Search Results (2,095)

Accurate forest soil organic carbon (SOC) monitoring is essential for forest soil quality assessment and carbon-sink evaluation. Visible-near-infrared (Vis–NIR) hyperspectral sensing provides rapid and information-rich measurements for SOC prediction, but obtaining sufficiently large labeled soil-spectral datasets remains difficult because field sampling, sample preparation, and reference SOC determination are labor and time intensive. This study developed a latent conditional diffusion-based data augmentation framework for SOC prediction from hyperspectral sensor data. A total of 248 forest red-soil samples from Guangxi, China, were measured using laboratory Vis–NIR reflectance spectroscopy over 350–2500 nm and divided by the Kennard-Stone algorithm into a 174-sample modeling set and a fixed 74-sample validation set. Four generative models, including VAE, GAN, WGAN-GP, and the proposed hyperspectral latent conditional denoising diffusion implicit model (HsDDIM), were evaluated using spectral visualization, t-SNE distributions, maximum mean discrepancy (MMD), Fréchet Inception Distance (FID), and downstream prediction performance. Unlike joint spectral-label generation, HsDDIM treats SOC as an external condition and generates spectra in the latent space under specified SOC conditions; the SOC condition itself is not generated by the diffusion process. Among the compared augmentation strategies, HsDDIM showed the closest distributional agreement with the real spectral samples according to MMD and FID, with values of 0.0806 and 0.5281, respectively. Without augmentation, FD1-SVR achieved the best validation result (R² = 0.83, RMSE = 4.71 g kg⁻¹). After 300% HsDDIM augmentation, 1D-CNN achieved R² = 0.91, RPD = 3.39, and RMSE = 3.40 g kg⁻¹. These results suggest that the SOC-conditioned latent DDIM framework can improve small-sample hyperspectral SOC prediction under the present fixed-validation protocol. Full article

A decline in CO₂ emissions is often treated as evidence of successful climate policy, although the same outcome may reflect technological modernization, economic contraction, demographic decline, deindustrialization, or external shocks. Accordingly, the objective is to test whether comparable headline reductions reflect different transition mechanisms when historical drivers, prospective indicative gross carbon-cost exposure, and investment capacity are evaluated jointly. We hypothesize that Germany is closest to managed decarbonization, Poland follows carbon-intensive decoupling, and Ukraine’s decline is predominantly structural or shock-driven and that the joint diagnostic indicates greater contraction risk where multiple investment-capacity measures are weakest, without implying a causal prediction of firm response. This study develops a three-stage diagnostic framework combining additive Logarithmic Mean Divisia Index (LMDI) decomposition, indicative gross sectoral carbon-cost exposure, and multiple macroeconomic investment-capacity indicators to distinguish managed decarbonization from carbon-intensive decoupling and structural or shock-driven emissions decline. The framework is applied to Germany, Poland, and Ukraine for 1990–2024, with subperiods reflecting major structural transformations and shocks. Germany reduced emissions by 482.48 Mt CO₂ despite rising population and GDP per capita, as lower energy and carbon intensity more than offset these upward pressures. Poland reduced emissions by 103.48 Mt CO₂ mainly through lower aggregate energy intensity, while carbon-intensity improvements were more limited. Ukraine recorded the largest decline (564.03 Mt CO₂), but substantial contributions came from lower GDP per capita and population, while carbon intensity increased emissions over the period. At a common benchmark of EUR 75.28/tCO₂, Ukraine has the highest indicative gross exposure among the integrated-steel cases, whereas Poland has the highest indicative gross exposure among the electricity cases. The findings show that emissions outcomes should be interpreted jointly with their drivers, sectoral carbon-price transmission, and the capacity to finance technological response. Full article

Background/Objectives: Reported performance on public brain tumour MRI benchmarks is hard to interpret because of near-duplicate train/test overlap, ImageNet pretraining bias, and single-seed evaluation. We address this with a leakage-aware evaluation protocol and a compact model trained entirely from scratch. Methods: WICA-Net-M is a 2.47 M-parameter CNN whose gated Haar Discrete Wavelet Transform (DWT) separates low- and high-frequency components and fuses them through a learnable gate, complemented by Coordinate Attention. We evaluate it on the standard and image-level deduplicated splits of the Nickparvar brain tumour MRI dataset under a three-seed, leakage-aware protocol, benchmark it against five ImageNet-pretrained baselines and conduct a near-duplicate overlap audit against BRISC 2025. Results: WICA-Net-M reaches 99.42 ± 0.16% accuracy on the standard V1 split and 95.25 ± 0.32% accuracy/95.17 ± 0.31% macro F1 on the deduplicated V2 split, closely matching five ImageNet-pretrained baselines (95.17–95.67%) with fewer parameters, with sub-half-point differences across three seeds indicating comparable rather than superior accuracy. The audit identifies 861 exact SHA-256 pairs involving 857 of the 1000 BRISC test images against the full Nickparvar collection. Alongside perceptual-hash candidate pairs, this exact overlap shows that BRISC cannot serve as independent external validation. Conclusions: The descriptive 4.17-point V1-to-V2 difference reflects the combined effects of duplicate removal, class rebalancing, and altered sample composition, with none separable from public releases, though it shows that a scratch-trained compact model can approach pretrained performance on the controlled split. Patient-level leakage remains unresolved because patient identifiers are unavailable. Leakage-aware, multi-seed evaluation should be standard before clinical translation. Full article

Cross-event fire-severity mapping requires a model to delineate the burned footprint and grade severity within it across wildfires whose spectral expression varies with vegetation and observation conditions. Nominal multiclass models treat the classes as parallel alternatives and leave the order implicit. We introduce Co-Burn, a bi-temporal Siamese model that adds a pre-to-post dNBR channel to the post-fire branch as an NIR-SWIR change anchor and uses a conditional ordinal head to estimate burn presence before high-severity assignment. Ten methods were compared across 14 New South Wales wildfires against a Sentinel-2 FESM-derived three-class target, with 4 complete fires held out from model development and selection. Co-Burn ranked first under both pixel-pooled and event-mean aggregation, reaching 0.520 ± 0.017 and 0.524 ± 0.028 external burned mIoU. Event-level factorial contrasts showed that burned-mIoU effects varied among fires, while dNBR anchoring reduced false-burn rate on all four external events. At the fixed operating point, Co-Burn assigned 19.9% of reference-unburned pixels to burned classes, against 28.9% for the reflectance-only nominal variant. On the hardest held-out fire, limited false-burn expansion coexisted with a downward shift across the ordered severity classes. Co-Burn supports ordered three-class mapping of previously unseen forest fires before target-fire labels become available. Full article

Invasive candidiasis (IC) causes substantial mortality in critically ill children, yet pediatric data from Southeast Asia remain limited. We retrospectively studied 117 children aged ≤ 18 years with proven IC at a Thai tertiary center (2009–2023). Thirty-day mortality was 25.6%, rising to 47.1% in neonates; death occurred a median of 5.5 days after diagnosis. Firth penalized logistic regression identified septic shock (aOR, 4.89; 95% CI, 1.77–14.88) and thrombocytopenia (aOR, 3.74; 95% CI, 1.17–15.35) as independent mortality predictors, with septic shock remaining significant across all analytical frameworks, including Fine–Gray competing-risks analysis. Apparent mortality associated with absence of antifungal therapy reflected reverse causation: in all six untreated children who died, Candida was reported only 2–7 days after death. Candida albicans (43.6%), C. tropicalis (30.8%), and C. parapsilosis (24.8%) predominated; C. glabrata was absent. Non-albicans Candida was already common at the outset and showed no statistically detectable increase over 15 years. Among 45 consecutive pediatric bloodstream isolates (2021–2026), amphotericin B and echinocandins largely retained activity, whereas reduced azole susceptibility was concentrated in C. tropicalis (47.8% fluconazole-susceptible; 26.1% posaconazole wild-type). The reduced azole susceptibility of local C. tropicalis isolates argues for periodic reassessment of institutional susceptibility data rather than reliance on historical or external epidemiology; in comparable settings, an echinocandin or amphotericin B is a more reliable empiric choice than an azole, pending species identification and susceptibility results. Full article

Background/Objectives: Video laryngoscopy improves the glottic view, yet difficult intubation still occurs. Whether it reflects a poor view or arises despite an adequate one is rarely quantified in otorhinolaryngological surgery. We described this pattern and its preoperative predictors. Methods: In a single-center retrospective cohort of 1932 adults undergoing otorhinolaryngological surgery with non-channeled video laryngoscopy (2018–2021), difficult intubation was defined a priori as two or more laryngoscopic attempts (blade insertions), a lower bound on difficulty. The glottic view (Cormack–Lehane grade) and intubation time described the difficulty, and predictors were identified by logistic regression. An unweighted count of four anatomical bedside factors, selected on rationale (not data-driven significance) at pre-existing clinical thresholds, was the primary risk score; a five-factor version added a resident operator. Results: Difficult intubation occurred in 120 patients (6.2%; 95% CI 5.2–7.4). Most occurred despite an adequate view (64% at Cormack–Lehane grade I–II; 45% at grade I alone); the rate was 4.3% (95% CI 3.5–5.4) versus 29.3% (95% CI 22.5–37.1) for good versus poor views (odds ratio 9.2, 95% CI 6.0–14.0), and difficult intubations took about twice as long (median 90 versus 40 s). Independent predictors were shorter inter-incisor and thyromental distances and a resident operator. The four-factor count graded risk from 4.6% to 12.9% (odds ratio 1.63 per factor; apparent area under the curve 0.596, 95% CI 0.546–0.645, not corrected for optimism; positive predictive value 8.8% at one or more factors), and a five-factor count performed similarly (0.631). Within the good-view subgroup, difficulty appeared to relate more to operator inexperience than to anatomy (adjusted odds ratio 2.48 versus 1.13). Conclusions: Difficulty was far more likely with a poor view, but because a good view occurred in 92.4% of patients, most difficult intubations arose despite an adequate one. We hypothesize that difficulty then lies in tube delivery rather than visualization; because the analysis conditions on the view, this needs prospective testing. Discrimination was modest and the count exploratory; these single-center findings require external validation. Full article

Large language model (LLM)-based artificial intelligence is increasingly used in ethically consequential human decision-making, yet fully autonomous machine ethics remains unrealistic, motivating architectures that support rather than replace human ethical judgment. This study introduces the Cognitive–Reflective Equilibration Architecture (CREA), a cognitively grounded artificial moral advisor that operationalizes the Cognitive–Reflective Equilibration Model (CREM), in which reflective reasoning guides ethical judgment from intuitive cognition toward a more advanced equilibrium among competing values, drawing on Piaget and Rawls. CREA implements CREM’s 20-step process through four stage-aligned reasoning agents—Cognitive, Reflective, Equilibration, and Evaluation—coordinated via multi-LLM orchestration, in which auxiliary models independently explore principles, generate counterarguments, and score supporting and opposing considerations to externalize reflective deliberation. The architecture was empirically evaluated by comparing four configurations—single-agent, multi-agent, multi-LLM, and multi-LLM with knowledge- and reasoning-bank augmentation—across four indicators of advice quality using 500 matched execution units per configuration. All comparisons are system-internal: advice quality was scored by CREA’s own multi-LLM measurement pipeline rather than by human ethicists, so the findings reflect relative differences among architectures under LLM-based self-evaluation, not normative validity. Within that scope, distributing reflective reasoning across multiple models was associated with higher reason-giving (justifiability) and normative-alignment scores relative to simpler configurations. CREA therefore offers an empirically characterized, auditable advisor architecture whose potential to scaffold human ethical judgment remains a hypothesis for user-centered validation rather than a demonstrated outcome. Full article

Cellular senescence (CS) is increasingly recognized as an important cell-state programme involved in breast cancer progression and therapeutic response, but its context-dependent molecular heterogeneity limits its application in prognostic assessment. In this study, GeneCompass-based all-gene in silico perturbation analysis was performed to identify candidate genes predicted to induce senescence or rejuvenation, thereby expanding the known senescence-related gene set. Machine learning further established a seven-gene prognostic signature that may serve as an adjunctive tool for prognostic assessment across multiple cohorts. The time-dependent AUCs at 1, 3, and 5 years were 0.707, 0.700, and 0.684 in the training cohort; 0.657, 0.661, and 0.629 in the test cohort; and 0.611, 0.646, and 0.637 in the external validation cohort, respectively. Single-cell and spatial transcriptomic analyses suggested that the risk component of the prognostic signature reflects not only malignant epithelial cell states but also stromal–vascular remodelling in the tumor microenvironment. Among the signature genes, ADGRF5 exhibited the most pronounced expression alteration, and its knockdown suppressed malignant phenotypes in breast cancer cells. These findings provide an AI-assisted strategy for senescence biomarker discovery and highlight ADGRF5 as a candidate functional risk gene associated with breast cancer progression. Full article

Oil palm plantations generate substantial biomass and by-products with potential to support the circular bioeconomy and crop–livestock integration. This study assessed biomass and nitrogen flows across three oil palm–cattle integration scenarios using material-flow analysis and nitrogen flow diagrams. Field observations were conducted in April 2026 in a seven-paddock grazing area. Scenarios A and B reflected existing conditions, whereas Scenario C represented a potential model incorporating assumed by-product utilization pathways. Cattle grazing was associated with an estimated 43.2% reduction in understory biomass relative to the non-integrated oil palm plantation. Under the assumed pathways, Scenario C indicated greater material-flow connectivity and nitrogen recycling, with the highest total N input, recycled N, and N recycling rate at 241 kg N/ha/year, 54.5 kg N/ha/year and 22.6%, respectively, while reducing unused oil palm by-products through palm kernel meal utilization. External N input remained 180 kg N/ha/year across scenarios, while product N output increased from 67.2 kg N/ha/year in Scenario A to 71.1 kg N/ha/year in Scenario C. Overall, cattle integration increased biomass utilization and internal nitrogen recirculation, although substantial N surpluses, estimated N₂O-related emissions, and unutilized residues remained. These findings indicate opportunities to improve nutrient recovery and by-product utilization without implying reduced dependence on external fertilizer inputs. Full article

This study proposes a 2.45 GHz low-power microwave drilling system for precise bone drilling near sensitive organs with a tunable parallel metal impedance-matching sleeve. The system is based on a quarter-wavelength coaxial resonant cavity integrated with a movable parallel impedance tuning sleeve, which can mitigate impedance mismatch caused by dielectric variations in ablated tissue during drilling. To further ensure matching accuracy and stability, a solid-state source is employed to track the minimum reflection frequency within the operating bandwidth. Broadband electromagnetic and transient thermal simulations are conducted to evaluate the impedance response and thermal characteristics. A prototype microwave drilling system is fabricated and experimentally validated in fresh cartilage tissue, fresh cortical bone, and cooked bovine femur samples. The results demonstrate that the proposed tuning sleeve and adaptive source effectively regulate the response to the reflection coefficient. At 2.45 GHz and an input power of 20 W, the proposed system achieves a measured reflection coefficient below −12 dB and produces a localized drilled region within 10 s. Compared with previously reported microwave drilling systems operating at power levels of up to approximately 200 W and recent microwave drilling studies typically employing tens to hundreds of watts, the proposed system demonstrates bone drilling at a substantially reduced input power, while the integrated tunable sleeve provides structural impedance matching without relying solely on conventional external impedance tuners. Full article

Background: Severe COVID-19 reflects a multi-layered host response with systemic inflammation, thrombo-inflammation, tissue damage, oxidative stress and altered antiviral interferon biology. We performed an integrated exploratory analysis of first-wave hospitalized patients to identify biomarker patterns associated with adverse clinical evolution during established admission. Methods: We analyzed 60 hospitalized COVID-19 patients and 18 healthy controls for soluble IFNAR2 (sIFNAR2) comparison. Biomarker samples were obtained during hospitalization, approximately seven days after symptom onset. Outcomes were final clinical status, severe respiratory involvement, post-sampling clinical worsening and death. Analyses included non-parametric testing, false-discovery-rate adjustment, effect-size estimation, exploratory ROC curves, parsimonious regression, penalized internal validation, composite scores and molecular-structure analyses. Results: Final status was favorable outcome in 31 patients, severe non-fatal disease in 22 and death in 7. sIFNAR2 was higher in patients than in healthy controls and highest among non-survivors, but did not distinguish favorably from severe non-fatal disease. The most consistent severity-associated signals were IL-6, D-dimer, total thiols, IL-10, LDH, ferritin, leukocytes and IL-1RA. D-dimer, IL-6 and ferritin yielded the largest exploratory univariable AUCs for severe respiratory involvement, whereas ferritin, IL-10, IL-1RA and sIFNAR2 predominated in event-limited mortality analyses, which were based on only seven deaths. Composite multi-axis scores and PLS-DA were tools requiring external validation. Conclusions: Biomarker patterns during admission were associated with adverse evolution. Conventional markers remained the most practical signals, while cytokines, redox markers and sIFNAR2 provided complementary biological information. sIFNAR2 should be interpreted as an exploratory complementary marker of the interferon receptor axis, mainly linked to mortality, not as a stand-alone clinical test or functional measure of IFNAR signaling. Full article

Background: Transparent measurement is essential when organizational capacity is compared across public health institutions. The supplied analytic dataset did not contain repeated annual scores or indicator-level evidence for the previously proposed five-domain maturity model. We therefore conducted a reproducible cross-sectional analysis limited to the variables available in the dataset. Methods: Twenty Romanian public health directorates (PHDs)—10 in counties with a university medical center (CU) and 10 in counties without one (FU)—were compared using structural indicators, eight binary COVID-19 re-sponse measures, and a composite capacity–response index. The index was the equally weighted arithmetic mean of four sample-normalized components: employees per 100,000 inhabitants, budget per capita, previous-crisis experience, and the number of documented COVID-19 measures. Results are presented as medians and interquartile ranges. Mann–Whitney U and Fisher exact tests were used, with Holm correction within comparison families. Results: The composite index was higher in CU than FU institutions (55.0 [38.0–66.3] vs. 22.0 [19.3–31.3]; U = 93.0; p = 0.0013; Holm-adjusted p = 0.0077; rank-biserial r = 0.86). CU institutions also had larger populations, more employees, larger total budgets, higher per capita budgets, more previous-crisis experience, and more documented response measures. Employees per 100,000 inhabitants did not differ significantly. After multiplicity correction, vaccination-center involvement and communication/education were more frequently documented in the CU group. Digital platforms were recorded for only two institutions and cannot be interpreted as a digital-maturity domain. Conclusions: The available data support descriptive evidence of institutional disparities associated with the university-center context, but they do not support causal inference or validation of a maturity model. The composite index should be treated as an exploratory, sample-dependent benchmarking measure pending prospective instrument development, documented scoring rules, independent assessment, and external criterion validation. Its group difference largely reflects the four constituent components, which should remain the primary basis for interpretation. Full article

Baseflow cannot be measured directly, so many separation methods exist, and they disagree. We review 16 methods spanning digital filters, graphical partitioning, recession analysis, and conductivity mass balance (CMB). We show that most recursive filters are special cases of a generalized three-parameter equation, which separates linear reservoir from signal processing families and identifies the Boughton and Eckhardt filters as algebraically equivalent. All 16 are implemented in baseflowx, an open-source Python (version 3.9 or later) package, with eight of them validated against independently published results. The methods rank the 398 reference catchments similarly but differ in level: pairs correlating above 0.93 diverge by up to 0.22 in mean absolute baseflow index (BFI). We compared 15 streamflow-only methods with a CMB reference at 31 screened catchments, with 24 more as a robustness check. All estimated larger baseflows than CMB on average, reflecting the quantities separated: hydrograph-based methods count bank storage return, slow interflow, and other delayed water as baseflow, whereas CMB partitions by conductance. CMB-derived ${\mathrm{BFI}}_{\max }$ values were below the conventional 0.80 Eckhardt default at all 55 catchments. Parameter choice also produced most of the variation attributed to method choice. Streamflow-only methods can support relative comparisons and trends but cannot constrain an absolute baseflow fraction without an external reference. Full article

Volcanic reservoirs exhibit strong lithological heterogeneity and complex seismic responses, making lithofacies prediction between wells challenging, particularly in the Yingcheng Formation of the Songliao Basin. To address this issue, this study proposes an integrated workflow for volcanic lithofacies prediction based on Principal Component Analysis (PCA)-optimized multi-attribute seismic clustering. Seismic facies are first identified from reflection configuration and external geometry, and three types—chaotic, layered, and shield-like facies—are established and calibrated using well logs and core data. PCA is then applied to reduce attribute redundancy and optimize the attribute set. The selected attributes, including root mean square (RMS) amplitude, energy half-life, and gradient magnitude, are used for multi-attribute clustering. The results indicate that eruption facies dominate the study area and correspond to favorable reservoirs, accounting for the majority of high-quality reservoir zones. In contrast, overflow and volcanic sedimentary facies show comparatively lower reservoir potential. The predicted lithofacies distribution shows strong spatial consistency with well observations, demonstrating the method’s reliability. Overall, the proposed workflow improves lithofacies prediction accuracy and provides an effective tool for reservoir characterization and well deployment in complex volcanic settings. Full article

Digital transformation has come to the fore as a pivotal force behind corporate green innovation in the digital economy. Although previous research has documented that digital transformation can promote green innovation, the underlying organizational mechanisms remain theoretically fragmented. Moreover, the external conditions that shape the effectiveness of this relationship are not yet fully understood. To address these gaps, we develop an integrated analytical framework anchored in Organizational Information Processing Theory (OIPT) and the Resource-Based View (RBV). This framework explains how digital transformation promotes green innovation through two complementary organizational mechanisms—resource investment and technology application—and incorporates the moderating role of digital infrastructure. Using panel data on Chinese A-share-listed firms covering 2015–2024, this study examines the proposed relationships through a Bidirectional Fixed Effects Model, mediation analysis, and moderation analysis. The findings demonstrate a significant positive association between digital transformation and green innovation. This effect is partially mediated by increased R&D(Resource and Development) investment and deeper digital organizational embedding, which constitute complementary pathways reflecting resource investment and enhanced resource utilization, respectively. Additionally, digital infrastructure positively moderates this relationship by providing a more supportive external digital environment. This study enriches the existing literature by synthesizing previously dispersed perspectives on mechanisms into an OIPT-RBV framework, offering a more comprehensive account of how and under what circumstances digital transformation facilitates green innovation. Our findings also carry practical implications for managers seeking to strengthen innovation capabilities and for policymakers aiming to advance digital infrastructure and firms’ sustainable development. Full article