model—comprising VGG16, ResNet50V2, and DenseNet121—provides robust feature extraction and strong generalization capabilities for accurately classifying normal, benign, and malignant (cancerous) cases. To actively mitigate data imbalance and reduce model overfitting, real-time data augmentation alongside rigorous class weighting was implemented. The results show that, with 97.27% accuracy and a 97.24% F1-score, the WGM ensemble of these models performs exceptionally well. Furthermore, Gradient-weighted Class Activation Mapping (Grad-CAM) visualization was investigated on CT images to provide an exploratory qualitative visualization of the image regions associated with model predictions. While the proposed framework shows promise as an effective tool for automated lung cancer diagnosis, its validation is currently limited to the IQ-OTH/NCCD dataset. External dataset evaluation will be essential to fully establish robustness and clinical applicability. Full article

Background/Objectives: Early and accurate discrimination of neurological conditions, dementia, stroke and healthy aging, remains a critical clinical challenge. Electroencephalography (EEG) is a non-invasive measure of brain dynamics and entropy-based features obtained from multichannel EEG have shown strong discriminative ability. However, existing deep learning approaches do not sufficiently address the combined challenges of small clinical cohorts and high-dimensional entropy feature spaces. In this study, a novel architecture is proposed for multi-class neurological EEG classification under extreme small-sample conditions. Methods: A novel dual-branch Channel-wise Transformer and Attention-Branch Network (EEG-ChTABNet) are pr to classify 19-channel EEG entropy features into three classes (dementia, stroke, healthy control; N = 45; 15 per class). The architecture suggests four new designs. First, the Channel Importance Attention (CIA) block, which adaptively learns to re-weight the importance of electrodes via squeeze-excitation. Second, the dual-branch encoder, which combines the global multi-head self-attention with the local depthwise-separable convolution. Third, the gated sigmoid fusion mechanism. Fourth, the bottleneck residual classification head, to solve overfitting. Eight entropy feature sets: Amplitude-Aware Permutation Entropy (AAPE), Attention Entropy (AttEn), Dispersion Entropy (DisEn), Distribution Entropy (DistrEn), Fluctuation-based Dispersion Entropy (FDispEn), Fuzzy Entropy (FuzEn), Linear Gaussian Estimation of the Conditional Entropy (LinEn), and Symbolic Dynamics (SyDy) were evaluated individually with stratified 5-fold cross-validation on within-fold SMOTE augmentation. Results: EEG-ChTABNet consistently outperformed the baseline Transformer on all 8 feature sets. DisEn and SyDy features yielded peak classification accuracy of 73.3% (AUC: 0.823 and 0.857, respectively) compared to the corresponding baseline of 57.8% and 55.6%. SyDy achieved the best overall AUC of 0.857 and the dementia detection sensitivity was up to 86.7% over multiple feature sets. Conclusions: EEG-ChTABNet shows the effectiveness of channel-adaptive, dual-branch Transformer Designs for EEG-based neurological classification from Small-Sample Entropy Feature Data, and Identifying SyDy and DisEn as the Most Discriminative Feature Representations for Three-Class Neurological EEG Classification. Full article

Glioblastoma (GBM) is one of the most aggressive human cancers, with therapeutic failure driven by pronounced intratumoral heterogeneity, microenvironmental plasticity, immune suppression, blood–brain barrier (BBB)-related pharmacological constraints, and adaptive resistance mechanisms. A major limitation in GBM research is the lack of a human-relevant experimental system able to reproduce these dynamic features while generating interpretable, multimodal datasets. In this context, we propose a testable organ-on-chip (OoC)-extracellular vesicle (EV)-deep learning (DL) framework in which patient-derived GBM cells, endothelial cells, astrocytes, pericytes, stromal cells, and immune components are organized within perfused microphysiological systems. EVs are selectively and temporally harvested from defined compartments, and imaging, barrier-function, sensor, and EV-cargo data are integrated through modality-specific and multimodal DL architectures. This framework is intended not as an immediately validated clinical tool but as an experimental roadmap for linking EV-mediated communication to measurable phenotypes such as BBB disruption, invasion, immune reprogramming, and drug response. We critically discuss the technical requirements of BBB-on-chip systems, EV source attribution, immune-component integration, DL model selection, data scarcity, overfitting, batch effects, domain shift, regulatory barriers, cost, throughput, and reproducibility. By repositioning OoC-EV-DL integration as a staged translational strategy rather than a clinically established solution, this work aims to define a realistic and biologically grounded route for advancing precision oncology in GBM. Full article

How does artificial intelligence (AI) adoption shape administrative efficiency and e-citizen integration in local governments, and what role does professional development play in mediating these relationships? Drawing on a survey of 500 municipal employees across Spanish municipalities, this study employs partial least squares structural equation modelling (PLS-SEM), with formal non-linearity testing via Warp3 algorithms, to test a theoretically grounded model. The conceptual framework integrates Digital Transformation Theory and Public Value Theory as primary explanatory lenses, while drawing on the Technology Acceptance Model (TAM) and Total Factor Productivity (TFP) logic as complementary background perspectives that contextualise rather than directly operationalise the micro-level findings. Structural results reveal that AI adoption exerts a strong direct (and statistically linear) effect on perceived administrative efficiency (β = 1.04, p < 0.001; the standardised coefficient exceeding 1.0 and R² > 1 are a legitimate WarpPLS warp-model fit index rather than evidence of model misspecification: the Warp3 warp functions inflate the variance of predicted efficiency and break the additive identity SST = SSM + SSE, with the high AI–PD collinearity (r ≈ 0.84) as the contributing mechanism (RSCR = 1.000, SSR = 1.000); a comparative re-estimation without the moderation term yields β = 0.87 and R² = 0.76; we adopt this parsimonious specification (β ≈ 0.87, R² = 0.76) as the substantively interpretable estimate, with predictive relevance confirmed by a high Stone–Geisser Q² = 0.685, indicating that the model fits and predicts well rather than overfitting, while simultaneously stimulating professional development (β = 0.84, p < 0.001, R² = 0.70). Professional development positively predicted both efficiency (β = 0.27, p < 0.001) and e-citizen integration (β = 0.26, p < 0.01). Efficiency is the primary driver of e-citizen integration (β = 0.54, p < 0.001, R² = 0.53). The proposed moderation of AI adoption by professional development on efficiency was not supported (β = −0.01, p = 0.44), suggesting additive rather than synergistic effects. Model fit was robust (GoF = 0.701; ARS = 0.749; APC = 0.495); convergent and discriminant validity were confirmed by composite reliability, average variance extracted, Fornell–Larcker, and HTMT criteria; and common method bias diagnostics (Harman’s single-factor test, full-collinearity AFVIF, and marker-variable analysis) indicated that systematic method variance was not a material threat. These findings offer micro-empirical evidence of the mechanisms linking AI adoption to citizen service outcomes via a professional development pathway and provide actionable recommendations for Spanish and European municipalities navigating AI-driven governance reform. Full article

A virtual training environment offers clear advantages for agricultural robotics. It provides a safe setting in which perception, navigation, and control algorithms can be evaluated without risking damage to either the robot or the crop. It also supports efficient data generation: large volumes of training data can be collected under diverse environmental conditions that would be costly, slow, and often season-dependent in real-world deployments. This broader variability improves model adaptability, reduces the risk of overfitting, and leads to more robust operation. In this paper, we argue that digital twin technology should therefore be understood not merely as a passive mirror of a physical robot, but as an active training environment in which multiple sensor-related subprocesses can be developed, tested, validated, and refined jointly. This paper is based on our experiences with digital twin technology used in the development of a vineyard robot, including a self-driving rover, sensor simulation, procedural map generation, and agriculture-specific movement models. Our contribution is threefold: we reinterpret the digital twin as a training space, propose a layered framework for training agricultural robots in virtual environments, and explain why agriculture is a particularly strong use case, given variable field conditions, expensive real-world experimentation, and persistent labor scarcity. To validate this framework, we present the simulation-based evaluation of an autonomous reinforcement learning agent. The agent has been trained entirely in this virtual environment, which successfully navigated to 155 out of 161 target points in a simulated vineyard demonstration environment. Full article