Search Results (2,272)

With the expanding application of synthetic aperture radar (SAR) in ocean monitoring and port regulation, nearshore ship detection based on SAR image faces notable challenges arising from strong background scattering, dense target occlusion, and large pose variations. Therefore, this paper proposes a two-stage oriented detection network named EARS-Net to improve the accuracy of ship detection in complex nearshore environments. Specifically, a lightweight convolutional block attention module (CBAM) is embedded into the high-level semantic stages of ResNet50 to enhance discriminative ship features while suppressing interference from port infrastructures and shoreline structures. Then, the dynamic angle regression loss (DAL) is proposed, and the angle weight function is designed according to the ship direction distribution characteristics, which allocates higher regression weight to the ship target with larger tilt angle, improving the defect of insufficient positioning accuracy for large angle ships. Moreover, a training strategy that combines focal loss, multi-scale training, and rotated online hard example mining (ROHEM) is employed to alleviate sample imbalance and improve generalization in dense scenes. Experimental results on the nearshore subset of the SSDD show that EARS-Net achieves an average precision (AP) of 0.903 on the test set, demonstrating reliable detection capability under complex backgrounds and dense target distributions. These results validate the effectiveness of our method and highlight its potential as a practical engineering solution for enhancing port situational awareness and coastal security monitoring. Full article

Background/Objectives: Peptide receptor radionuclide therapy (PRRT) is an established treatment for metastatic neuroendocrine tumors (NETs), yet long-term disease control occurs only in a subset of patients. Predicting progression-free survival (PFS) could support individualized treatment planning. This study evaluates laboratory, imaging, and multimodal deep learning models for PFS prediction in PRRT-treated patients. Methods: In this retrospective, single-center study 116 patients with metastatic NETs undergoing [177Lu]Lu-DOTATOC were included. Clinical characteristics, laboratory values, and pretherapeutic somatostatin receptor positron emission tomography/computed tomographies (SR-PET/CTs) were collected. Seven models were trained to classify low- vs. high-PFS groups, including unimodal (laboratory, SR-PET, or CT) and multimodal fusion approaches. Performance was assessed via repeated 3-fold cross-validation with area under the receiver operating characteristic curve (AUROC) and area under the precision–recall curve (AUPRC). Explainability was evaluated by feature importance analysis and gradient based saliency maps. Results: Forty-two patients (36%) displayed short PFS (≤1 year) and 74 patients displayed long PFS (>1 year). Groups were similar in most characteristics, except for higher baseline chromogranin A (p = 0.003), elevated $\gamma$-GT (p = 0.002), and fewer PRRT cycles (p < 0.001) in short-PFS patients. The Random Forest model trained only on laboratory biomarkers reached an AUROC of 0.59 ± 0.02. Unimodal three-dimensional convolutional neural networks using SR-PET or CT performed worse (AUROC 0.42 ± 0.03 and 0.54 ± 0.01, respectively). A multimodal fusion model integrating laboratory values, SR-PET, and CT—augmented with a pretrained CT branch—achieved the best results (AUROC 0.72 ± 0.01, AUPRC 0.80 ± 0.01). Explainability analyses provided insights into model predictions, with explainability patterns in the fusion model appearing physiologically plausible and predominantly tumor-focused. Conclusions: Multimodal deep learning combining SR-PET, CT, and laboratory biomarkers outperformed unimodal approaches for PFS prediction after PRRT. Upon external validation, such models may support risk-adapted follow-up strategies. Full article

Currently, the B-cell response patterns induced by viral antigens in avian disease models and their detailed immunological characteristics still require comprehensive elucidation at the single-cell level. In this study, we employed single-cell sequencing (scRNA-seq) and B cell library technology to conduct an in-depth analysis of B cells in the spleens of mice with inactivated duck hepatitis A virus type 1 (DHAV-1) as model antigen. This study aimed to investigate the immunological characteristics of the virus antigen in the mouse model and characteristics of B-Cell Receptors. The results showed that the DHAV-1 group had distinct changes in splenic B cell subset counts, proportions, and intercellular communication. Additionally, an increased trend in communication strength between Gm26917+B and Gm11837+B cells was observed, with enriched expression of C-X-C motif chemokine ligand (CXCL) and lymphotoxin (LT) detected in the DHAV-1 group. Furthermore, the DHAV-1 group exhibited a prominent combination of the IGHV1 family and IGHV3-1/IGHJ3 in the heavy (H) chain variable region. Compared with the CK group (negative control group), the amino acid sequence length and diversity of the CDR3 region in the DHAV-1 group exhibited a decreasing trend. In summary, our findings characterize the immunological features of splenic B cells in mice after immunization with inactivated DHAV-1, and provide a preliminary characterization of DHAV-1-induced B cell transcriptional states and BCR repertoire features, generating testable hypotheses for subsequent mechanistic investigations of B cell-mediated immune responses to viral antigens. Full article

Historic buildings often exhibit high energy use intensity (EUI), while conservation constraints limit envelope retrofits, making it difficult to identify robust and actionable operational predictors. Using four in-use historic buildings in Shenyang, China, this study presents a pilot methodological demonstration with a controlled-comparability workflow consisting of two linked layers: (i) a Driver layer of intervenable operational variables and (ii) a Pathway layer of calibrated EnergyPlus heat-balance terms for physics-informed interpretation. Three importance approaches (Spearman, wrapper RFE with XGBoost, and Random Forest) are compared; rankings are fused via reciprocal rank fusion, and stability is tested using cross-period rolling validation across Top-K feature sets. After similarity screening, EUI variation is better explained by operational predictors and the corresponding simulated loss channels than by macro-scale structural heterogeneity. Infiltration-related indicators and envelope/infiltration loss components remain consistently prominent, while Spearman importance is less stable in the Pathway layer under seasonal switching and nonlinear coupling. A Top-10 subset provides a favorable accuracy–stability trade-off. The proposed Driver–Pathway mapping supports conservation-compatible prioritization hypotheses within a simulation-consistent interpretive framework; findings are associational and context dependent and should be validated through field measurements and experimental or quasi-experimental studies before prescriptive claims are made. Full article

The increasing deployment of Internet of Things (IoT) devices introduces significant security challenges, while privacy concerns limit centralized data aggregation for intrusion detection. Federated learning (FL) offers a decentralized alternative, yet the interaction between feature representation, model architecture, and data heterogeneity remains insufficiently understood in IoT malware detection. This study provides a controlled comparative analysis of centralized and federated learning, optionally using amino acid encoding, under IID and Non-IID conditions using a 10,000-sample subset of the CTU–IoT–Malware–Capture dataset. First, we evaluate raw tabular features versus amino acid-based feature encoding, followed by a lightweight multi-layer perceptron (2882 parameters) versus a deeper residual network (70,532 parameters), across binary and multi-class classification tasks. In the binary setting, centralized training achieved up to 98.6% accuracy, while federated IID training reached 98.6%, with differences within statistical variance. Under Non-IID conditions, performance decreased modestly (0.1–0.5 percentage points), and accuracy was consistently lower when using encoded features compared with raw features. The degradation is smaller in deeper architectures and may offer improved stability under highly skewed federated conditions. In the four-class setting, the complex network achieved up to 97.8% accuracy with raw features, while amino acid encoding achieves up to 93.3%. The results show that federated learning can achieve performance comparable to centralized training under moderate heterogeneity, that lightweight architectures are sufficient for low-dimensional IoT traffic features, and that feature compression via amino acid encoding does not inherently mitigate Non-IID effects. These findings clarify the relative impact of representation, heterogeneity, and architectural capacity in practical FL-based IoT intrusion detection systems. Full article

Accurate tea-shoot detection in real tea gardens is essential for intelligent harvesting, yet dynamic illumination (low light, strong light, and shadows) can cause brightness/contrast fluctuations and feature distribution shifts, degrading detection stability and localization accuracy. This paper proposes LA-YOLO, a dynamic-light tea-shoot detector based on YOLOv11. First, we construct a dynamic-light benchmark dataset and a difficulty-stratified evaluation protocol with four single-light subsets (A–D) and a mixed-light subset (E). Second, we design LA-CSNorm, an input-side brightness-adaptive preprocessing module that applies gated enhancement to dark samples followed by channel-selective normalization to reduce illumination-induced drift. Third, we propose RECA, a residual efficient channel-attention module to enhance discriminative channels and improve localization stability. Ablation studies show that LA-CSNorm and RECA provide complementary gains, and their combination improves the YOLOv11 baseline to 0.831 mAP@0.5 and 0.621 mAP@0.5:0.95, with only 0.01 M additional parameters. On the mixed-light subset E, LA-YOLO achieves 0.816 mAP@0.5 and 0.613 mAP@0.5:0.95, and consistently outperforms mainstream YOLO variants (e.g., YOLOv11m) under dynamic lighting conditions. These results demonstrate that LA-YOLO offers a robust and deployment-friendly solution for tea-shoot detection in complex natural illumination. Full article

Background: Transcatheter tricuspid edge-to-edge repair (T-TEER) is the most widely used treatment option for patients with tricuspid regurgitation (TR). In real-world practice, a substantial proportion of referred patients are not eligible for T-TEER or do not achieve an adequate early TR reduction and may therefore require alternative transcatheter tricuspid valve interventions (TTVI)—orthotopic or heterotopic tricuspid valve implantation. The aim of the study was to characterize patients with severe TR referred for transcatheter treatment, and identify patients in whom alternative TTVI strategies may be required. Methods: The CAPTURE Study (NCT 06838611) enrolls consecutive patients referred for TR treatment. All patients undergo clinical and echocardiographic assessment to determine eligibility for T-TEER. Candidates for alternative TTVI strategies were defined as patients disqualified from T-TEER due to anatomical ineligibility or those with unsuccessful T-TEER, defined as next-day TTE showing TR more than moderate. This pilot analysis includes patients enrolled from November 2023 to December 2024. Results: 147 patients were enrolled, 77 (52.4%) patients were qualified for T-TEER and the procedure was performed in 71 (48.3%) patients, with successful TR reduction in 55 cases (77.5% of treated patients); a subset of 34 patients (23.1%) was identified as potential candidates for alternative TTVI strategies. These patients exhibited more advanced TR (torrential TR 76.5% vs. 18.2%; p < 0.001) and right heart failure symptoms (ascites 44.1% vs. 12.7%; p < 0.001). Additionally, they had significantly higher bilirubin concentration (1.09 [1.20] mg/dL vs. 0.61 [0.42] mg/dL; p = 0.003), lower hemoglobin level (11.8 [1.7] g/dL vs. 12.3 [1.7] g/dL; p = 0.017) and platelet count (161.0 [51.0] × 109/L vs. 183.0 [79.0] × 109/L; p = 0.015), suggesting an increased bleeding risk. Conclusions: In this preliminary single-center real-world cohort, approximately half of the patients with severe TR were eligible for T-TEER, whereas more than 20% emerged as potential candidates for alternative TTVI strategies. This subgroup was characterized by more advanced right-sided remodeling and laboratory features suggestive of hepatic dysfunction and increased bleeding risk, which may have important implications for Heart Team decision-making and procedural planning. Full article

Attribute reduction, also referred to as feature selection, focuses on seeking a minimal attribute subset and is an important topic in rough set theory (RST). Classical rough set-based attribute reduction is restricted to decision systems with a single label level and fails to obtain attribute reducts across hierarchical label levels, leading to low computational efficiency. To tackle this issue, this paper proposes a novel method to derive bidirectional reducts across label levels. We first establish the reduction relationship from the coarse label level to the fine label level by analyzing the relationships among decision classes at different label levels. Correspondingly, we construct the reduction relationship from the fine label level to the coarse label level by investigating the connections between positive regions across different label levels. Based on these relationships, two efficient attribute reduction algorithms are developed, which can rapidly compute a reduct at one label level from the reduct at another level. Experimental results on twelve UCI datasets demonstrate that the proposed algorithms require less running time than the other four methods while still achieving high classification accuracy. Full article

Background: Multidrug-resistant (MDR) Escherichia coli in intensive pig production represents a persistent animal health and One Health concern. Here, we integrated quantitative phenotypic susceptibility data with whole-genome sequencing (WGS) to characterize the resistome and its inferred genomic context (chromosomal vs. plasmid-predicted contigs and mobile genetic element (MGE)-proximal regions) in swine-associated MDR E. coli from Hungary. Methods: A total of 203 E. coli isolates from large-scale pig farms were tested by broth microdilution. Based on resistance-oriented screening from an extended-spectrum β-lactamase (ESBL)-screen-positive pool, 116 isolates were subjected to whole-genome sequencing (WGS) as a resistance-enriched subset. Resistance determinants were annotated using the Comprehensive Antibiotic Resistance Database (CARD). Results: Resistance-oriented screening indicated frequent β-lactamase activity and ESBL screening positivity (110/203 and 127/203 isolates, respectively), consistent with strong antimicrobial selection pressure in the source population. Across the full phenotypic panel, 78/203 isolates (38.4%) met the MDR definition (non-susceptible to ≥3 antimicrobial classes), with marked between-farm variation (p < 0.001) but no age-group effect (p = 0.75). Non-β-lactam minimum inhibitory concentration (MIC) distributions showed pronounced, site-dependent high-MIC “tails”, most notably for tetracyclines, trimethoprim–sulfamethoxazole, fluoroquinolones, and colistin. In the WGS cohort (n = 116), we detected 82 distinct resistance determinants (5433 total occurrences), featuring a conserved chromosomal backbone enriched for intrinsic multidrug resistance components and lipid A modification pathways, alongside common plasmid- and MGE-associated acquired ARG modules involving tetracycline (tetA/tetB), sulfonamide/trimethoprim (sul/dfrA), aminoglycoside-modifying enzymes, and phenicol determinants (floR/cat). High-priority mobile determinants were rare but present, including mcr-1 (3/116; plasmid-associated) and plasmid-mediated quinolone resistance qnrB5 (2/116). Conclusions: Importantly, mobility/context inferences are restricted to this ESBL-screen-enriched WGS subset. Swine-associated E. coli from Hungarian large-scale farms harbors complex resistance architectures shaped by co-selection of mobile ARG modules on top of a pervasive chromosomal resistance backbone. Mobility-aware surveillance and stewardship are warranted to mitigate dissemination risks at the animal–environment–human interface. Full article

Competitive eSports impose substantial cognitive workload, yet performance evaluation still emphasizes post-match statistics without considering players’ cognitive states. We reviewed 30 papers that recorded physiological signals using sensors and utilized machine learning (ML) for predicting cognitive states and/or game performance. Findings showed that cardiovascular monitoring (heart rate variability/HRV) was the most prevalent modality (20/30 studies), followed by oculometry (10), electrodermal activity/EDA (9), and electroencephalogram/EEG (5); however, no standardized protocols (device/pre-processing/feature subset) were observed across HRV studies despite it being the most common measure. The best outcomes per construct (measure, accuracy) were: mental workload (pupillometry, ~82%), stress/arousal (EDA, p < 0.001), cognitive fatigue (pupil diameter/EEG, ~88%), expertise (EEG, ~92%), and tilt (EDA/HRV/eye-tracking, ~82–87%). Notably, current studies used small samples and were gender-imbalanced, while ML studies often lacked cross-validation. Only 2 of 30 studies examined flow state—a mental state of optimal performance characterized by total immersion and effortless execution—and interestingly, HRV showed decreases during stress/workload but increases during flow, suggesting context-dependent autonomic regulation. To address this gap, a new framework for flow detection is presented. This review will be of interest to game developers, eSports players, and coaches, and the reported findings may help towards improving player experience and game performance. Full article

This study developed an interpretable framework, RFE-SHAP, designed for yarn quality prediction. It integrates Recursive Feature Elimination (RFE) with SHapley Additive exPlanations (SHAP) theory to refine feature selection and mitigate data redundancy in small-sample environments. With Support Vector Regression (SVR) serving as the foundational evaluator, the RFE process iteratively identifies critical variables. Distinct from conventional methods, our approach employs SHAP values to quantify both the primary effects of individual features and the complex synergistic interactions among variables. This yields a transparent and intuitive strategy for identifying optimal feature subsets for two key quality indicators: yarn strength and hairiness H-value. To assess performance, a comparative analysis was performed between the traditional SVR-RFE method and the proposed RFE-SHAP method, using both as inputs for a Back-Propagation Artificial Neural Network (BP-ANN). The experimental results based on authentic production data demonstrate that the RFE-SHAP-BP model significantly enhances prediction reliability. Notably, compared to the baseline SVR-RFE-BP model, the proposed approach reduced the Mean Absolute Percentage Error (MAPE) by 0.73 and 1.01 percentage points for yarn strength and hairiness H-value, respectively. The final MAPE values reached 2.10% and 2.78%, confirming the model’s superior precision. These findings indicate that the RFE-SHAP method is highly feasible and effectively elevates prediction performance in data-limited industrial scenarios. Full article

Recent studies on EEG-based automated depression detection have primarily depended on complex deep learning models. While these methods improve classification performance, their practical application is limited by high computational complexity, challenging training processes, and poor interpretability. This paper proposes an efficient method for depression recognition, which extracts multi-domain features from preprocessed EEG signals and selects the most discriminative feature subset by integrating the rapid preliminary screening capability of RankSearch with the interactive optimization ability of the Genetic Algorithm (GA). Our approach first eliminates redundant features efficiently through RankSearch, then deeply explores inter-feature relationships via GA, significantly enhancing classification performance while maintaining feature-level interpretability. Using the optimized feature subset, we evaluate performance with multiple machine learning classifiers (Decision Tree, KNN, Random Forest, SVM, XGBoost). Experiments on the public HUSM dataset demonstrate superior performance under rigorous cross-validation (accuracy = 95.08%, sensitivity = 95.99%, specificity = 94.30%, F1-score = 95%, AUC = 0.9514), with feature importance analysis further confirming interpretability. Compared to existing models, our method achieves lower computational complexity and higher clinical practicality, offering a more efficient technical solution for objective depression diagnosis. Full article

Feature selection is essential for improving classification performance and reducing overfitting in high-dimensional learning tasks. However, conventional importance-based methods often suffer from instability, model bias, and sensitivity to threshold settings. To address these limitations, we propose EFSHB (Ensemble Feature Selection using Hierarchical Binning), a hybrid ensemble framework that integrates importance-based sorting, bin-level greedy evaluation, iterative hierarchical refinement, and union-based integration of model-wise selected features. At each iteration, five tree-based models independently perform bin-wise greedy selection, and their selected subsets are merged through a union operation to form the feature set for the next iteration. This iterative process progressively refines the feature space while mitigating model-specific bias and promoting robust predictive performance across heterogeneous models. EFSHB was evaluated on nine high-dimensional benchmark datasets, including biomedical gene-expression, synthetic, proteomics, and speech-feature data. Across all datasets, EFSHB achieved the highest or near-highest classification accuracy, outperforming traditional Greedy Feature Selection (GFS), binning-based GFS (GFSB), and hierarchical binning GFS (GFSHB). On average, EFSHB improved accuracy for all classifiers, achieving mean gains of 14.0% over GFS and 13.3% over GFSHB. EFSHB also provided balanced feature reduction by avoiding excessive feature retention while preserving complementary informative features identified across models. In terms of computational efficiency, EFSHB reduced average feature selection time from 266 min (GFS) to 11 min, corresponding to a 24-fold speed-up. These results demonstrate that EFSHB achieves robust predictive performance and high computational efficiency, making it suitable for diverse high-dimensional applications. Full article

A major limitation in studying gastroesophageal adenocarcinoma (GEA) has been the lack of reliable models that represent the disease’s complexity. We present lessons learned from a comprehensive large-scale biobanking effort combining traditional sample collection with several in vitro models, including 3-dimensional patient-derived organoids (PDOs), 2-dimensional cancer-associated fibroblasts (CAFs), tumor-infiltrating lymphocytes (TILs), and/or in vivo xenografts. This initiative started in 2018, integrating multiple advanced ex vivo models such as PDOs, patient-derived xenografts (PDXs), and organoids (PDXOs). This unique resource now includes tumor avatars from over 380 consented patients, making it the world’s largest living GEA biobank. We achieved a >90% success rate in creating per-patient models, including 227 tumor-derived and 203 neighboring normal PDOs. These organoids accurately mirror key features of the original tumors, such as their histology (e.g., microsatellite instability), mutations, and drug response across treatment points. Notably, PDOs can predict individual patient responses to chemotherapy within five weeks, underscoring their clinical relevance. Furthermore, high-throughput drug screening on PDO subsets with known genetic landscapes generates personalized chemosensitivity profiles for 22 drugs. Through a process of continued refinement of culture techniques and tumor sampling approach, our large-scale comprehensive collection of GEA avatars represents a unique and valuable preclinical experimental resource for precision oncology. Full article

Polycystic ovary syndrome (PCOS) is associated with reproductive, metabolic, and inflammatory disturbances. Alterations in T-cell subpopulations—particularly increased T helper 17 cells (Th17) and decreased regulatory T cells (Treg)—have been reported in PCOS; however, data on normoandrogenic phenotype D remain limited. Hypoxia-inducible factor 1α (HIF-1α), a key regulator of hypoxic response, also influences immune and metabolic processes and may affect the Treg/Th17 balance. To assess Treg and Th17 abundance, HIF-1α expression within these cells, and their ratios in women with phenotype D PCOS compared with healthy controls. The study included 49 women with phenotype D PCOS and 40 controls comparable in terms of age and BMI. Anthropometric, hormonal, metabolic, and inflammatory parameters were evaluated. Peripheral T-cell subsets and intracellular HIF-1α expression were analyzed by multiparameter flow cytometry. Absolute numbers of Treg and Th17 cells did not differ between groups. However, PCOS patients showed significantly higher Treg/Th17 and HIF-1α-positive Treg/HIF-1α-positive Th17 ratios. HIF-1α-positive Treg cells correlated positively with adiposity and insulin resistance markers; however, after False Discovery Rate (FDR) correction, correlations no longer remained statistically significant. Despite normoandrogenemia, PCOS patients exhibited higher hs-CRP levels. Phenotype D PCOS is characterized by altered immune cell ratios rather than absolute T-cell differences, suggesting distinct immunological features and persistent low-grade inflammation. Full article