Browse Articles
Discover research articles across all indexed journals
Pre-existing venous structural remodeling and arteriovenous fistula maturation assessed by conventional histology and nonlinear optical microscopy
Ultrasound-promoted Cu/ZnO-graphene oxide catalyzed synthesis of spirooxindole-fused quinazolines and evaluation of their antioxidant and antibacterial activities
CNN-transformer-based model explained by SHAP and multi-head attention weights for time series forecasting
Abstract Convolutional Neural Networks (CNNs) and transformer architectures offer strengths for modeling temporal data: CNNs excel at capturing local patterns and translational invariances, while transformers effectively model long-range dependencies via self-attention. This paper proposes a hybrid architecture integrating convolutional feature extraction with a multi-head attention backbone to enhance multivariate time series forecasting. The CNN module first applies a hierarchy of one-dimensional convolutional layers to distill salient local patterns from raw input sequences, reducing noise and dimensionality. The resulting feature maps are then fed into the prediction model, which applies multi-head attention to capture both short- and long-term dependencies and to weigh relevant covariates adaptively. We evaluate the proposed CNN-transformer-based model on a hydroelectric natural flow time series dataset. Experimental results demonstrate that the proposed model outperforms well-established deep learning models, with a mean absolute percentage error of up to 2.2%. The explainability of the model is obtained by a proposed SHapley Additive exPlanations (SHAP) and Multi-Head Attention Weights (MHAW). Our novel architecture, named CNN-Transformer-SHAP-MHAW, is promising for applications requiring high-fidelity and multivariate time series forecasts.
Investigation of the marginal impact of electricity generation sources on CO2 emissions in highly emitting countries by daily data and KRLS approach
Abstract Considering countries’ efforts to combat climate change and the critical role of energy use in carbon dioxide (CO 2 ) emissions, this study empirically analyzes the impact of electricity generation (EG) sub-types on CO 2 emissions. In this vein, the study focuses on the five highly emitting countries (namely, China, the United States, India, Russia, & Japan), uses daily data from 1st January 2019 to 30th June 2024, and implements a kernel regularized least squares (KRLS) approach to consider average and marginal impacts in empirical analyses. The empirical outcomes demonstrate that (i) gas and nuclear EG decreases CO 2 emissions for India; hydro EG is curbs CO 2 emissions in India and Japan; oil EG decreases CO 2 emissions at lower percentiles in the USA; (ii) coal EG cause an increase in CO 2 emissions in the all countries; (iii) wind EG provides a decrease in CO 2 emissions at lower percentiles in the USA and Japan, whereas its effect becomes ineffective across remaining percentiles and countries; (iv) solar EG is the unique source that has a declining impact across countries; (v) the KRLS approach has a high estimation capacity (R 2 : 93.53%). Overall, the study reveals average and marginal impacts of EG subtypes on CO 2 emissions across percentiles and countries. Accordingly, through consideration of empirical outcomes (e.g., solar EG is the most beneficial EG source for all countries) while some other EG sub-types are partially beneficial (e.g., gas EG in China and India; hydro EG in all countries; nuclear EG in all countries except Russia; wind EG in United States, India, and Japan), the study argues various policy implications (e.g., prioritizing the most helpful EG sources in providing further governmental support, focusing on first solar EG and then followed by hydro, nuclear, and wind EG to stimulate new capacity installation).
Meteorological modulation of ambient air quality in Delhi and Lahore during the November 2024 haze event
Abstract The Indo-Gangetic Plain in South Asia frequently experiences severe haze during the post-monsoon season. The November 2024 haze event recorded the highest daily PM 2.5 concentrations in Delhi (696 µg/m³) and Lahore (621 µg/m³) over the past 5 years. These elevated PM 2.5 concentrations coincided with a prolonged period of regional-scale atmospheric stagnation and a shift in peak fire activity from the afternoon to the evening. Here, we show that both cities had elevated pollutant concentrations with correlations between PM 2.5 and planetary boundary layer height, lower-tropospheric stability, atmospheric heat deficit, relative humidity, and fire radiative power, consistent with accumulation of emissions from multiple sources, including regional biomass-burning emissions, under stagnant atmospheric conditions. We also highlight that positive geopotential height anomalies and the mechanism of anticyclonic stagnation and weak winds favored a high-pressure system, suppressing turbulent ventilation and accumulating pollution in the lower troposphere. Furthermore, air mass back trajectories overlapped the stubble-burning region ,and higher evening fire counts observed by a geostationary satellite emphasize the likely role of increased evening fire activity alongside other local and regional emissions in elevated air pollution concentrations. These results demonstrate that local and synoptic meteorology strongly aggravates haze events, suggesting emission control strategies should be guided by meteorological forecasts across regional airsheds.
Longitudinal trends in modifiable cancer risk factors in the Generations Study cohort in the United Kingdom
Abstract Understanding how modifiable cancer risk factors change across the life course is crucial for evidence-based prevention. However, surveillance of lifestyle factors often relies on cross-sectional studies, unable to capture within-person patterns across the life course. Longitudinal studies assessing long-term exposure patterns are scarce. Within the Generations Study, a prospective cohort of > 113,700 UK women (median recruitment age: 48.4 years; range: 16–102) recruited between 2003 and 2009, we explored life course trajectories of modifiable cancer risk factors, leveraging up to six self-reported exposure measurements collected retrospectively and prospectively. Linear mixed-effects models with natural cubic splines for age were fitted for alcohol consumption, smoking, physical activity and body mass index (BMI). Mean velocity curves were derived from differentiated spline curves. Interactions between age and birth cohort assessed generational differences. Each risk factor showed distinct, nonlinear patterns in the level and rate of change across the life course. Alcohol consumption, smoking and physical activity peaked in late adolescence, whilst BMI increased with age, peaking in the late 60s. Younger cohorts (born 1960–2003) exhibited progressively higher BMI trajectories and earlier peaks in alcohol consumption and smoking intensity compared to older cohorts (born 1908–1959). Modifiable cancer risk factor exposure varied throughout life, highlighting the importance of repeated exposure measurements to capture these patterns. Our findings revealed key life stages for behavioural transitions. This insight can guide targeted prevention and intervention efforts.
Research on the mechanical properties and modification mechanism of loess treated by consolid system
Effective removal of methyl violet dye by montmorillonite-incorporated carboxymethyl cellulose-g-poly(methacrylic acid-co-acrylamide) nanocomposite hydrogel
Quality and reliability of videos about pneumoconiosis on TikTok and Bilibili in China
The integrating analytical framework of the GxE interactions in tomato: a comprehensive multi-analytical approach toward across-elevations
Clinico-demographic and molecular characteristics of familial Mediterranean fever in Egypt: a multicenter study
Abstract Familial Mediterranean Fever (FMF) is an inherited autoinflammatory disease characterized by recurrent episodes of fever and serositis caused by mutations in the MEFV gene. FMF primarily affects individuals of Mediterranean ancestry. Typical manifestations include short-lasting attacks of abdominal pain, chest pain, and arthritis. Long-term complications, such as amyloidosis, might be prevented with colchicine treatment. This study aimed to assess the clinical, demographic and molecular features of FMF in Egyptian patients. This multicenter prospective study included 280 clinically suspected FMF patients referred to the outpatient clinics of participating centers. Patients were enrolled based on recurrent episodes of fever and abdominal pain suggestive of FMF. The diagnosis of FMF was established based on the Eurofever/PRINTO criteria, in addition to molecular genetic confirmation of MEFV mutations. Patients who did not meet diagnostic criteria were excluded. The study included 173 female patients, and 107 male patients with age range from 2 up to 60 years. All patients were descending from different families. Parental consanguinity was found in 24.6% while positive family history was seen in 33.2%. The duration of the attack in almost two third of the patients was less than 48 h, only 6% of the patients suffered attacks longer than 72 h. About half the patients achieved a marked reduction or complete cessation of FMF attacks, along with normalization or significant reduction of inflammatory markers (e.g., serum amyloid A), following colchicine therapy at a daily dose of 1.5–3 mg., only 6% needed higher doses. The initial serum amyloid A was normal in 47.5% in the patients and elevated in the remaining patients. Fever was documented in only 19.6% of patients at presentation; abdominal pain was among the most common presentation, seen in 62.1%. The most frequently observed MEFV allele was E148Q (143, 39.5%) followed by M694I ( n = 59, 16.3%), A744S and V726A ( n = 44, 12.2% for each), then M680I ( n = 35, 9.7%). In our multicenter study focused on Egyptians, 93.6% of the enrolled FMF patients carried at least one MEFV variant. The E148Q allele was the most frequently observed followed by M694I. The study also revealed that Egyptian patients exhibited a mild form of the disease with a female predominance. This mild presentation might be attributed to the high E148Q prevalence and low rate of amyloidosis in our cohort.
A Transformer-Based Deformable Convolution UNet for Adaptive Arbitrary Style Transfer
Abstract In recent years, image style transfer has matured significantly in the field of computer vision. However, current methods for image style transfer still face the challenge of balancing between content edge contours and style texture strokes, as it is difficult to control the degree of stylization. To address this issue, we propose a UNet based on Transformer feature fusion, named Trans-DCUNet. The network adaptively integrates content and style features by taking advantage of the self-learning characteristics of the cross-attention mechanism in the Transformer. The network combines Transformer and CNN, which takes advantage of the global context capture ability of Transformers and the local modeling ability of convolutional neural networks to fully learn image features at multiple levels, thereby generating stylized images. To enhance texture generation, we introduce a deformable convolutional residual module, which allows the convolution kernel to adapt to varying image features, capturing fine texture details more effectively. Additionally, we augment the traditional perception loss with edge detection loss and frequency perception loss, aiming to better preserve the edge contours of the content image and learn the texture strokes of the style image. Our experiments were conducted on the Microsoft COCO and WikiArt datasets. Experimental results show that our method achieves a content retention SSIM of up to 0.8655 and a style similarity LPIPS of 0.5655, outperforming most competing methods, while generating more artistic stylized images with significantly improved visual effects.
Improved accuracy of PCG signal classification for myocardial infarction biomarker using automatic feature selection and boosting process
Abstract Myocardial infarction (MI) is a leading global health concern, typically diagnosed using ECG, biomarkers, or imaging, which costly or unavailable in low-resource settings. This study presents a non-invasive, machine learning-based approach using phonocardiogram (PCG) signals for classifying normal, ST-elevation MI (STEMI), and non-ST-elevation MI (NSTEMI). The proposed approach leverages the acoustic signatures of cardiac mechanical activity, capturing subtle variations in heart sound morphology and timing associated with ischemic myocardial dysfunction. The processing pipeline included PCG acquisition via electronic stethoscope, band-pass filtering, envelope-based segmentation, feature extraction, and selection using Mutual Information and K-best ranking. We evaluated the method using a diverse dataset of 104 subjects from Indonesia and Japan to ensure generalizability across ethnic and physiological variations. Eighteen key features with mutual information values up to 0.82 bits were used to train AdaBoost and Gradient Boosting models. Without parameter tuning, these models achieved 88.30% and 93.00% accuracy, respectively. After optimization, AdaBoost reached 94.00% accuracy, and Gradient Boosting achieved 98.30% accuracy and a 95.10% F1-score. These results outperform previous bagging-based methods of 86.00% accuracy, demonstrating improved accuracy and robustness. This work highlights the potential of PCG-based MI detection as a low-cost, non-invasive diagnostic alternative, particularly valuable for early screening and triage in resource-limited healthcare settings.
Propofol maintains more stable cochlear amplifier responses than isoflurane anesthesia does in a porcine model
Sports injury prevention and management in university physical education: a qualitative study of lecturers and students in Sichuan Province, China
Treatment reliability of solar septic tank systems under comparative field conditions
Adaptive self-supervised knowledge transfer for computationally efficient thoracic disease screening from chest X-ray images
Vitamin a modulates neurogenesis-associated pathways and cholinergic signaling in Alzheimer’s disease: potential role of reactive astrocytes via NGN2/SOX-11 and SIRT-1
Abstract Alzheimer’s disease (AD) is a progressive neurodegenerative disorder lacking effective disease-modifying therapies. A promising regenerative approach involves enhancing endogenous neurogenic capacity within the injured brain. Reactive astrocytes—stellate-like cells in the AD brain—may contribute to a pro-neurogenic environment through transcription factors (TFs) such as neurogenin 2 (NGN2) and SOX-11. This process is tightly regulated by epigenetic mechanisms, particularly SIRT-1, a neuroprotective histone deacetylase that modulates TF activity and neuronal fate. Vitamin A (V A ), a key regulator of differentiation and epigenetic remodeling via its active metabolite retinoic acid, is stored in astrocytes and hepatic stellate cells (HSCs). We hypothesized that AD-related astrocyte activation depletes cerebral V A , mobilizes hepatic stores, contributes to liver fibrosis, and that V A supplementation may restore astrocytic function, activate endogenous TFs via SIRT-1, and drive cholinergic neuron regeneration. In a scopolamine (SCO)-induced AD rat model, V A biodistribution was traced using confocal microscopy. Brain and liver V A deficiency were confirmed via retinol-binding protein (RBP) and ALDH1A1 expressions. Rats received V A (1500, 3000, or 4500 IU/kg/day) or donepezil. Outcomes included neurogenesis (DCX), NGN2/SOX-11 expression, SIRT-1 activation, cholinergic regeneration, amyloid-β deposition, and serum tau. Liver fibrosis was assessed via TGF-β, hydroxyproline and histopathologically. AD induced systemic V A depletion and liver fibrosis. Medium-dose V A (VAMD) significantly enhanced neurogenesis, TF expression, SIRT-1 activation, cholinergic regeneration, and reversed liver fibrosis. VAMD demonstrated neuroregenerative and antifibrotic effects, indicating a possible therapeutic role in AD.
A hybrid forecasting and fuzzy comprehensive evaluation approach for graded early warning of experimental task frequency
Abstract Accurate prediction and graded warning of experimental task frequencies are critical for resource optimization and operational safety in large-scale scientific facilities. To address this, we propose a novel graded early-warning framework that seamlessly integrates deep sequential forecasting with fuzzy comprehensive evaluation. A hybrid Transformer-LSTM neural network is developed to capture both long-range dependencies and local temporal patterns in multivariate task frequency sequences. To address the sensitivity of deep models to hyperparameters, we employ the bioinspired Black-winged Kite Algorithm (BKA) to jointly optimize the learning rate, hidden layer units, and regularization coefficients. BKA is selected for its Cauchy mutation and adaptive leadership mechanisms, which effectively avoid local optima and balance exploration-exploitation. Preliminary tests confirm that BKA outperforms classical methods such as PSO and GA. Subsequently, a fuzzy comprehensive evaluation model, incorporating entropy-based weights and membership functions derived from prediction residuals, maps the forecasting uncertainties onto a four-level early warning scale (Safe, Mild, Moderate, Severe). Digital simulation verification demonstrates that the proposed method achieves high prediction accuracy, with a root mean square error (RMSE) of 2.18 counts $$/\textrm{h}$$ and a coefficient of determination ( $$R^2$$ ) of 0.774 on the test set, outperforming baseline LSTM and PSO-LSTM models. Furthermore, the fuzzy warning system attains a precision of 0.91 and a recall of 0.93 for the Severe level, with an overall area under the ROC curve (AUC) of 0.94, confirming its reliable graded warning capability. The integration of data-driven prediction and fuzzy logic provides a powerful framework for active anomaly management in complex dynamic systems.