ABSTRACT Microwave ablation (MWA) has become an increasingly important minimally invasive treatment option for pulmonary lesions, though its clinical performance remains limited by tissue heterogeneity and inconsistent energy deposition across patients. This study systematically investigates the therapeutic mechanisms of MWA, combining well‐documented thermal coagulation effects with the potential non‐thermal bioelectromagnetic effects discussed in existing preclinical research, through an integrated framework of computational modeling and orthogonal experimental design. Our Multiphysics simulations quantified the differential effects of antenna wavelength ( λ ), geometry (length‐to‐width, l/w ratio), and power ( P ) on specific absorption rate (SAR) distribution across pulmonary parenchyma. Particle swarm optimization (PSO) identified =70 W and λ = 0.1 m as optimal parameters, achieving peak SAR (78.48 W/kg) with statistically significant dominance of power ( p < 0.01) and wavelength effects on necrotic volume ( p = 0.02). Therapeutically, our analysis suggests that MWA's efficacy may derive from synergistic mechanisms: instantaneous thermal necrosis driven by 60°C–70°C protein denaturation, and potential nonthermal bioelectromagnetic effects that could disrupt cellular integrity and enhance anti‐tumor immunity. This combined mechanism may be particularly beneficial for improving ablation completeness in heterogeneous lung tissues, where conventional single‐factor thermal models often struggle to account for variable tissue properties. Our findings establish evidence‐based antenna optimization guidelines, demonstrating that precise modulation of P and λ can standardize ablation zones while minimizing collateral damage. The potential nonthermal effects summarized in this analysis provide new avenues to address persistent challenges in pulmonary MWA applications, which could ultimately help advance toward more predictable clinical outcomes, though further preclinical validation is still needed to confirm these effects.

ABSTRACT As universities undergo accelerated digital transformation, social media algorithms—while streamlining campus services—have emerged as a pivotal yet under examined source of student privacy vulnerabilities and cyber‐attack exposure. The opaque, recommendation‐driven “black box” mechanisms of these platforms incentivize users to over‐disclose sensitive personal information, while the aggregation of heterogeneous behavioral data across platforms intensifies risks of cross‐platform re‐identification and comprehensive user profiling. This process catalyzes a latent but systemic risk transmission chain: algorithm‐induced disclosure → multi‐source behavioral fusion → targeted exploitation . To accurately assess and block this chain, this paper constructs a privacy risk assessment mechanism for universities to evaluate the risks to university network security and student data privacy. First, the platform's digital footprint is structured into a 128‐dimensional feature vector; second, a weighted linear risk scoring function is designed based on the “Personal Information Protection Law of the People's Republic of China” and weighting; finally, based on high‐confidence association rules, the Drools engine is embedded to achieve dynamic early warning. Empirical results show that this mechanism can effectively identify high‐risk individuals (the top 18.01% covers 81.19% of leakage events) and reveals patterns such as a 267.1% surge in high‐risk behaviors among second‐year students during exam weeks. This study makes the implicit transmission chain of social media algorithms explicit, providing a feasible technical path and decision support for universities to build a proactive, time‐based, and hierarchical privacy protection system.

ABSTRACT In recent years, impact resistance has become a critical performance metric for fiber‐reinforced concrete (FRC), particularly in safety‐critical structures such as bridges, bunkers, and nuclear facilities. While drop‐weight impact tests provide a reliable evaluation, they are costly, destructive, and impractical for large‐scale design optimization. The aim of this study is to predict the impact resistance (N1 and N2) of fiber‐reinforced concrete using hybrid artificial neural network (ANN) models trained with genetic algorithm (GA) and Levenberg–Marquardt (LM) optimization techniques. The dataset comprises 117 experimental concrete samples, including 63 reinforced with polypropylene fibers, 45 with steel fibers, and 9 plain mixes for comparative assessment. Input features include fiber characteristics (type, length, tensile strength, aspect ratio), and concrete mix parameters. The results indicate that the developed ANN models successfully predicted the impact resistance of FRC with moderate predictive performance. Using the LM algorithm, the coefficients of determination (R) were as follows: N1—training (0.84094), testing (0.83542), validation (0.89993); N2—training (0.97655), testing (0.90333), validation (0.75404). Sensitivity analysis revealed that in the N1 scenario, Cement had the greatest influence, followed by fiber dosage and sand content, whereas in the N2 scenario, Fiber Dosage was the most influential parameter. These findings can assist engineers and designers in optimizing fiber‐reinforced concrete mix designs by focusing on the most influential parameters. Moreover, the developed model can significantly reduce the time and cost associated with extensive experimental testing and enable non‐destructive prediction of impact resistance across diverse mix designs and fiber types.

ABSTRACT The evolution of smart grid and the deepening of the power market have significantly enhanced the key position of precise electricity price estimation in industry decision‐making and the optimal resource allocation. It is necessary to implement short‐term electricity price forecast in the power market that accurately characterizes the dynamic change laws of electricity prices in a market environment with significant high volatility and nonlinear characteristics. To address this demand, the paper proposes an Attention‐based long short‐term memory model called ATT‐LSTM, which intends to improve the accuracy and interpretability of electricity price forecast, and provide support for power market operation analysis and trading decision making. In the model, multi‐dimensional feature engineering and data preprocessing are first performed, including meteorological data imputation, outlier removal, and time series sliding window construction. An attention layer is introduced to dynamically allocate importance weights to different time steps and features, allowing the model to adaptively focus on key influencing variables. This model is particularly suitable for electricity price forecast scenarios with significant seasonality and short‐term disturbances. Experiments are based on measured data from the Australian National Electricity Market, with comparisons conducted using multiple benchmark models, that is, LSTM, BPNN, and ATT‐LSTM. The results show that the ATT‐LSTM model achieves the optimal performance in terms of multiple metrics (e.g., MAE and RMSE) and prove that introducing the attention mechanism into the LSTM can effectively improve the accuracy and transparency of electricity price forecast.

ABSTRACT This study focuses on capturing the physiological state and action characteristics of badminton players in the high‐intensity and fast‐changing sports scene. It proposes a badminton fitness training evaluation system that combines smart wearable devices and deep learning. Through multi‐dimensional data fusion and innovative network architecture, the management from data collection to physical fitness evaluation is realized. First, the monitoring principle of the intelligent exercise bracelet is deeply explored. Through an adaptive filtering algorithm, the heart rate monitoring error for high‐intensity exercise is effectively reduced. Furthermore, the AMste‐Transformer network is innovatively proposed. Based on the time–space mixed attention mechanism, the network addresses the cross‐domain mapping challenge of millimeter‐wave radar data to three‐dimensional (3D) attitude. With adaptive convolution kernel sizes and an Efficient Channel Attention (ECA) feature enhancement module, this study improves estimation accuracy for critical joints including the shoulders, elbows, and wrists. Experiments are conducted on the Human 3.6 M dataset under Protocol 1 and Protocol 2. The model uses 8 attention heads, an embedded feature channel dimension of 512, and a network depth of 4. The experimental results show that the Mean Per Joint Position Error (MPJPE) of AMste‐Transformer in Protocol 1 is 54.3 mm, which is 8.7% lower than that of the second‐best PoseFormer. For badminton‐specific movements, the key joint error is 17.7% lower than that of the baseline Transformer method. The improved smart sports band has achieved high‐precision heart rate monitoring in badminton scenarios with various exercise intensities. Its Pearson correlation coefficient with the medical Holter gold standard reaches up to 0.99. This study provides a scientific tool for badminton training and a technical framework for physical fitness evaluation of other high‐intensity intermittent sports.

ABSTRACT Accurate geometrical representation of stenosis is essential for stent design, surgical planning, and computational fluid dynamics (CFD) simulations, as even minor shape variations significantly alter hemodynamic predictions. This review systematically compiles and classifies the diverse stenosis geometries proposed in prior studies, creating a foundational reference for researchers. We catalog models ranging from idealized analytical shapes (e.g., axisymmetric cosine, Gaussian, and asymmetric profiles), patient‐specific reconstructions to parametric frameworks, highlighting their mathematical formulations, hemodynamic implications, and clinical applications. By consolidating these geometries, this article enables researchers to: (1) identify the most suitable existing model for their specific study, (2) understand inconsistencies in results across studies due to geometrical differences, and (3) develop new models informed by prior morphological variations. Crucially, we emphasize how stenosis geometry governs key hemodynamic parameters such as wall shear stress and pressure gradients and influences stent performance, underscoring why shape selection cannot be overlooked. This structured classification is intended to guide model selection and study design by aligning geometric fidelity with specific research objectives and practical constraints, rather than to imply predictive accuracy or clinical superiority.

ABSTRACT Global access to neutron scattering is under increasing pressure as research reactors close and demand for spallation sources outpaces supply. Compact accelerator‐driven neutron sources (CANSs) offer a pathway to more distributed access, yet current designs are often limited by scale, cost, and performance, and no turnkey solutions presently exist. This work explores the potential of next‐generation CANS based on high‐gradient electron linear accelerators that leverage technology originally developed for high‐energy physics applications. By combining mature ‐, ‐, and ‐band RF structures with commercially available RF power sources and compact injectors, we identify and evaluate accelerator configurations capable of delivering tens of MeV, kilowatt‐scale electron beams within a few meters of linac length and at accessible cost. To ground the study, we optimize the accelerator in line with the requirements of VULCAN, a concept for a turnkey CANS‐based facility optimized for stress diffractometry and suitable for industrial and university environments. We show that competitive neutron yields and spectra can be attained while maintaining compactness and affordability. These findings open up a previously unexplored region of the CANS design space and highlight the potential of high‐gradient accelerator technology to democratize access to neutron scattering through a new generation of compact, affordable, and high‐performance neutron facilities.

ABSTRACT Constructing underground tunnels and associated infrastructure in tectonically sensitive mountainous regions such as Bilaspur, Himachal Pradesh, presents significant engineering challenges due to variable rock mass properties, complex deformational behavior and weak litho‐structural conditions. Detailed pre‐investigation of geological and structural characteristics is therefore essential to evaluate rock stability prior to infrastructure development, particularly in steep terrains that exhibit elevated geotechnical and environmental risks. This study assesses rock mass quality using the rock mass rating (RMR) geomechanical classification system based on detailed field investigations, including intact rock strength, rock quality designation (RQD), discontinuity spacing, discontinuity condition, groundwater condition, and discontinuity orientation. To enhance the reliability of stability assessment, geographic information system (GIS)‐based spatial analysis was integrated to evaluate terrain slope, fault distribution, lineaments and regional seismicity. Although GIS parameters were not used directly in RMR computation, they were employed as an independent spatial validation framework to examine the consistency between geomechanical classification and regional structural controls. The results indicate RMR values ranging from 21 to 30, corresponding to Class IV (poor rock mass quality), primarily controlled by closely spaced discontinuities and unfavorable structural orientations. The novelty of this study lies in integrating field‐based geomechanical classification with GIS‐derived tectonic and geomorphological analyses to provide spatial validation of tunnel stability conditions in a tectonically sensitive Himalayan terrain, enabling improved assessment of engineering feasibility beyond conventional RMR evaluation. The findings suggest that tunnel construction in the study area requires enhanced support systems, including the New Austrian Tunneling Method (NATM), rock bolting, steel ribs, and shotcrete to ensure long‐term stability. This integrated framework provides a practical basis for planning railway tunnel infrastructure in structurally complex mountain environments.

ABSTRACT Dynamic process modeling is essential for simulating time‐evolving biochemical systems, particularly those with multistate interactions and combinatorial complexity. Traditional Ordinary Differential Equation (ODE) models offer mechanistic clarity but struggle with scalability and context‐sensitive encoding. Rule‐Based Modeling (RBM) frameworks address these limitations through modular rule abstraction, yet require manual specification and lack adaptive learning. This study introduces algorithmic innovations within the Neural Ordinary Differential Equation (Neural ODE) paradigm to bridge the gap between mechanistic interpretability and scalable expressivity. Neural ODEs can be considered as a revolutionary approach in the field of modeling dynamic biochemical interactions. They have made it possible to create models of such interactions that are flexible enough to adapt to different scenarios and do so without requiring any manual intervention in terms of rule encoding or predefined reaction schemes. This is achieved by employing differential solvers within the framework of neural networks, thus enabling a learning process that is in accordance with the behavior of the system. Using the DARPP‐32 signaling network—a benchmark system characterized by multivalent phosphorylation and dynamic perturbations—the proposed Neural ODE framework demonstrates the ability to replicate key dynamic behaviors observed in ODE and RBM models. Comparative simulations under baseline and perturbed conditions reveal that Neural ODEs maintain trajectory fidelity while offering enhanced modularity and computational efficiency. Feature importance analysis and latent space visualizations further validate the model's interpretability and robustness. Unlike ODEs and RBMs, Neural ODEs adapt to structural mutations and binding schemes through latent trajectory learning, enabling flexible simulation of biochemical variability without manual rule encoding. This work establishes Neural ODEs as a viable and scalable alternative for modeling complex biochemical systems, combining the strengths of data‐driven learning with the interpretability of differential equations.