Search Results (9,126)

The in-service assessment of marine propulsion engines requires more than nominal rating comparison because operating severity is shaped by propeller demand, resistance growth, air-path response, and thermal state. This study develops a quantitative benchmarking method for the regime-dependent performance assessment of a low-speed two-stroke Wärtsilä 6RT-flex58T-D engine installed on a 31,000 DWT multi-purpose container vessel. The method integrates certified sea-trial measurements, endurance-test records, manufacturer load-diagram constraints, and a 15% service-margin projection within one reference framework. Three representative regimes are evaluated: a measured light-running baseline (SR1), a measured thermally stabilised sustained regime (SR2), and a projected heavy-running regime derived from the baseline using a 15% sea-margin assumption (R2). Comparison is performed using indicators of operating-point position, shaft torque, propeller-law consistency, selected air-path and thermal variables, load-diagram proximity, and corrected specific fuel oil consumption where available. The SR1 baseline followed the fitted propeller law with deviations not exceeding 1.18%, confirming a coherent light-running reference. In SR2, corrected SFOC decreased from 174.4 to 172.0 g/kWh, while the exhaust temperature before turbine increased from 359 °C to 435 °C, and the corresponding thermal margin decreased from 156 °C to 80 °C. Under the +15% service-margin projection, the required shaft power at the 100% trial point increased from 12,046.0 to 13,852.9 kW, exceeding the 13,560 kW installation MCR by 2.2%, with corresponding 15% increases in torque and BMEP. These results demonstrate that measured baseline operation, sustained-load severity, and projected heavy-running demand can be distinguished quantitatively within one installation-specific load-diagram-based benchmarking framework. Full article

Despite significant advances, multimodal sentiment analysis still faces critical challenges in modeling complex cross-modal interactions and extracting discriminative sentiment features. To address these limitations, this paper proposes a hierarchical multimodal sentiment analysis framework. Specifically, a cross-modal feature enhancement module is first introduced to capture deep correlations among textual, visual, and acoustic modalities via cross-attention mechanisms, thereby obtaining context-aware fused representations. Subsequently, an attention-gated feature disentanglement approach is employed to effectively separate sentiment-relevant information from content-specific features within the fused representations; an independence loss is further imposed to enforce orthogonality between these two feature subsets, thereby mitigating noise induced by repetitive visual frames and textual stop words. Finally, all disentangled features are integrated to facilitate high-level sentiment reasoning through a multimodal logical inference module, where supervised contrastive loss is incorporated to enhance the discriminability of sentiment expressions. Extensive experiments conducted on two public benchmarks, CMU-MOSI and CMU-MOSEI, demonstrate that the proposed framework achieves improvements of 2–6% across multiple evaluation metrics compared with state-of-the-art methods. Full article

Open-pit lignite mining causes significant environmental alterations, particularly through the removal of soil deposits and the creation of external dumps, which necessitate effective reclamation to restore landscape structures. This study evaluates the potential of using multi-temporal remote sensing data to assess the effectiveness of forest reclamation on selected external dumps of the Adamów, Bełchatów, and Turów Lignite Mines in Poland. Using Landsat imagery spanning five decades from 1976 to 2023, the study monitors vegetation development through the Normalized Difference Vegetation Index (NDVI) and the Soil-Adjusted Vegetation Index (SAVI). Reclaimed forest stands were compared against undisturbed reference forests within a 30 km buffer zone, with recovery defined as achieving 95% of the reference values. The results indicate that most studied sites reached a state of recovery, with success closely linked to the specific reclamation measures implemented and the age of the forest stands. Notably, the Adamów mine, which utilized Bender’s target species method, demonstrated rapid results, achieving high similarity to reference forests early in the analyzed period. In contrast, recovery in Bełchatów and Turów was more gradual, following trajectories influenced by pioneer and biodynamic afforestation methods. Ultimately, the study confirms that remote sensing is a highly efficient tool for monitoring extensive post-mining areas over long periods, providing a general assessment of biological restoration success. Full article

Accurate rice mapping is critical for grain yield estimation and food security, yet traditional methods often struggle with asynchronous data quality and the inherent statistical gap between SAR and optical signals. To bridge this gap, we propose MMDFRNet, a novel multi-modal deep learning framework that synergistically integrates Sentinel-1 SAR and Sentinel-2 optical imagery. Unlike conventional static fusion approaches, MMDFRNet features a dual-stream modality-specific encoder architecture designed to decouple structural backscattering signals from spectral reflectance. Central to this framework is the multi-modal feature fusion (MMF) module, which employs an adaptive attention mechanism to dynamically align and recalibrate features based on their reliability, effectively mitigating noise from compromised modalities. Additionally, a multi-scale feature fusion (MSF) module is incorporated to coordinate hierarchical semantic information, enhancing boundary delineation in fragmented landscapes. Extensive experiments conducted across multiple study areas in China demonstrate the superiority of MMDFRNet. The model achieves a Precision of 0.9234, an IoU of 0.8612, and an F1-score of 0.9252. Notably, it consistently outperforms state-of-the-art benchmarks (e.g., UNetFormer, STMA, and CCRNet) by margins of up to 11.72% (Precision) and 7.39% (IoU) compared to classic baselines. Furthermore, rigorous ablation studies and degradation analyses confirm the model’s robustness, verifying its ability to transform the degradation paradox into a performance booster through pixel-wise adaptive alignment. Consequently, MMDFRNet offers a promising solution for precise rice area statistics and long-term monitoring in complex agricultural landscapes. Full article

High-performance medical segmentation models are often benchmarked on high-end GPUs. Such benchmarks do not provide useful performance insights for point-of-care low-end devices. This work, firstly, posits that to achieve improved clinical adoption of AI-powered segmentation models, especially in reduced manpower settings like rural hospitals, we need benchmarks that provide actionable insights on the degree to which high-performance models address five deployment constraints viz: resource-effectiveness for low-end computing devices, clinically acceptable accuracy, clinically compatible execution times, localization of user data, and user-based finetuning. In this work, five state-of-the-art foundation segmentation models and one target-specific model were systematically evaluated on three multi-organ medical datasets. Furthermore, the best-ranking foundation model and target-specific model were benchmarked on three low-end devices. Our findings show that lightweight foundation models provided the best performance trade-off and are easily user-fine-tuned on custom datasets. Target-specific models provide high accuracy out-of-the-box, but may require significant optimisation to deliver comparably fast execution times and user-based finetuning on low-end devices. The methods and results from this research provide actionable insights on high-performance medical segmentation models for low-end computing devices, as a necessary step towards improved adoption in resource-limited clinical settings. Full article

Background: Post-approval manufacturing changes for biologics require rigorous comparability assessments to ensure uninterrupted quality and clinical performance. CMAB007 (Aomaishu^®), a China-approved (2023) omalizumab biosimilar, underwent process enhancements—including media optimization and anion-exchange chromatography substitution—yielding a 5-fold increase in production without altering the host cell line. Methods: A novel three-arm tiered strategy was adopted to compare post-change CMAB007, pre-change CMAB007, and reference (Xolair^®) products. Critical quality attributes (CQAs) were classified into tiers based on risk impact, with tier-specific acceptance criteria. Comprehensive analytics assessed structure, post-translational modifications, purity/impurities, activity, and Fc-mediated functions. Forced degradation (lyophilized/reconstituted states) and accelerated stability studies were evaluated. Based on the high degree of CMC similarity and to prevent “biological drift”, the pharmacokinetic (PK) and safety comparability of the post-change CMAB007 versus the reference product (Xolair^®) was confirmed in a randomized, double-blind, two-arm study in healthy males (N = 114; single 150 mg subcutaneous administration). The pre-change product was not included in this clinical PK study. Results: Post-change CMAB007 exhibited analytical similarity within tiered acceptance criteria for all CQAs. Stability studies confirmed enhanced robustness under stress conditions. PK equivalence was demonstrated for AUC_0–inf (GMR: 99.82%; 90% CI: 91.46~108.94%), AUC_0–t (99.54%; 91.40~108.41%), and C_max (101.88%; 95.21~109.01%). Immunogenicity (ADA incidence: 10.5% vs. 12.5%, p = 0.742) and safety profiles were comparable. Conclusions: This study pioneers a tiered three-arm comparability strategy for post-approval changes, integrating advanced analytics, risk-based quality assessment, and clinical validation. The approach mitigates “biological drift” risks, ensuring biosimilar quality, efficacy, and safety while enabling sustainable production scalability. Full article

Reliable and timely estimation of air pollution exposure at high spatial and temporal resolution remains challenging in complex urban environments, where pollutant concentrations vary due to traffic emissions, urban morphology, and meteorological conditions. This study presents a physics-informed machine learning framework for near-real-time estimation of NO₂ concentrations at fine spatial scales. The approach combines a limited set of steady-state computational fluid dynamics (CFD) simulations with operational meteorological and air-quality data. CFD simulations under specific wind directions are first used to characterize site-specific dispersion patterns. These outputs are then scaled using hourly meteorological observations to generate physics-based concentration descriptors. A machine learning predictor, implemented using Random Forest and Extreme Gradient Boosting, is trained to refine these estimates by incorporating additional environmental and observational features. The method is applied to a 1 km × 1 km urban district in Antwerp, Belgium, within the FAIRMODE intercomparison framework. Validation against measurements from 105 passive samples collected over one month shows substantial improvement compared to standalone dispersion modeling, with coefficients of determination up to R² = 0.965 and reduced bias across locations. These findings demonstrate that integrating physical modeling with machine learning enables accurate and computationally efficient high-resolution exposure assessment in urban settings. Full article

Background: Language impairment is a core feature of Major Neurocognitive Disorder (MND), yet the domain-specific relationship between language functioning and everyday functional status remains insufficiently characterized. Methods: We conducted a retrospective observational study in 125 older adults diagnosed with MND according to DSM-5 criteria with mild-to-moderate cognitive impairment measured with Mini-Mental State Examination (MMSE). Language performance was assessed using semantic, phonemic verbal fluency and confrontation naming. Functional status was evaluated using basic (BADL) and instrumental activities of daily living (IADL). Ordinal logistic regression models examined associations between language domains and functional outcomes, adjusting for global cognitive status (MMSE), demographic variables, multimorbidity, and depressive symptoms. Model fit was evaluated using the Akaike Information Criterion. Results: Semantic fluency emerged as the best-performing predictor of BADL across all hierarchical models, remaining statistically significant after full adjustment for MMSE and clinical covariates (β ≈ 0.60, p < 0.05). Phonemic fluency showed the most robust association with IADL, with a stable effect across models, reaching a trend toward statistical significance in the fully adjusted analyses (β ≈ 0.22–0.27, p = 0.069). Naming ability did not influence functional outcomes. All observed associations persisted after controlling for MMSE, demographic variables, multimorbidity, and depressive symptoms. Conclusions: Language abilities showed differential associations across language domains with functional status in this sample of patients with MND. Semantic fluency was associated with basic self-care, while phonemic fluency showed a trend toward association with instrumental daily activities. These relationships remained observable after adjustment for global cognitive impairment, suggesting verbal fluency as a potentially sensitive marker of functional vulnerability. Full article

Disease staging is a critical component of clinical diagnosis, treatment, and prognosis assessment. However, structured clinical data typically exhibit high-dimensional, nonlinear feature interactions; stage-specific dominant features; and threshold-based discontinuities. These characteristics make it challenging for a single model to achieve both global feature modeling capability and local discriminative power, thereby limiting further improvements in prediction accuracy. To address this limitation, we propose a novel deep ensemble learning framework, ETGB-SEF (Entmax-TabNet Gradient Boosting Stacked Ensemble Framework), for multiclass disease staging. First, at the base model level, Entmax-1.5 replaces Sparsemax in TabNet, thereby enabling an adjustable sparse feature selection mechanism that enhances the ability to model weakly correlated clinical features while preserving interpretability. Second, at the model-fusion level, a stacked ensemble architecture in the probability space is developed. This architecture integrates the modified TabNet with Gradient Boosting Decision Trees (GBDT) in a complementary way, enabling the former to capture global nonlinear semantic dependencies while the latter captures threshold-based discriminative boundaries among clinical features. Extensive experiments on real-world datasets demonstrate that the proposed method consistently outperforms existing state-of-the-art approaches. Full article

The design of stretchable energy harvesting systems entails complex multiphysics coupling between electromagnetic and mechanical domains, typically requiring engineers to proficiently use disparate simulation tools and optimization algorithms. This steep learning curve, combined with the absence of integrated workflows, poses a substantial obstacle to efficient design. To overcome these challenges, we present StretchCopilot, a multi-agent collaborative framework driven by Large Language Models (LLMs) for the generative design of stretchable radio frequency (RF) energy harvesters operating in the 2.45 GHz band. In contrast to conventional approaches dependent on manual iteration or isolated algorithmic methods, our framework utilizes a graph-based state machine architecture (LangGraph) to coordinate specialized agents. It interprets high-level user instructions, such as “design a robust energy harvester capable of withstanding 15% strain”, and autonomously manages domain-specific solvers, including inverse design networks and rectifier circuit synthesis tools, through a unified interface. Experimental evaluations indicate that the framework effectively streamlines the design workflow, allowing users to produce desired rectenna (rectifying antenna) systems via natural language interactions. Case studies confirm that, once the underlying surrogate models are fully trained, the proposed approach compresses the marginal design time from several hours to within minutes, while ensuring consistent energy harvesting performance under mechanical deformation. Full article

Differentiating Alzheimer’s disease (AD) from frontotemporal dementia (FTD) using resting-state electroencephalography (EEG) remains clinically challenging because of their overlapping electrophysiological characteristics. Although EEG suits large-scale dementia screening, current method often overestimates performance because of epoch-level data leakage and multiclass feature competition in unified models. We propose a task-decoupled, two-stage hierarchical deep learning framework utilizing multiband EEG microstate dynamics. Continuous microstate sequences, modeled via Hungarian matching to preserve fine-grained temporal information, are processed using a normalizer-free 1D convolutional neural network (1D-CNN-NFNet) integrated with multi-head attention. By decoupling the workflow, Stage 1 performs generalized dementia screening using alpha and delta microstates, achieving an area under the curve (AUC) of 0.851. Stage 2 disentangles AD from FTD using delta and theta dynamics, yielding an AD-locking specificity of 86.1%. Evaluated under a strict subject-level leave-one-subject-out (LOSO) cross-validation protocol, the two-stage framework achieved 63.9% balanced accuracy, outperforming the single-stage baseline (55.4%) with a negligible inference latency of 0.733 ms. Furthermore, attention-based interpretability analysis links frequency-specific microstate alterations to underlying cortical disconnection syndromes. These results demonstrate that the framework provides a reproducible and interpretable auxiliary reference for dementia screening and subtyping in clinical neurology. Full article

Software-defined networking (SDN) provides a flexible solution to manage complex networks on demand by centralized control and programmability. However, efficiently optimizing network configurations to achieve load balance and improve service quality remains challenging. In this paper, we propose a novel SDN network state awareness and dynamic routing optimization method, termed DGSFN-DR. Hereby, we leverage a Graph Attention Network (GAT) to model the spatial dependencies of the network topology for its link graph. Then, we employ a Recurrent Neural Network (RNN) to capture the temporal dependencies of link states, including the lagged temporal features induced by routing algorithms, to improve the prediction accuracy of future link states. Our algorithm dynamically adjusts routing strategies to optimize network performance according to the predicted link weights with the dual graph spatiotemporal fusion prediction network (DG-SFN). Experimental results demonstrate that our DGSFN-DR outperforms other methods in various network traffic intensities and topologies. Specifically, it achieves improvements of $4\%$ to $15\%$ in latency, jitter, packet loss, and available bandwidth. In particular, the DGSFN-DR exhibits superior adaptability and optimization potential under high traffic loads and complex network topologies. This work expands dynamic routing optimization theory for SDN and new insights for practical network management. Full article

Background/Objectives: Mental health service utilization gaps remain a persistent global public health challenge. Among the 61.5 million adults with any mental illness in the United States, nearly half went without treatment in the past year, and dropout rates from outpatient services among those who do enter care range from 19.7% to 30.8%. Only 30 to 60% of individuals with lifetime mental illness are in active recovery at any given time. Existing theoretical frameworks, including Andersen’s Behavioral Model, the Health Belief Model, and the COM-B framework, each address isolated phases of the care continuum but offer no unified structure for understanding the complete, sequential journey from first contact through sustained recovery. This article introduces the Access, Initiation, Engagement, Retention, and Recovery (AIERR) model to address this theoretical gap. Methods: A conceptual review was conducted following Hulland’s framework for theory development through narrative synthesis. Literature was identified through targeted searches in PubMed, PsycINFO, and Google Scholar, prioritizing peer-reviewed empirical studies, systematic reviews, and foundational theoretical frameworks. Sources were assigned to AIERR stages using predefined decision rules corresponding to each phase’s defining characteristics. Results: AIERR maps five sequential, interconnected stages: Access (structural, cultural, and systemic conditions enabling service reach), Initiation (the transition from provider identification to first appointment attendance), Engagement (active and meaningful treatment participation), Retention (sustained continuity of care), and Recovery (long-term reclamation of life quality and community belonging). For each stage, the framework identifies individual-level and structural-level barriers, facilitating conditions, and targeted intervention points. Conclusions: AIERR advances mental health services theory by unifying previously siloed frameworks, establishing stage-specificity as a core theoretical principle, and reorienting research and intervention strategy toward the upstream structural conditions that produce downstream utilization failures. These theoretical contributions require empirical testing to confirm. Implications for health equity research, clinical practice, and health systems design are discussed. Full article

Background: Recovery following total joint arthroplasty varies substantially among patients, and psychological factors may partly account for this variability. Although anxiety and depression have been widely investigated, the specific contribution of preoperative surgical fear to postoperative pain and quality of recovery remains unclear. This study aimed to examine the association between preoperative surgical fear and postoperative pain intensity and quality of recovery. Methods: This prospective, hospital-based observational study enrolled 89 patients undergoing primary total knee or hip arthroplasty. Preoperative surgical fear was measured using the Surgical Fear Questionnaire (SFQ). Pain intensity was assessed with the Numeric Rating Scale (NRS) preoperatively and at three postoperative time points. Recovery quality at 24 h was evaluated using the Quality of Recovery-40 (QoR-40). Pearson correlation and multiple linear regression analyses were performed to evaluate associations and identify variables independently associated with recovery outcomes, controlling for potential confounders, including age, sex, ASA physical status, and type of surgery. Results: The mean SFQ score was 26.62 ± 15.19, and the mean QoR-40 score was 157.63 ± 16.66. Surgical fear was moderately and negatively correlated with overall recovery quality (r = −0.546, p < 0.001). In multiple linear regression analysis, surgical fear was most strongly associated with poorer overall recovery quality (β = −0.563, p < 0.001), within a model explaining 30.3% of the variance (adjusted R² = 0.303). At the subscale level, surgical fear was significantly associated with emotional state, pain, physical comfort, and perceived support. Pain intensity at 12 h postoperatively was significantly associated with reduced physical independence (β = −0.218, p = 0.038). Pain intensity peaked at 12 h postoperatively (p < 0.001). Conclusions: Higher levels of preoperative surgical fear are associated with poorer quality of recovery following total joint arthroplasty. These findings highlight surgical fear as a potentially relevant perioperative factor and support the integration of routine psychological assessment into perioperative care pathways in relation to early postoperative recovery outcomes. From a clinical perspective, early identification of patients with high surgical fear may facilitate targeted perioperative counseling and supportive interventions by healthcare professionals, potentially improving recovery outcomes. Full article

Background and Objective: Previous studies have shown that muscle mass and visceral fat are interrelated and affect metabolic health. However, there is limited research exploring machine learning (ML) models that can help us understand the relationship between muscle mass and the risk of adiposity in the adult population. The objective of this study was to identify predictors of obesity on the basis of data from 13,663 adults assessed via body composition analysis via optimal and interpretable ML algorithms. Methods: A cross-sectional design was used to analyze data from 13,663 adults, comprising men (n = 6877) and women (n = 6786). The variables were obtained via 8-point multifrequency BIA under standardized clinical protocols with an Inbody^® Model 770 device validated for the adult population. To illustrate the interaction between body composition components, a probability heatmap was generated on the basis of the values predicted from the logistic model. The decision boundary was defined via the metabolic risk probability gradient, allowing visualization of the two-dimensional transition between low- and high-risk states. Statistical processing and figure generation were performed via Python software v.3.10. Results: The evaluation of the 10 algorithms demonstrated exceptional predictive performance, with the multilayer perceptron (MLP) standing out as the superior model in both sexes. The AUC-ROC was 0.981 for men and 0.993 for women, with F1 scores of 0.912 and 0.969, respectively. Overall, systematically higher accuracy was observed in the female cohort, exceeding 95% accuracy in most models. Conclusions: Muscle mass has been shown to act as a metabolic mediator, modulating and reducing the risk associated with visceral adiposity. It also concludes that the use of ML algorithms, specifically neural networks, is a good model for analyzing the risk associated with excess visceral fat. Full article