Search Results (881)

Objective: To develop and validate an interpretable deep learning model based on the TabNet architecture for predicting doxorubicin-induced cardiotoxicity (DIC) in patients with breast cancer through integration of multidimensional clinical data. Methods: This retrospective study included 2034 patients who received doxorubicin-based chemotherapy at The Fourth Affiliated Hospital of Harbin Medical University between January 2021 and December 2023. Clinical, biochemical, electrocardiographic, and echocardiographic parameters were incorporated into six predictive algorithms: logistic regression, decision tree, random forest, gradient boosting machine, XGBoost, and TabNet. Model discrimination, calibration, and clinical utility were assessed using AUC, C-index, calibration plots, and decision curve analysis. Model interpretability was evaluated through attention-based feature importance and SHAP analysis. Results: TabNet achieved the best overall predictive performance, with an AUC of 0.86 and a C-index of 0.80 in the validation cohort, demonstrating superior discrimination, calibration, and generalization compared with all baseline models. Decision curve analysis confirmed its higher net clinical benefit across threshold probabilities. The model identified eight dominant predictors—cumulative anthracycline dose, LVEF, QTc interval, lactate dehydrogenase, creatinine, glucose, hypertension, and platelet count—that collectively reflected myocardial contractility, electrophysiological stability, and systemic metabolic stress. Correlation and clustering analyses revealed that high-risk patients exhibited concurrent QTc prolongation, metabolic disturbance, and LVEF decline, defining a distinct cardiometabolic injury phenotype. These findings highlight TabNet’s ability to uncover complex feature interactions while maintaining transparent and clinically interpretable outputs. Conclusions: The TabNet-based multidimensional model provides an accurate, stable, and interpretable tool for individualized prediction of doxorubicin-induced cardiotoxicity, supporting early intervention and precision management in breast cancer patients receiving anthracycline therapy. Full article

Background/Objectives: Early-stage diabetes often presents with subtle symptoms, making timely screening challenging. This study aimed to develop an interpretable and robust machine learning framework for early-stage diabetes risk prediction using integrated statistical and machine learning–based feature selection, and to evaluate its generalizability using real-world hospital data. Methods: A Union Feature Selection approach was constructed by combining logistic regression significance testing with ReliefF and MRMR feature importance scores. Five machine learning models—Decision Tree, Naïve Bayes, SVM, KNN, and Neural Network—were trained on the UCI Early Stage Diabetes dataset (N = 520) under multiple feature-selection scenarios. External validation was performed using retrospective hospital records from the University of Phayao (N = 60). Model performance was assessed using accuracy, precision, recall, and F1-score. Results: The union feature-selection approach identified four core predictors—polyuria, polydipsia, gender, and irritability—with additional secondary features providing only marginal improvements. Among the evaluated models, Naïve Bayes demonstrated the most stable external performance, achieving 85% test accuracy, balanced precision, recall, and F1-score, along with a moderate AUC of 0.838, indicating reliable discriminative ability in real-world hospital data. In contrast, SVM, KNN, and Neural Network models, despite exhibiting very high internal validation performance (>96%) under optimally selected ML features, showed marked performance decline during external validation, highlighting their sensitivity to distributional shifts between public and clinical datasets. Conclusions: The combined statistical–ML feature selection method improved interpretability and stability in early-stage diabetes prediction. Naïve Bayes demonstrated the strongest generalizability and is well suited for real-world screening applications. The findings support the use of integrated feature selection to develop efficient and clinically relevant risk assessment tools. Full article

Forests are a critical component of global carbon sequestration, biodiversity, and ecosystem services, making accurate mapping essential for long-term monitoring. In North Korea, limited field access, rugged topography, and inconsistent national statistics necessitate reliable remote sensing–based observation. However, frequent cloud contamination challenges the use of optical time-series imagery for forest monitoring. This study introduces a framework that integrates a ConvLSTM-based autoencoder into a U-Net segmentation model to improve tree cover classification from Sentinel-2 time-series data. The autoencoder was pretrained to reconstruct cloud-contaminated or missing observations using multi-octave Perlin-noise perturbations, providing standardized inputs that enhanced segmentation robustness under noisy conditions. Results show that tree cover accuracy exceeded 96% when all five time steps were available and remained stable (94–95%) even with one missing step. Accuracy declined below 90% with three missing steps but remained above 80%, enabling draft classifications under limited data. Confidence analysis further indicated that model certainty is a practical quality-control metric. Annual mapping for 2019–2024 showed a general increase in tree cover, aligning with reported afforestation efforts in North Korea. Taken together, the framework advances long-term monitoring, carbon accounting, and risk assessment in North Korea, while also enabling robust, region-adapted monitoring in cloud-prone, data-limited settings. Full article

Background and Objectives: The objective of this study is to assess the efficacy of a novel software risk score, PulsIn, in predicting cardiovascular diseases within an independent study conducted on subjects from the western region of Romania. Accurate prediction of cardiovascular events in hypertensive patients remains challenging when relying solely on traditional risk scores. This study proposes PulsIn, a composite risk score that integrates classical, echocardiographic, inflammatory, renal, and metabolic markers, combined with machine learning, to refine cardiovascular risk stratification. Materials and Methods: In a prospective cohort of 300 hypertensive adults without prior major cardiovascular events, we collected demographic and clinical data, standard risk factors, laboratory biomarkers (including homocysteine, paraoxonase-1 activity, microalbuminuria, and lipid profile), and advanced echocardiographic parameters (3D left ventricular ejection fraction, diastolic function, global longitudinal strain, and left atrial strain). PulsIn was constructed as an extended composite score and used as input to machine learning models (random forest, XGBoost, and other tree-based algorithms) to predict incident major cardiovascular events. Model performance was assessed by receiver operating characteristic curves, discrimination, calibration, and feature importance and compared with established risk scores (SCORE2, Framingham, QRISK, and others). Results: PulsIn-based models showed improved predictive performance compared with traditional scores, with XGBoost and random forest achieving area under the curve values up to approximately 0.85–0.88, versus 0.60–0.78 for conventional scores. Echocardiographic indices of subclinical cardiac damage, microalbuminuria, homocysteine, and paraoxonase-1 activity emerged as key predictors, particularly enhancing reclassification in patients at intermediate risk by traditional tools. Conclusions: The PulsIn composite risk score, integrating multimodal clinical, echocardiographic, and biomarker data within a machine learning framework, offers more accurate cardiovascular risk prediction than conventional algorithms in hypertensive patients. External validation in larger, independent, and more diverse populations is required before routine clinical implementation. Full article

Hepatitis viruses continue to pose major global health challenges, necessitating the development of reliable and well-characterized experimental models. Tree shrews are increasingly recognized as a valuable small animal model because of their natural susceptibility to hepatitis viruses and close phylogenetic relationship with primates. Recent identification of a high prevalence of natural hepacivirus infections in tree shrews underscores the urgent need for routine viral screening of the animals used in hepatitis studies. Undetected infections may confound experimental results, undermine data integrity, and pose risks to laboratory biosecurity. Integrating systematic screening and standardized reporting practices will minimize these risks, enhance reproducibility, and safeguard the integrity of research findings. Moreover, a consistent assessment of the infection status will enhance the translational potential of tree shrews for studying viral hepatitis pathogenesis and evaluating antiviral interventions. This opinion paper emphasizes that ensuring the virological status of tree shrews is not merely a procedural recommendation but also a methodological standard essential for advancing hepatitis virus research. Full article

The comprehensive evaluation of factors that increase the difficulty of autonomous driving in various complex traffic situations and diverse roadway geometries within living lab environments is of great interest, particularly in developing sustainable urban mobility systems. This study introduces a novel methodology for assessing autonomous driving vulnerabilities and identifying urban traffic segments susceptible to autonomous driving risks in mixed traffic situations where autonomous and manual vehicles coexist. A microscopic traffic simulation network that realistically represents conditions in a living lab demonstration area was used, and twelve safety indicators capturing longitudinal safety and vehicle interaction dynamics were employed to compute an integrated risk score (IRS). The promising weighting of each indicator was derived through decision tree method calibrated with real-world traffic accident data, allowing precise localization of vulnerability hotspots for autonomous driving. The analysis results indicate that an IRS-based hotspot was identified at an unsignalized intersection, with an IRS value of 0.845. In addition, analytical results were examined comprehensively from multiple perspectives to develop actionable improvement strategies that contribute to long-term sustainability, encompassing roadway and traffic facility enhancements, provision of infrastructure guidance information, autonomous vehicle route planning, and enforcement measures. Furthermore, this study categorized and analyzed the characteristics of high-risk road sections with similar geometric features to systematically derive effective traffic safety countermeasures. This research offers a systematic, practical framework for safety evaluation in autonomous driving living labs, delivering actionable guidelines to support infrastructure planning and validate sustainable autonomous mobility. Full article

Landslide risk assessment along high-altitude transportation corridors is critical for infrastructure resilience. This study presents an integrated framework combining Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) deformation data and machine learning (ML) to systematically identify and assess landslide susceptibility along the entire 245.5 km West Kunlun Highway. We first compiled a landslide inventory through visual interpretation and SBAS-InSAR analysis. Subsequently, fourteen causative factors were selected to construct and compare six ML models: random forest (RF), K-nearest neighbours (KNN), artificial neural network (ANN), gradient boosting decision trees (GBDT), support vector machine (SVM), and logistical regression (LR). Research findings indicate that along the Hotan–Kangziva Highway in the Western Kunlun Mountains, there exist 21 potential risk points for small-scale landslides, 12 for medium-scale landslides, and 5 for large-scale landslides, with hazard identification accuracy reaching 80%. The random forest model demonstrated outstanding performance, classifying areas with 5.10%, 4.55% and 4.96% probability as extremely high, high and medium susceptibility, respectively. This work provides a robust methodology and a high-accuracy assessment tool for landslide risk management in the data-scarce Western Kunlun Mountains. Full article

Artificial Intelligence (AI) is increasingly used to enhance project management practices, especially in risk analysis, where traditional tools often lack predictive capabilities. This study introduces an AI-based tool that supports project teams in identifying and interpreting risks through machine learning and integrated documentation features. A synthetic dataset of 5000 project instances was generated using deterministic rules across 27 input variables, enabling the training of multi-output Decision Tree and Random Forest models to predict risk type, impact, probability, and response strategy. Due to the rule-based structure of the dataset, both models achieved near-perfect classification performance, with Random Forest showing slightly better regression accuracy. These results validate the modelling pipeline but should not be interpreted as real-world predictive accuracy. The trained models were deployed within a web platform offering prediction visualization, automated PDF reporting, result storage, and access to a structured risk management plan template. Survey feedback highlights strong user interest in AI-assisted mitigation suggestions, dashboards, notifications, and mobile access. The findings demonstrate the potential of AI to improve proactive risk assessment and decision-making in project environments. Full article

Background: HER2-positive breast cancer patients receiving chemotherapy and targeted therapy (including anthracyclines and trastuzumab) face an elevated risk of cardiotoxicity, which can lead to long-term cardiovascular complications. Identifying predictive biomarkers is essential for early intervention. Circulating microRNAs (miRNAs), known regulators of gene expression and cardiovascular function, have emerged as potential indicators of cardiotoxicity. This study aims to evaluate the differential expression of circulating miRNAs in HER2-positive breast cancer patients undergoing chemotherapy and to assess their prognostic ability for therapy-induced cardiotoxicity using machine learning models. Methods: Forty-seven patients were assessed for cardiac toxicity at baseline and every 3 months, up to 15 months. Blood samples were collected at baseline. MiRNA expression profiling for 84 microRNAs was performed using the miRCURY LNA miRNA PCR Panel. Differential expression was calculated via the 2^−∆∆Ct method. The five most upregulated and five most downregulated miRNAs were further assessed using univariate logistic regression and receiver operating characteristic (ROC) analysis. Five machine learning models (Decision Tree, Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting Machine (GBM), k-Nearest Neighbors (KNN)) were developed to classify cardiotoxicity based on miRNA expression. Results: Forty-five miRNAs showed significant differential expression between cardiac toxic and non-toxic groups. ROC analysis identified hsa-miR-155-5p (AUC 0.76, p = 0.006) and hsa-miR-124-3p (AUC 0.75, p = 0.007) as the strongest predictors. kNN, SVM, and RF models demonstrated high prognostic accuracy. The decision tree model identified hsa-miR-17-5p and hsa-miR-185-5p as key classifiers. SVM and RF highlighted additional miRNAs associated with cardiotoxicity (SVM: hsa-miR-143-3p, hsa-miR-133b, hsa-miR-145-5p, hsa-miR-185-5p, hsa-miR-199a-5p, RF: hsa-miR-185-5p, hsa-miR-145-5p, hsa-miR-17-5p, hsa-miR-144-3p, and hsa-miR-133a-3p). Performance metrics revealed that SVM, kNN, and RF models outperformed the decision tree in overall prognostic accuracy. Pathway enrichment analysis of top-ranked miRNAs demonstrated significant involvement in apoptosis, p53, MAPK, and focal adhesion pathways, all known to be implicated in chemotherapy-induced cardiac stress and remodeling. Conclusions: Circulating miRNAs show promise as biomarkers for predicting cardiotoxicity in breast cancer patients. Machine learning approaches may enhance miRNA-based risk stratification, enabling personalized monitoring and early cardioprotective interventions. Full article

Feature selection is pivotal in enhancing the efficiency of credit scoring predictions, where misclassifications are critical because they can result in financial losses for lenders and exclusion of eligible borrowers. While traditional feature selection methods can improve accuracy and class separation, they often struggle to maintain consistent performance aligned with institutional preferences across datasets of varying size and imbalance. This study introduces a FastTree-Guided Genetic Algorithm (FT-GA) that combines gradient-boosted learning with evolutionary optimization to prioritize class separability and minimize false-risk exposure. In contrast to traditional approaches, FT-GA provides fine-grained search guidance by acknowledging that false positives and false negatives carry disproportionate consequences in high-stakes lending contexts. By embedding domain-specific weighting into its fitness function, FT-GA favors separability over raw accuracy, reflecting practical risk sensitivity in real credit decision settings. Experimental results show that FT-GA achieved similar or higher AUC values ranging from 76% to 92% while reducing the average feature set by 21% when compared with the strongest baseline techniques. It also demonstrated strong performance on small to moderately imbalanced datasets and more resilience on highly imbalanced ones. These findings indicate that FT-GA offers a risk-aware enhancement to automated credit assessment workflows, supporting lower operational risk for financial institutions while showing potential applicability to other high-stakes domains. Full article

Biological invasions are major threats to global biodiversity, and mapping their distribution is essential to prioritizing management efforts. The Pinaceae family (hereafter pines) includes invasive trees, particularly in Southern Hemisphere regions where they are non-native. These invasions can increase the severity of fires in wildland–urban interfaces (WUIs). We mapped pine invasion in the Bariloche WUI (≈150,000 ha, northwest Patagonia, Argentina) using supervised land cover classification of Sentinel-2 imagery with a Random Forest algorithm on Google Earth Engine, achieving 90% overall accuracy but underestimating the pine invasion area by about 25%. We then assessed in which main vegetation context pine invasions occurred relying on major vegetation units across the precipitation gradient of our study area. Invasions cover 2% of the study area, mainly in forests (61%), steppes (25.4%), and shrublands (13.4%). Most invaded areas (89.1%) are on private land; nearly 70% are on large properties (>10 ha), where state financial incentives could support removal. Another 13.5% occur on many small properties (<1 ha), where awareness campaigns could enable decentralized, low-effort control. Our land cover map can be developed further to integrate invasion dynamics, inform fire risk and behavior models, optimize management actions, and guide territorial planning. Overall, it provides a valuable tool for targeted, scale-appropriate strategies to mitigate ecological and fire-related impacts of invasive pines. Full article

Accurate fault risk assessment is essential for maintaining wind turbine reliability. Traditional failure modes and effects analysis (FMEA)-based approaches struggle to handle the fuzziness, uncertainty, and conflicting nature of multi-criteria evaluations, which may lead to delayed fault detection and increased maintenance risks. To address these limitations, this paper proposes an enhanced risk assessment framework that integrates subjective-objective weighting and fuzzy decision-making. First, a combined subjective–objective weighting (CSOW) model with adaptive fusion is developed by integrating the analytic hierarchy process (AHP) and the entropy weight method (EWM). The CSOW model optimizes the weighting of severity (S), occurrence (O), and detection (D) indicators by balancing expert knowledge and data-driven information. Second, a fuzzy decision-making model based on interval-valued intuitionistic fuzzy numbers and VIKOR (IVIFN-VIKOR) is established to represent expert evaluations and determine risk rankings. Notably, the overlap rate between the top 10 failure modes identified by the proposed method and a fault-tree-based Monte Carlo simulation incorporating mean time between failures (MTBF) and mean time to repair (MTTR) reaches 90%, substantially higher than other methods. This confirms the superior performance of the framework and provides enterprises with a systematic approach for risk assessment and maintenance planning. Full article

Subway stations are enclosed spaces with high passenger density and complex evacuation conditions. Fires in such environments can escalate rapidly and cause severe consequences. This study proposes a dynamic risk assessment model grounded in dual symmetries. The first symmetry is a balanced “Human–Machine–Environment–Management” analytical structure. The second is a coherent model transformation from a Fault Tree (FT) to a Bayesian Network (BN). Shuanggang Station on Nanchang Metro Line 1 serves as a case study. This work establishes a comprehensive evaluation system based on 4 first-level indicators of man–machine–environment–management, 9 secondary indicators, and 27 tertiary indicators. FT analysis identified 117 minimal cuts and 14 minimal paths, pinpointing core risk nodes such as flammable materials and oxidizers, electrical equipment overheating, and fire management deficiencies. The model was then symmetrically converted into a BN using GeNle Academic 4.1 software to support dynamic probability inference. The results show that prevention measures at Shuanggang Station reduce the fire occurrence probability from 0.000249 to 0.00007 (a 71.9% reduction). The probability importance of rescue escape routes is 0.00223. This indicates that the accessibility of rescue routes constitutes a highly sensitive hazard. The symmetric framework and modeling approach offer a scientific basis for targeted fire prevention, control, and evacuation management in the Nanchang Metro and similar stations. The findings support improvements in the safety and resilience of metro operations. Full article

Urban densification in post-socialist cities has drastically reduced open and green spaces in high-rise housing areas (HRHAs), intensifying heat stress and degrading outdoor thermal comfort (OTC). These neighborhoods—shaped by socialist-era planning and, later, market-led infill—combine high built density, low greenery, and limited ventilation, making them critical testbeds for climate-adaptive regeneration. This study presents the first empirically validated ENVI-met assessment of blue–green infrastructure (BGI) performance in a post-socialist HRHA, using a representative courtyard in Niš, Serbia, during the 14 August 2024 heatwave. A 24 h field campaign (air temperature, humidity, wind speed, and mean radiant temperature) validated the model with high accuracy (R² = 0.92, RMSE = 1.1 °C for air temperature; R² = 0.88, RMSE = 3.5 K for Physiological Equivalent Temperature (PET). Four retrofit scenarios were simulated: S0 (existing), S1 (grass), S2 (grass + trees), and S3 (S2 + shallow pool). Across all scenarios, daytime PET indicated strong–extreme heat stress, peaking at 61.9 °C (16:00 h). The best configuration (S3) reduced PET by 2.68 °C (10:00 h) but <1 °C at peak hours, with acceptable comfort limited to 04:00–07:00 h. The results confirm that small-scale surface-level greening provides negligible thermal relief under a dense HRHA morphology. Urban morphological reform—optimizing height, spacing, ventilation, and integrated greening—is more effective for heat mitigation. Future work should include multi-seasonal field monitoring and human thermal-perception surveys to link microclimate improvement with exposure and health risk. Full article

Global increases in drought frequency and severity pose growing risks to forest resilience, particularly for long-lived endemic tree species such as kauri (Agathis australis). Building on prior leaf-level work, this study assessed the utility of multitemporal canopy-scale hyperspectral imaging to characterise water stress in both controlled nursery and field conditions. Two complementary experiments were undertaken: (i) a 10-week controlled-environment experiment comparing drought and control groups, and (ii) a field-based assessment of juvenile kauri trees across multiple time points with contrasting soil volumetric water content. In the controlled-environment experiment, drought-treated seedlings exhibited delayed physiological responses, with reductions in stomatal conductance and assimilation emerging only after three weeks. In contrast, time-series analysis of narrow band hyperspectral indices (NBHIs) revealed detectable stress signatures within one week after drought initiation, with early sensitivity driven by structural and pigment-related indices. As stress progressed, pigment-specific indices became the dominant predictors. These findings were consistent with the field-based experiment. Variation in leaf equivalent water thickness (EWT) was strongly explained by pigment-sensitive indices, including Pigment Specific Simple Ratio Carotenoid (PSSRc) and Carotenoid Reflectance indices (CRI700 and CRI550), which together accounted for ca. 87% of the variance. Structural indices such as the Normalised Difference Vegetation Index (NDVI) also ranked among the top 20 predictors, but had comparatively lower explanatory power (<75%). Overall, the two experiments show that canopy-based hyperspectral imaging provides early, sensitive, and consistent detection of water stress in kauri. The findings highlight a scalable approach for monitoring drought impacts on kauri and offer a foundation for developing operational forest health tools under increasing climate pressure. Full article