Search Results (376)

Water deficit is a major constraint on apple production in arid and semi-arid regions, and the selection of drought-adapted rootstocks is an effective strategy for improving orchard sustainability. This study evaluated the drought resistance of ten apple rootstocks and identified promising germplasm for apple production in the arid region of the Loess Plateau in Ningxia, China. A field experiment was conducted in Yanchi County, where annual precipitation is approximately 200 mm. Twelve leaf anatomical traits, nine stomatal traits, and ten physiological and biochemical parameters were measured and integrated for multivariate analysis. Significant differences were observed among rootstocks in epidermal, mesophyll, and stomatal traits. Qingzhen No. 1 exhibited the greatest upper and lower epidermal thicknesses, Qingzhen No. 2 had the greatest leaf thickness, and Pajam had the greatest mesophyll thickness. Stomatal traits also differed substantially among rootstocks, with SH40 showing the highest stomatal density, whereas Qingzhen No. 1 showed the largest stomatal size and opening characteristics. Pajam also exhibited relatively high proline content and POD activity, indicating coordinated osmotic and antioxidant responses under severe aridity. Principal Component Analysis (PCA) of 21 leaf anatomical and stomatal traits extracted five principal components explaining 91.888% of the total variance, while PCA of the 10 physiological and biochemical indicators extracted four principal components explaining 87.394% of the total variance. Membership function and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) analyses consistently ranked Pajam highest, followed by M26 and B9. The rootstocks were classified into high (Pajam and M26), moderate (B9, Qingzhen No. 1, SH40, JM7, T337, and Qingzhen No. 2), and low (M7 and Nic29) drought-resistance categories. These findings support environment-specific rootstock selection for apple production under water-limited conditions. Full article

Background/Objectives: We assessed changes in nonfatal disease burden and risk attribution in Romania between 1990 and 2023. Methods: We analysed Global Burden of Disease (GBD) 2023 estimates of years lived with disability (YLDs), related nonfatal measures and risk attribution for 199 causes and 61 risks, with seventeen comparator locations. Results: The age-standardised YLD rate changed by −4.1% (95% uncertainty interval [UI] −10.1 to 1.3), and total YLDs by −3.9% (95% UI −9.2 to 0.7). Both intervals included zero. The crude rate rose by 17.0%. The population size and age structure components of the decomposition were approximately −552,000 and +545,000 YLDs; their uncertainty was not estimated. Healthy life expectancy increased by 5.2 years. Its difference from life expectancy rose from 9.2 to 10.2 years, without an uncertainty interval for this derived change. Low-back pain remained the leading level 3 cause. Anxiety disorders increased by 41.7% (95% UI 4.5 to 101.8) and ranked second by point estimate; rank uncertainty was not estimated. The joint risk-attributable fraction fell from 31.2% to 28.3%. Conclusions: Romania had more disability per inhabitant and a larger share at older ages in 2023. Musculoskeletal disorders remained the main source of nonfatal burden, while anxiety disorders and diabetes increased. Full article

Background: Early broad-spectrum empirical therapy is life-saving in critical illness, but its unnecessary continuation drives resistance, toxicity, and cost; antibiotic de-escalation is the principal stewardship strategy for resolving this tension, yet its evidence base is fragmented and its practice inconsistent. This scoping review mapped the evidence on the definitions, timing, eligibility, implementation, safety, and clinical, microbiological, and resistance outcomes of de-escalating cephalosporins, carbapenems, colistin, and tigecycline in critically ill adults with suspected or confirmed Gram-negative infection. It translated this into a framework for a tertiary-care intensive care unit (ICU). Methods: We conducted a focused scoping review informed by JBI methodological guidance and reported according to PRISMA-ScR. The Web of Science Core Collection was searched for English-language publications from 1 January 2016 to 4 August 2026. Following deduplication, two reviewers independently screened titles and abstracts, and subsequently assessed potentially eligible full texts. Disagreements were resolved through discussion or consultation with a third reviewer. The review was designed to map the characteristics and range of the identified evidence rather than to provide an exhaustive systematic assessment or quantitative synthesis of intervention effects. (PROSPERO CRD420261478424). Results: We included 51 publications (35 empirical studies; 16 reviews, editorials, or consensus statements), with the empirical evidence being predominantly observational, including a single randomized trial. The definitions were heterogeneous, and the spectrum-ranking systems were non-uniform; reassessment typically occurred at 48–72 h. The reported de-escalation proportions ranged from approximately 10% in broadly defined treated populations to 71% in a selected, extractable ICU subgroup. These values were not directly comparable because studies used different definitions, eligibility criteria, time points, and denominators, including all patients treated with antibiotics, empirical-treatment episodes, microbiologically documented infections, and patients considered clinically eligible for de-escalation. Direct comparative studies did not identify a consistent increase in mortality following de-escalation; however, the predominantly observational evidence was vulnerable to confounding by indication, survivor bias, and treatment-selection bias, and did not establish equivalence, non-inferiority, or a survival benefit. Conclusions: De-escalation appears safe but rests on low-certainty, heterogeneous evidence. We propose an evidence-informed framework, a structured 48–72 h time-out, an eligible-patient denominator and a minimum monitoring dataset for tertiary ICUs, and identify standardized definitions and resistance-focused trials as research priorities. These components represent an evidence-informed implementation proposal developed by the authors and require prospective local validation. Full article

In developing countries, railway networks often create operational bottlenecks by combining older single-track sections with modern, double-track, signaled infrastructure. This study addresses train scheduling in such hybrid networks using a segment-based rather than a train-based approach. A mixed integer linear programming (MILP) model is developed that minimizes total waiting time across all journey segments, incorporating directional safety intervals separating single-track and double-track sections and co-directional and counter-directional movements. Train-type priority is included in the objective function as a class-dependent, age-adjusted weight, rather than a rigid priority rule; thus preventing the unlimited delay that a rigid priority ranking might otherwise impose on lower-priority trains in heavily shared resources. Since an exact MILP solution is impractical on a national scale, a segment-based heuristic method is developed; the age-weighted heuristic method, validated against exact solutions on smaller subnetworks, is found to require 35% more time to reach full optimum. This approach, applied to Turkey’s national railway network (109 routes, 997 stations or siding sections, 2281 transit segment between two stations or siding sections, 1539 daily services), has reduced daily waiting times from 2508 h to 609 h. This corresponds to an annual saving of approximately $13.2 million in fuel and labor costs and a reduction of 24,260 tons/year in CO₂ emissions. Since diesel-fueled services account for more than 99% of the cost savings, sensitivity analyses show that these estimates are robust against fuel price assumptions. The study offers a low-cost solution for improving railway capacity before costly infrastructure investments and is expected to be generalized to similar heterogeneous national networks. Full article

Ultrasonic pulse-echo imaging is one of the most widely used non-destructive evaluation (NDE) modalities for monitoring the structural integrity of reinforced-concrete civil infrastructure such as bridge decks, tunnel linings, and dam walls. In this acquisition geometry, a single low-frequency transducer is mechanically raster-scanned over the accessible top surface of the specimen and records one A-scan per scan position, simultaneously serving as transmitter and receiver. However, commonly used reconstruction algorithms such as the Synthetic Aperture Focusing Technique (SAFT) and Reverse Time Migration (RTM) tend to produce reconstructions of limited quality on this class of data because they do not adequately model the round-trip propagation kernel that is specific to the monostatic geometry, they do not separate the strong near-surface direct-arrival reflection from the bulk image, and they do not account for the persistent aggregate-induced clutter that contaminates every A-scan in concrete media. In this paper, we propose a clutter-aware reconstruction method for monostatic ultrasound acquisition (CARMA), whose main innovation is the joint integration of a monostatic-specific round-trip propagation model, a dedicated near-surface direct-arrival subspace, and a data-adaptive low-rank clutter subspace within a unified model-based reconstruction framework. Unlike existing reconstruction approaches, CARMA explicitly accounts for the co-located transmit–receive geometry through a squared-cosine round-trip directivity model while simultaneously separating scan-dependent direct-arrival contributions and aggregate-induced clutter from the desired reflectivity image. To verify the method under fully controlled and repeatable conditions, we generate intensive, physically realistic full-wave simulations with the k-Wave pseudo-spectral acoustic solver that reproduce a representative civil-infrastructure inspection scenario: three reinforced-concrete specimens with a stepped back wall of varying thickness, ten embedded ground-truth defects spanning steel tendon ducts and low-impedance polystyrene inclusions, a monostatic raster-scanned pulse-echo acquisition, and randomly distributed aggregate scatterers that reproduce the clutter of real concrete. Results on these intensive k-Wave simulations indicate that CARMA reconstruction yields approximately $2\times$ lower localization error than RTM and approximately $4\times$ lower localization error than SAFT, while recovering the deepest embedded defect with substantially better localization and contrast than the comparison methods. Full article

Images from high-resolution cameras are mapped onto a sparse pattern of low spatial resolution and intensity in the retina, which limits visual perception in retinal prosthetic vision. When the entire scene is converted into phosphenes, it may allow unnecessary background information to be retained and may cause visual clutter, which may make it hard for prosthetic vision users to interpret the scene. In order to tackle this issue, this paper presents a context-, depth-, and user-preference-aware method for selecting the objects of interest in the generation of phosphene images. The proposed method does not show all the objects equally but learns to sort the objects according to their relevance to prosthetic vision. Manual annotation of a subset of COCO images was conducted where the most salient object was selected based on environment type, scene type, user mode, safety, navigation relevance, task importance, and distance. All of the candidate objects are described by full-scene visual features, object-crop features, handcrafted priority features, context embeddings, and monocular depth features. To predict object-level importance scores and identify the Top-1 and Top-4 important objects in unseen scenes, a hybrid deep learning model combining twin ResNet-18 backbones for scene and object feature extraction with embedding-based context encoding was trained. Priority maps and phosphene images were then created using the selected object masks and were depth-weighted. Two types of phosphene representations were also produced: Canny-edge-based and direct full images. The proposed framework is designed to suppress irrelevant background areas and improve important and closer objects in order to obtain a simplified and informative prosthetic-vision representation of the scene. The experimental evaluation, including Top-1 accuracy, Top-3 accuracy, mean reciprocal rank (MRR), and visual comparison, demonstrates the effectiveness of the proposed framework, achieving a Top-1 accuracy of 90.12%, a Top-3 accuracy of 97.45%, and an MRR of 0.9368. Furthermore, the proposed Canny-priority phosphene representation achieved an average human-participant recognition accuracy of approximately 86%. The proposed method offers a user-adaptive strategy for selecting and visualizing the information of a scene under the severe constraint of the bandwidth of retinal prosthetic vision. Full article

Background and Objectives: Sarcopenia is common among patients receiving maintenance hemodialysis, although prevalence estimates vary according to the diagnostic approach. We assessed sarcopenia defined according to the revised European Working Group on Sarcopenia in Older People (EWGSOP2) criteria, its baseline correlates, and its association with 24-month all-cause mortality. Materials and Methods: In this single-center prospective cohort at a tertiary nephrology center, 176 patients enrolled in December 2022 were followed for 24 months. Confirmed sarcopenia required both low handgrip strength and a low appendicular skeletal muscle index assessed by bioelectrical impedance. Physical performance was not assessed. Because the proportional-hazards assumption was not satisfied, restricted mean survival time (RMST) was used as the primary measure of the survival difference. Results: Sarcopenia was present in 46 of 176 participants (26.1%, 95% CI 20.2–33.1). Older age, lower serum albumin and phosphorus concentrations, lower Kt/V, and longer time on maintenance hemodialysis were independently associated with sarcopenia. Thirty-one participants died during follow-up, and survival was lower among those with sarcopenia (log-rank p = 0.035). The adjusted RMST difference was −3.88 months (95% CI −6.49 to −1.28; p = 0.003). A Cox sensitivity analysis yielded a hazard ratio of 2.62 (95% CI 1.06–6.45). There was no evidence of a sarcopenia-by-sex interaction in the adjusted RMST analysis (p = 0.964). Conclusions: Confirmed sarcopenia was present in approximately one quarter of patients and was associated with approximately four months shorter restricted mean survival over 24 months. EWGSOP2-based assessment may provide prognostic information in this setting; whether routine screening or management based on sarcopenia status improves clinical outcomes remains to be determined. Full article

Radiation fields around collimated or shielded sources exhibit strong gradients whose accurate delineation is critical for worker protection and emergency response. Mobile robots can survey such fields, but they sample sparsely and irregularly along their trajectories, and it remains unclear which reconstruction method can be trusted, and where. Using a single dominant collimated source in a two-dimensional indoor setting, this study shows that the answer depends decisively on sampling geometry, and proposes a physics-guided Gaussian process (GP) that performs reliably under trajectory-constrained sampling. A tracked robot combining light detection and ranging (LiDAR)-based simultaneous localization and mapping (SLAM) with a $\gamma$ dose-rate detector surveyed a collimated Cs-137 field in seven independent runs, and all methods were evaluated under both random hold-out (interpolation near visited locations) and spatial block cross-validation (extrapolation into unvisited regions); truth-referenced evaluation against a dense reference field is provided by Poisson-sampled simulations, while experimental accuracy is cross-validated on held-out measurements. Under uniform sampling, a multilayer perceptron (MLP) robustly outperformed GP variants ($R^2=0.95$, stable across 18 seed combinations); under trajectory sampling, its advantage vanished at visited locations and reversed catastrophically in unvisited regions. The proposed physics-guided GP, which uses a fitted collimated-beam template as the GP mean with a Matérn 3/2 residual process, achieved the highest extrapolation $R^2$ (median $0.61$; best baseline $0.31$), reduced the extrapolation error by 32–69% relative to all eight baselines, recovered interpretable source parameters, and provided predictive intervals with approximately calibrated region-level coverage (point-wise error ranking remains weak); a runtime fit-quality gate further renders the correctness of the embedded prior an observable quantity, so the method flags when its own assumptions fail. These results offer quantitative guidance for method selection in robotic radiation mapping under the as low as reasonably achievable (ALARA) principle. Full article

Conventional inverse synthetic aperture radar (ISAR) imaging techniques can produce high-resolution imagery from complete observation data. However, in practical scenarios, incomplete data caused by undersampling or missing data often leads to defocused results with traditional methods. While compressive sensing or low-rank reconstruction approaches have been proposed to address this challenge, existing techniques frequently fail to fully exploit both the sparsity and low-rank properties inherent in ISAR scenes. Moreover, they typically rely on convex approximations that introduce estimation bias, weaken sparsity promotion, and increase computational complexity, ultimately degrading imaging performance. To overcome these limitations, this work presents an enhanced sparse ISAR imaging method that jointly enforces non-convex sparsity and low-rank constraints for incomplete data recovery. The imaging model incorporates both inherent sparsity priors and a low-rank constraint. The resulting non-convex optimization problem is solved via an efficient iterative algorithm based on the alternating direction method of multipliers, where the sparse component is reconstructed using an iterative reweighted scheme with a regularizer and the low-rank component is recovered through truncated singular value decomposition. Experimental results on both simulated and measured data demonstrate the efficacy and superior performance of the proposed method. Full article

The informal waste recycling sector provides livelihoods for roughly 0.2% of the urban population in developing economies, yet its formalization amid economic development creates trade-offs among social welfare, environmental performance, and fiscal sustainability. This study formulates the informal-to-formal recycling transition as a three-objective problem: minimizing the residual informal waste-picker headcount (an exit/displacement proxy rather than a measure of integration success), maximizing recycling rates, and minimizing government expenditure. We solve this problem with NSGA-II and SPEA2 on a stylized model whose parameters are calibrated for order-of-magnitude consistency with publicly documented Chinese municipal waste-management benchmarks, a directional sanity check rather than a formal parameter fit; each algorithm runs 30 times. SPEA2 attains a modestly higher hypervolume than NSGA-II (0.733 vs. 0.714; two-sided Mann–Whitney U rank test, $p\approx 3\times {10}^{-11}$), and both algorithms converge within roughly 50 generations on this tractable two-variable problem. An unconstrained random search remains highly competitive, underscoring how readily this low-dimensional frontier can be approximated. The decision space comprises two policy levers, the formalization target $\alpha$ and the enforcement intensity E; a subsidy dimension is excluded because it would enter the fiscal objective only as a cost without a benefit channel. All fiscal values are expressed in abstract model currency units. A one-at-a-time sensitivity analysis identifies the formalization gain coefficient ${\gamma }_1$ as the parameter with the highest hypervolume coefficient of variation (CV = 13.2%), pinpointing where empirical estimation effort matters most within this model. Scenario simulations under boom, recession, and transition conditions shift the Pareto frontier in distinct ways. A bi-objective ablation indicates that the three objectives are largely independent, since only a small share of bi-objective solutions is strictly dominated in the full three-objective space; a quantitative hidden-cost penalty is not established. A representative compromise strategy ($\alpha \approx 0.97$, $E\approx 0.33$) reaches a 44.5% recycling rate with near-complete formalization at modest fiscal cost. The framework thus serves as a transparent, reproducible decision-support tool for urban waste governance in developing economies, making explicit the trade-offs that the model implements. Full article

The coupled effect of in situ temperature and stress complicates the permeability evolution of coal reservoirs, which restricts the exploration and evaluation of deep coalbed methane (CBM). Two high-rank coal samples were collected from the Sihe (SH) and Zhaozhuang (ZZ) mining areas, and multi-gradient coupled temperature–stress seepage experiments (20–50 °C, 8–32 MPa) as well as supporting triaxial mechanical tests were carried out to investigate the temperature and stress sensitivity of coal permeability. Combined with coal mechanical deformation characteristics, the transition depth mechanism of permeability evolution with burial depth was revealed. Experimental results indicate that coal permeability follows a negative exponential decay trend with increasing effective stress, and the evolution process can be divided into three stages: rapid attenuation, slow decline and stabilization. Temperature rise can weaken the stress attenuation degree of coal permeability under continuous effective stress loading and effectively reduce the stress sensitivity of coal reservoirs. Under constant confining pressure, permeability decreases linearly with rising temperature; the temperature-induced damage effect is prominent at low effective stress, while the regulatory effect of temperature is greatly weakened when fractures are compacted under high effective stress. An exponential function between permeability and burial depth was established based on coupled temperature–stress experimental data, and the critical burial depth of permeability transition depth in the study area was determined to be 550–600 m. The abrupt change interval of elastic modulus against confining pressure is consistent with the burial depth of permeability transition depth, which acts as the key mechanical factor dominating the nonlinear transition of reservoir permeability. This study provides experimental and theoretical support for the development of deep CBM in the study area. The results represent non-adsorbing gas (nitrogen) permeability under the investigated temperature–stress window (20–50 °C, 8–32 MPa) and should not be extrapolated to methane-bearing CBM reservoirs without adsorption–swelling corrections. The transition depth of approximately 550–600 m is a laboratory-derived estimate rather than a field-verified reservoir threshold. Full article

Background/Objectives: Traumatic intracranial hemorrhage (tICH) in older adults following low-energy falls (LEF) represents a common yet diagnostically challenging condition in the emergency department (ED), where predicting injury prior to computed tomography (CT) remains difficult. Machine learning (ML) has been proposed to support CT decision-making, but its feasibility using routinely available pre-CT clinical variables in this specific population remains unclear. This study presents a systematic exploratory benchmarking of ML pipeline configurations for pre-CT tICH prediction in a well-defined retrospective cohort of older emergency patients following LEF. Methods: We performed a secondary analysis from a retrospective observational bicentric study from two university hospital EDs, including 2250 patients aged ≥65 years presenting after an LEF and undergoing cranial CT. Clinical data were extracted manually from electronic health records (EHRs). Eighteen pre-CT clinical features retrieved from electronic health records were selected based on routine availability and ≤10% missingness. Overall, 1224 valid ML pipeline configurations, combining nine classification algorithms, six imputation strategies, four class-balancing approaches, and optional hyperparameter tuning, were evaluated using 10-fold stratified cross-validation on a training set. The 20 highest-ranked configurations by cross-validation AUC were then assessed on a previously inspected exploratory hold-out test set (n = 563); training-derived rule-out operating points were evaluable for 17 of these 20, as three tuned SVM configurations lacked stored out-of-fold predictions. Results: tICH prevalence was 7.0% (n = 158). Across the 20 highest-ranked configurations, hold-out AUC ranged from 0.517 to 0.585, with Matthews correlation coefficient near zero and balanced accuracy of approximately 50% throughout, indicating differences in operating point rather than in discriminative ability. Some of these top-ranked pipelines reached higher cross-validation AUC (up to 0.679) but detected no cases at the default 0.5 threshold—an effect of the decision threshold under class imbalance rather than of the models’ rank-order discrimination, which was itself limited (hold-out AUC of 0.517–0.585). Conclusions: Despite comprehensive exploratory benchmarking across 1224 ML pipelines, routinely available pre-CT clinical features did not provide sufficient discriminatory signal to develop a clinically useful tICH prediction model in this cohort of CT-imaged older adults following LEF. These findings indicate that none of the evaluated configurations produced clinically adequate performance; this near-chance result persisted across all pipelines and most plausibly reflects a combination of limited feature signal, low outcome prevalence, and a sample size below the level required for reliable model development at this event rate. Future studies should target substantially larger prospective multicenter cohorts and evaluate additional feature domains, including structured clinical examination findings, point-of-care biomarkers, and imaging features. Full article

The Gulf Cooperation Council (GCC) countries face the challenge of balancing their reliance on hydrocarbon resources with ambitious renewable energy transition goals, including initiatives such as Saudi Vision 2030. Despite these commitments, renewable energy deployment in the region remains relatively limited, highlighting the need to better understand the factors associated with renewable energy consumption. This study investigates the macroeconomic factors associated with renewable energy consumption in the six GCC countries over the period 2000–2024 using a hybrid methodology combining panel econometric methods and machine learning. Panel data were compiled from the World Bank and the International Energy Agency. The econometric results identify research and development (R&D) expenditure and trade openness as the two most important predictors of renewable energy consumption, jointly explaining approximately 63% of the variation (adjusted R² = 0.629). The random forest model supports these findings by ranking R&D expenditure as the most influential predictor, followed by trade openness. GDP, foreign direct investment, and inflation were not found to be significantly associated with renewable energy consumption in the final models and the structural characteristics of GCC economies. The random forest model also achieved an out-of-sample R-squared of 0.425 with a low RMSE (0.032), suggesting satisfactory predictive performance for an initial model. Sub-period analysis further suggests that the associations of R&D expenditure and trade openness strengthened after 2015, coinciding with the implementation of national energy transition strategies across the GCC. These findings suggest that technological innovation and economic openness may support the region’s energy transition, with greater investment in R&D and stronger international trade integration potentially facilitating renewable energy adoption. Full article

Background: Lung cancer remains the leading cause of cancer-related mortality, with non-small cell lung cancer (NSCLC) accounting for approximately 85% of cases. Over the past decade, immune checkpoint inhibitors have become a core component of first-line treatment for advanced-stage NSCLC lacking driver mutations. Programmed death-ligand 1 (PD-L1) expression is currently used as the standard biomarker, yet its predictive value remains limited, and in most immunotherapy trials, treatment efficacy has been observed independently of PD-L1 expression status. Enhancer of zeste homolog 2 (EZH2), an epigenetic regulator, promotes immune escape by suppressing antigen presentation and impairing CD8+ T-cell function, thereby generating an immune-cold tumor microenvironment, positioning it as a promising candidate biomarker. Methods: We retrospectively analyzed 102 NSCLC patients treated with immunotherapy at a single center between 2018 and 2024. EZH2 expression was assessed by immunohistochemistry. A cohort-derived 25% threshold was used for the primary exploratory analysis, and the analyses were repeated using a 50% threshold as a sensitivity analysis. Patients were classified as EZH2-high (45.1%) and EZH2-low (54.9%) at the 25% threshold. Results: At the 25% threshold, objective response rate (ORR) was 53.6% in the EZH2-low group and 45.7% in the EZH2-high group (Fisher’s exact p = 0.551), while disease control rate (DCR) was 64.3% and 60.9%, respectively (p = 0.837). Median overall survival (OS) was 37 versus 27 months (log-rank p = 0.323), and median progression-free survival (PFS) was 15 versus 12 months (p = 0.387). No significant correlation was found between EZH2 and PD-L1 expression (r = 0.167, p = 0.138). In treatment-line-adjusted Cox models, EZH2 expression was not associated with OS (hazard ratio (HR) 0.953, 95% confidence interval (CI) 0.533–1.704; p = 0.870) or PFS (HR 0.996, 95% CI 0.573–1.733; p = 0.989), whereas squamous histology was an independent predictor of survival. Results remained non-significant at the 50% threshold. Early progression was uncommon and did not differ significantly by EZH2 status overall or within PD-L1 strata. Conclusions: In this real-world cohort, EZH2 expression was not independently associated with response, early progression, OS, or PFS, and showed no correlation with PD-L1. These exploratory findings do not support the clinical use of EZH2 as a biomarker at this stage; prospective, multicenter studies using predefined thresholds and standardized immunohistochemical methods are needed to clarify its potential role as a marker complementary to PD-L1. Full article

To investigate the development characteristics of winding interturn insulation partial discharge in a full-scale converter transformer, an interturn insulation defect specimen was installed inside a ±400 kV full-scale converter transformer. A partial-discharge test was conducted under stepwise alternating-current (AC) voltage, while pulse-current, high-frequency, ultra-high-frequency, ultrasonic, sound-pressure, and vibration signals were acquired synchronously. To reduce the subjectivity of discharge-stage classification, a dimensionless discharge-development index was constructed by integrating the 95th-percentile apparent charge, discharge pulse count, and phase occupancy, and the stage-transition points were identified using piecewise-linear change-point analysis. Two change points at approximately 1.9 and 4.3 min divided the discharge process into the inception, development, and severe stages. The apparent charge increased from approximately 1.5 × 10² pC in the inception stage to the order of 10³ pC in the development stage, and then rose sharply to approximately 1.6 × 10⁵ pC in the severe stage. Meanwhile, the discharge activity changed from an intermittent low-intensity state to a continuous high-intensity state. After the test, insulation-paper ablation and carbonization, conductor exposure, and a continuous breakdown channel were observed in the defect region. These post-test observations confirm severe final damage to the interturn insulation, although the time sequence of the individual damage features could not be determined from the final morphology alone. Under the present sensor arrangement, the HF channel provided the earliest identifiable response at approximately 1.9 min, whereas the UHF and internal ultrasonic channels supplied complementary evidence during subsequent discharge development. The external ultrasonic, sound-pressure, and vibration responses were more strongly affected by propagation paths, structural coupling, sensor location, and background disturbance and were therefore treated as supplementary indicators. These findings support a staged multisensor interpretation strategy for interturn partial discharge in a full-scale converter transformer rather than a universal ranking of sensor performance. Full article