Browse Articles

Assessing the capability of large language models in answering pediatric critical care board-style questions

Optimized machine learning and artificial neural networks for NIRS-based prediction of mango internal quality

Abstract A robust and updated approach for non-destructive prediction of key quality attributes in intact mangoes was developed by integrating near-infrared reflectance spectroscopy (NIRS) with advanced machine learning and artificial neural network (ANN) algorithms. In this study, NIR spectral data of 136 mango samples were collected and processed using standard normal variate (SNV) correction. Five regression methods namely partial least squares regression (PLSR), support vector machine regression (SVMR), random forest regression (RFR), extreme gradient boosting (XG-Boost), and generalized regression neural network (GRNN), were optimized and evaluated for predicting total acidity (TA) and ascorbic acid (AA). Results indicate that while traditional linear methods like PLSR achieved reasonable predictive power (RPD > 2.5), nonlinear models, especially XGBoost and GRNN, significantly outperformed PLSR, with GRNN models achieving the highest accuracy, with maximum performance reaching R 2 up to 0.98 and RPD > 5.5 across the evaluated parameters (TA and AA). The findings demonstrate that optimized machine learning and ANN models offer robust, accurate, and practical solutions for rapid, non-invasive mango quality assessment. This integrated methodology supports advanced quality control, sorting, and breeding programs, providing substantial benefits for industry and supply chain management through rapid, reliable assessment of fruit nutritional and chemical properties.

Enhancing prediction accuracy for Parkinson’s disease using advanced machine learning models

Effect of multicomponent support intervention on medication adherence and self-efficacy levels in hypertension patients

Exploring hemodynamic measurements from the Tromsø Study for prediction of cardiovascular disease using traditional statistical models and machine learning approaches

Abstract In Norway, NORRISK2 is the government-recommended risk model for predicting an individual’s 10-year probability of getting cardiovascular disease (CVD). This study aims to investigate the potential for improvement of CVD prediction by using hemodynamic measurements from a non-invasive beat-to-beat blood pressure monitor, taken as part of pain sensitivity assessment with the cold-pressor test (CPT) during the Tromsø6 Study (2007–2008). Using 6694 recordings, ultra-short-term pulse rate variability (PRV) and baroreflex sensitivity (BRS) obtained during the CPT were added as additional variables into the existing NORRISK2 survival model (extended model). In addition, the time-series data was used in a machine learning (ML) model without the NORRISK2 background variables. Both models were compared to a recalibration of the original NORRISK2 model. The predictions from the recalibrated NORRISK2 model and the ML model were then combined with logistic regression. The statistical models performed similarly on the test set, with an area under the receiver operating characteristic (AUROC) of 0.8 (95% CI: 0.71–0.86), 0.79 (0.71–0.85) and 0.77 (0.69–0.84) (original, recalibrated and extended NORRISK2, respectively). The ML model using only hemodynamic measurements obtained a test set AUROC of 0.73 (0.67–0.80). Combining the NORRISK2 and ML model did not increase the AUROC. Adding ultra-short-term PRV and BRS derived from Tromsø6 did not improve the prediction of the NORRISK2 model either. Although with lower accuracy, the beat-to-beat time series of hemodynamic variables from a CPT had a significant (p < 0.01) ability to predict future CVD without any other person-specific data.

Patterns of compliance with COVID-19 preventive measures in Armenia: results from a cross-sectional survey

Efficient and interpretable maximal frequent fuzzy pattern mining with multi phase pruning and ternary search

Abstract The exponential growth of quantitative data across various domains has intensified the need for efficient pattern mining techniques that can handle numerical uncertainty while maintaining interpretability. Traditional fuzzy frequent pattern mining algorithms suffer from pattern explosion in dense datasets, generating overwhelming numbers of redundant patterns that hinder practical analysis. This study introduces a novel Maximal Frequent Fuzzy Pattern Mining (MFPM) framework that integrates fuzzy set theory with maximal pattern representation to address these limitations. The proposed methodology employs a multi-phase approach that incorporates aggressive pruning strategies, including maximum cardinality selection and early termination, to reduce the dimensionality of the search space. Evaluation on three datasets (Chess, Connect and Mushroom) demonstrates consistent gains in both effectiveness and efficiency. Time-wise, MFPM accelerates discovery where classical algorithms are slowest: dense regimes and permissive supports. A ternary search algorithm efficiently identifies the longest patterns, while an Anti-Apriori strategy with superset pruning ensures the extraction of only non-redundant maximal patterns. Experimental evaluation on benchmark datasets demonstrates remarkable effectiveness, achieving up to 94.97% pattern reduction compared to traditional FTDA algorithms while maintaining equivalent knowledge representation. Computational efficiency improved by over 65% in challenging low-support scenarios. The framework generates concise, semantically interpretable patterns that capture the most significant relationships in quantitative data, facilitating informed decision-making across diverse application domains, including healthcare analytics, business intelligence, and web usage mining.

Event-preserving feature engineering for intermittent demand forecasting using SHOS

Identification of TBXAS1 as a candidate biomarker and potential microglia-associated inflammatory regulator in Parkinson’s disease

Anchusa azurea enhances cisplatin efficacy in oral and bone cancers through IL-17 and TNF-α pathway modulation: a metabolomic and network pharmacology approach

Abstract Anchusa species have traditionally been used to treat arthritis, gout, rheumatism, and skin wounds. Cisplatin (Cis) is a widely used chemotherapy drug associated with serious adverse effects. The study aimed to evaluate the potential synergistic anticancer effects of Anchusa azurea methanol extract (AAME) in combination with cisplatin against bone, skin, and oral cancer cell lines. This study involved a comprehensive metabolomic profiling of AAME, alongside cytotoxicity assays, cell cycle analysis, autophagy assessment, and evaluation of IL-17 and TNF-α pathway-related protein expression. AAME inhibited the proliferation of MG63 and HNO97 cancer cells while sparing HSF normal cells. AAME and Cis displayed synergistic effects (combination index < 1), especially in HNO97 cells. Treatments led to a synergistic decrease in TNF-α, p/t-JNK, IL-17, pNFκB/tNFκB, TRAF6, pMAPK/tMAPK ratios, and AP1 expression, also increased Casp3 and Casp8 levels, cell cycle arrest, and enhanced autophagy. The TPC and TFC of AAME are 5.46 mgGAE/gE and 0.13 mgRE/gE respectively, reflecting on its radical scavenging activity (EC50 209.67 µ g/mL). HRLC-MS/MS leading to the annotation of 50 metabolites, including phenolics and flavonoid derivatives, notably with a prevalence of rosmarinic acid, quercetin, and kaempferol. In network pharmacology, the 90 genes are common between AAME constituents and oral cancer. A. azurea enhances cisplatin’s anticancer effects by modulating IL-17, TNF-α, and apoptotic pathways, offering a promising adjuvant therapeutic strategy. Further in vivo investigations are warranted to validate the observed in vitro synergistic anticancer effects of A. azurea in combination with Cis.

Clinical ethical practice and associated factors among health professionals in public hospitals in addis ababa, Ethiopia

THR-DMPOM model for reliable high-precision evaluation to enhance real-time performance management and optimize resource allocation

The role of metacognition in fibromyalgia impact and psychological distress: a cross-sectional study in Turkish women with fibromyalgia

Quantitative characterization of thermo-optically modulated microbend loss in silica fibers for high-temperature deep-well telemetry

Abstract To mitigate the nonlinear amplification of fiber micro-bend loss in high-temperature deep-well environments (303.15 ~ 483.15 K), this study establishes an equivalent refractive index model integrating thermal, mechanical, and optical fields. The model incorporates thermo-optic effects(TOEs) directly into the micro-bend equivalent refractive index expression, facilitating unified modeling of coupled thermal and bending perturbations. Finite element eigenmode analysis was employed to quantify the synergistic effects of bending radius (0.1 ~ 1.0 mm) and temperature on radiation leakage and mode coupling. Results demonstrate that temperature does not act as an independent loss term. Instead, it reconfigures waveguide confinement by modulating the core-cladding refractive index profile, thereby altering mode-coupling pathways. Multimode fiber (MMF) exhibits oscillatory loss governed by inter-modal phase matching, with attenuation coefficients displaying non-monotonic temperature dependence. Conversely, single-mode fiber (SMF) demonstrates threshold-type loss surge at temperatures T  ≥ 440 K. Spectral and phase analyses reveal that MMF undergoes significant phase decorrelation and topological discontinuity under stochastic mode coupling, whereas SMF maintains phase continuity even under high-loss conditions, demonstrating superior coherence stability. Furthermore, wavelength scanning shows that long wavelengths (> 1.6 μm) significantly enhance SMF micro-bend sensitivity, while the stochastic coupling in MMFs is better suited for short-distance, intensity-modulated telemetry. This study establishes a quantitative correlation between temperature, bending, and wavelength, providing essential physical criteria and engineering guidelines for fiber selection and loss compensation in extreme downhole environments.

Photoacoustic device fingerprints induce bias in deep learning models

Abstract Deep learning (DL) models developed for established medical imaging modalities have shown increasing performance and reliability as a result of scaling efforts. In contrast, model development for emerging modalities such as photoacoustic imaging (PAI) remains challenged by data sparsity, which limits model generalizability and raises the susceptibility to bias. While recent studies in PAI have started to investigate subject-related confounders, the impact of hardware-related confounders remains unexplored, posing a critical risk for failure in multicentric deployment scenarios. We are the first to provide a multicentric analysis of hardware-induced bias in PAI. We analyzed device-specific characteristics in images from four device instances and two peripheral artery disease studies, and trained DL models to classify device origin and disease under varying levels of device–health correlations in the data. We showed that 1) multiple instances of the same PAI device type embed identifiable fingerprints in the images, 2) that DL models can leverage these fingerprints to reach $$100\,\%$$ accuracy in device detection and critically, 3) when a correlation between device instance and health status is present, models trained for disease diagnosis exploit these device-specific signatures as shortcuts, thereby producing biased and clinically misleading predictions. This research highlights the risk of overestimating algorithm performance when such confounding is overlooked, emphasizing the importance of bias evaluation and explainable artificial intelligence methods to identify potential shortcuts, finally enabling multicentric PAI studies.

Assessing Zn12O12 nanocage for the elimination of ciprofloxacin from drinking water using DFT simulations

Evolutionary game-theoretic modeling of electricity market dynamics for elastic load optimization

Analysis of stability and chaotic trajectories in nonlinear fluid wave interactions under forcing effects

Extended Maxwell equations derived from spacetime geometry

Abstract This paper presents a unified relativistic description of electromagnetic and gravitational interactions in the classical regime, addressing the fundamental question of why like charges repel while like masses attract. We formulate a flat spacetime geometry tailored to special relativity and introduce intrinsic operations of spacetime product and spacetime curl, yielding a concise formalism arising directly from relativistic spacetime properties. Based on three axiomatic postulates—relativistic properties of spacetime, energy conservation, and standard charge condition—we rigorously derive four extended Maxwell equations and their corresponding force formulas by employing our formalism, without ad hoc parameters or phenomenological fitting. These simultaneously recover electromagnetic and gravitational interactions, along with their mirror-symmetric counterpart, thereby naturally resolving the aforementioned fundamental question. A thought experiment demonstrates consistency with the principle of relativity. We propose a unified charge system to facilitate the unified description. The framework remains strictly classical, does not incorporate spin and radiative-reaction effects, and applies to gravity in quasi-charge conditions where the motion of matter can be treated as a conserved Lorentz four-vector current. This axiomatic derivation provides a geometrically grounded unified description of long-range interactions and a transparent foundation for theoretical extensions.

Stochastic Grey Wolf Optimization for Hyperparameter Tuning of LSTM and RNN Models in Energy Forecasting

Abstract Accurate photovoltaic (PV) power forecasts are required to support the stable and efficient incorporation of solar power into modern electric power grids. Even though the use of recurrent neural network models such as RNNs and LSTMs for time series forecasting has proven successful, these model’s ability to make accurate predictions is heavily influenced by proper hyper parameter selection. Therefore, the goal of this research is to introduce a stochastic grey wolf optimizer (SGWO) based hyper-parameter optimization system for RNN and LSTMs used for predicting PV power. A stochastic grey wolf optimizer will be added to the basic grey wolf optimizer to enhance the search capabilities of the algorithm. This new stochastic grey wolf optimizer introduces randomness into the optimization process which can help prevent premature convergence and increase exploration within the problem space. The performance of the proposed SGWO-based system will be tested on a real world PV data set. The comparison will include results from manual tuning, random searching, and standard GWO. Performance metrics will consist of root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) and will be used to evaluate how well each of the algorithms performed when making out-of-sample predictions. Results from testing showed that the SGWO-LSTM configuration produced the highest total out-of-sample prediction accuracy; MAE = 0.018, RMSE = 0.041, R 2  = 0.978.