Browse Articles

Attention-Enhanced CNNs and transformers for accurate monkeypox and skin disease detection

Correction: Experimental investigation and prediction of the flexural properties of FDM printed carbon fiber reinforced polyamide parts using optimized RSM and ANN models

Association between exacerbation history and airway bacterial community assessed by extended bacterial culture and sequencing approaches in stable COPD

Enhancing adverse drug reaction data quality in Canada: A high-precision pipeline for medication name standardization and enrichment

Background: The Canada Vigilance Adverse Reaction database is a vital pharmacovigilance tool, but its utility is severely limited by heterogeneity in medication nomenclature. A substantial portion (∼36.8%) of unique drug name variants in the database lack any mapping to an active ingredient, representing a critical data quality gap that can mask important adverse drug reaction (ADR) signals. Methods: We developed, validated, and publicly released a high-precision, automated pipeline to standardize and enrich medication names. The pipeline employs a cascaded matching strategy that leverages the RxNorm and Observational Health Data Sciences and Informatics (OHDSI) vocabularies. Standardized names are assigned a RxNorm Concept Unique Identifier (RxCUI) and enriched with active ingredient data and Anatomical Therapeutic Chemical (ATC) classifications via RxNav APIs. The pipeline’s accuracy was rigorously assessed by two independent experts on a balanced validation set of 200 cases. Results: The final pipeline successfully standardized 94.5% of the 46,585 unique drug names. A blinded expert validation confirmed high reliability, demonstrating a precision of 98.02% (95% CI: 0.9307–0.9946) and specificity of 97.22% (95% CI: 0.9043–0.9923). Case studies showed that standardization and aggregation of reports revealed known safety signals (e.g., mesalamine and asthenia) that were statistically undetectable in the raw data. Conclusion: Our transparent and reproducible pipeline effectively resolves medication name heterogeneity in Canada’s national ADR database. By transforming variable text into standardized concepts, it significantly enhances data quality, improves the sensitivity of safety signal detection, and facilitates interoperability with global health datasets. The publicly available tool provides a valuable resource for strengthening drug safety surveillance in Canada and beyond.

Rich vehicle routing optimization based on variable neighborhood descent and differential evolution algorithm

Vision transformer and Mamba-attention fusion for high-precision PCB defect detection

Defects in printed circuit boards (PCBs) are being detected using computer vision-based techniques. Defect-free PCBs are essential for the reliability of consumer electronics. However, deep learning-based methods often struggle with imbalanced defect distributions and limited generalization. To address these challenges, we propose ViT-Mamba, a hybrid framework that combines Vision Transformers with a Mamba-inspired attention mechanism for global feature extraction and precise defect segmentation. We further introduce an artificial defect generation module that systematically creates six types of PCB defects to improve robustness. A multiscale hierarchical refinement strategy is employed to enhance feature representation for accurate segmentation. Experiments on a public PCB defect dataset show that ViT-Mamba outperforms existing methods, achieving a mean Average Precision (mAP) of 99.69%.

AMIGO2 accelerates tumor progression by inducing a cancer stem cell-like phenotype

Tobacco, electronic nicotine delivery system, nicotine replacement therapy, and cannabinoid use during pregnancy: A descriptive cross-sectional survey

Introduction Tobacco smoking is associated with adverse health outcomes for both pregnant women and their offspring. Smoking cessation counseling is an effective method to help women quit smoking. Developing a targeted smoking cessation intervention could benefit those who struggle to quit tobacco and potentially reduce the harm due to any co-occurring tobacco use. Assessing the prevalence of tobacco, electronic nicotine delivery systems (ENDS), nicotine replacement therapy (NRT), and cannabinoid use in pregnancy is key to developing such interventions. Thus, we aimed to assess the prevalence and patterns of tobacco, ENDS, NRT, and cannabinoid use in pregnancy. We further aimed to assess the prevalence of smoking cessation counseling intervention. Materials and methods We conducted a cross-sectional survey among pregnant women attending regular clinical visits at Spitalzentrum Biel between February and May 2023 (n = 262). Frequency and proportion along with 95% confidence intervals (CI) were reported for tobacco, ENDS, NRT, and cannabinoid use in pregnancy. Results Tobacco use was reported among 7.6% (20/262, 95% CI: 4.2%−11.1%) of the included pregnant women. Tobacco cigarettes (conventional or roll-on) were used by 7.3% (19/262, 95% CI: 3.8%−10.7%) of the surveyed pregnant women, with 0.8% (2/262, 95% CI: 0.0%−3.4%) of them reporting use of cigarettes along with ENDS and 0.4% (1/262, 95% CI: 0.0%−3.8%) reporting use of the cigarettes with NRT. Cannabinoid use was reported by 3.8% (10/262, 95% CI: 1.1%−7.0%) of pregnant women and all of them used products with Cannabidiol (CBD) only. Additionally, only 25% (5/20, 95% CI: 10.0%−48.3%) of tobacco users had received smoking cessation counseling intervention. Conclusion The estimated prevalence of tobacco, ENDS, NRT, and cannabinoid use among the pregnant women in this survey was 7.6%, 0.8%, 0.4%, and 3.8% respectively. However, among tobacco users, only one-fourth received smoking cessation counseling intervention.

MRI grading of lumbar disc herniation based on AFFM-YOLOv8 system

Categorical consistency of parity and magnitude facilitates implicit learning of color-number associations

Perceiving high-frequency stimulus pairings may lead to implicit associative learning. Interestingly, category-level pairings, such as blue-even, may facilitate implicit learning relative to item-level pairings, such as blue-2 and blue-7. Such an advantage of categorical consistency has been previously demonstrated for associative learning with parity; here, we replicate this finding, and extend it to a second, more-often studied category, magnitude. In a parity experiment, participants reported the parity of single-digit numerals; numerals appeared in either blue or yellow, but throughout, participants were not given any information about color. In the novel magnitude experiment, the same participants reported the magnitude of single-digit numerals appearing in either purple or green. Associative learning was assessed through the comparison of response performance to congruent (high-frequency color-number parings; p = .9) vs. incongruent (low-frequency; p = .1) trials. A robust congruency effect was found at the category-level for both parity (accuracy: 8%; response time (RT): 54 ms) and magnitude (accuracy: 4%; RT: 37 ms), but not at the item-level. A third, novel parity-mix experiment, with purplish-blue and greenish-yellow, was also tested with these participants, in order to probe for potential interactions of colors associated across parity and magnitude dimensions. There was a congruency-effect advantage for parity-magnitude matching numerals vs. mismatching in terms of accuracy (4%), suggesting that color associations with conceptual categories may relate to each other. An explicit association report task revealed above-chance accuracy for the color of numerals for both parity and magnitude at the category-level, and for parity at the item-level. These results suggest that categorical consistency of multiple numerical concepts may facilitate implicit learning of both specific and multidimensional color-number associations.

Characterization of the effects of oxymatrine on myocardial hypertrophy in spontaneously hypertensive rats through transcriptomics and metabolomics

Short-term power prediction of photovoltaic power stations based on Kepler optimization algorithm and VMD-CNN-LSTM model

This study focuses on the short-term power prediction of photovoltaic power stations, aiming to address the intermittent and fluctuating problems of photovoltaic power generation, in order to improve the prediction accuracy and ensure the stable operation of the power system. Innovatively introduce the Kepler algorithm into this field, deeply analyze historical data, and mine the nonlinear relationships among various factors to lay a solid data foundation for subsequent predictions. The VMD-CNN-LSTM combined model is constructed. It is a model combining variational mode decomposition (VMD), convolutional neural network (CNN) and long short term memory network (LSTM), VMD adaptively decomposes the original power sequence based on frequency characteristics to reduce data complexity. CNN accurately extracts spatial features from the decomposed modal components; LSTM leverages its expertise in processing time series data to capture the dynamic trends of power changes, and the three work in synergy. Meanwhile, the Kepler optimization algorithm (KOA) is deeply integrated with this model to optimize the entire process of the model from data preprocessing to result correction. Verified by examples, compared with the traditional prediction model, the proposed method has significant optimization in evaluation indicators such as root mean square error and mean absolute error, which strongly proves its effectiveness and superiority. It provides an innovative idea and reliable method for the short-term power prediction of photovoltaic power stations and is of great significance for promoting the grid connection of photovoltaic power generation and the optimization of the power system.

Correction: Knowledge of pregnant women towards pre-eclampsia in South Gondar zone, 2023

Association of cognitive function with frailty, nutritional status, and quality of life in older adults with mild cognitive impairment

Background In older adulthood, mild cognitive impairment is an intermediate stage between normal aging and dementia, making its detection crucial. Therefore, this study aims to analyze the associations between cognitive function and frailty, nutritional status, and quality of life in older adults with mild cognitive impairment. Methods This work was conducted through a cross-sectional, analytical study involving 129 adults diagnosed with mild cognitive impairment, with a mean age of 68.07 ± 4.22. For cognitive assessment, the Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Isaac Verbal Fluency Test, Trail Making Test (TMT), D2 Test of Attention (D2 Test), and Digit Symbol Substitution Test (DSST) were completed; clinical and functional status was assessed using the frailty (FRAIL), Mini Nutritional Assessment (MNA), and life quality 36-Item Short Form Survey (SF-36) questionnaires. Results Regarding overall cognitive performance, the presence of mild cognitive impairment was confirmed in the sample, as was the slowing of executive functions. Regarding selective attention, participants obtained an average of 138.30 points [SD = 4.30] in the D2 test, while the average score for processing speed measured using the DSST was 43.60 [SD = 3.50]. Regarding clinical and functional variables, the average FRAIL score was 2.26 [SD = 1.67], suggesting a high prevalence of frailty and pre-frailty; the average nutritional status was 27.91 [SD = 1.88], a range of adequate nutritional status. Finally, quality of life showed an average of 61.40 [SD = 14.87], indicating a moderate level. Discussion This study shows that frailty, nutritional status, and quality of life are closely related to mild cognitive impairment. These results reinforce the need for early and multidimensional interventions that contribute to preserving the quality of life.

Preconception computed tomography exposure and risk of stillbirth in a nationwide population based cohort study

Characteristics and driving mechanisms of vegetation phenology variations in the Bosten Lake Basin, Xinjiang, China

Vegetation phenology functions as a highly sensitive biological metric for delineating the growth state of vegetation and mirroring environmental alterations. The exploration of its spatiotemporal variations and driving forces constitutes a key area within contemporary global change research. Leveraging MODIS NDVI data spanning from 2001 to 2023, this research derived phenology parameters for the Bosten Lake Basin in Xinjiang, China. The Sen+Mann-Kendall approach was utilized to analyze vegetation phenology trends, and the partial least squares path model (PLS-PM) was applied to quantitatively disclose the direct and indirect influences of climate, terrain, soil, and human activities on phenology. The findings indicated that over the past 23 years, the start of growing season (SOS) in the Bosten Lake Basin was predominantly concentrated between the 120th and 150th days, and it progressively advanced from the western to the eastern part. The end of growing season (EOS) fluctuated between the 240th and 270th days and gradually postponed from west to east. The length of the growing season (LOS) is mainly between 105 and 140 days and gradually lengthens from west to east. In the past 23 years, SOS advanced at an average rate of 8.9 days per decade, EOS advanced by 1.7 days per decade, and LOS extended by an average of 7.2 days per decade. The phenological parameters of diverse vegetation types demonstrated marked disparities. Needleleaf forests and cultivated vegetation exhibited earlier SOS, later EOS, and longer LOS, whereas alpine vegetation had the shortest growing season, approximately 73 days. With the increase of elevation, SOS was significantly delayed, EOS was significantly advanced, and LOS was significantly shortened (p < 0.05 for all). The direct impacts of climatic factors on SOS, EOS, and LOS were significant (p < 0.01), with total impacts of 1.03, −0.94, and −1.00, respectively. Among these climatic factors, temperature and precipitation are the most representative variables reflecting the influence of climate on phenology. Altitude indirectly affects phenology through climate, soil conditions, and human activities. Understanding the phenological change characteristics and driving mechanisms in the Bosten Lake Basin provides a basis for clarifying the ecological environment evolution and climate response in arid regions of China.

Fabrication and dimensional scaling of SiNW-FET array via AFM-LAO lithography and self-limiting oxidation

Study on axial compressive properties of CFRP-confined ultra-high toughness cementitious composites

In order to study the mechanical properties of carbon fiber reinforced polymer (CFRP)-confined ultra-high toughness cementitious composites (UHTCC) under axial compression load, the preparation of UHTCC and the treatment of CFRP confinement were completed according to the specifications, considering the influence of single fiber types and CFRP confinement layers. After the axial compression load failure test, the failure phenomenon and stress-strain curves of the test were obtained. The influence mechanism of different single fiber types and different CFRP confinement layer on the axial compressive properties of CFRP-confined UHTCC was analyzed, and the stress-strain curves of the test was compared with the existing stress-strain models. The research results show that the ultimate compressive strength of UHTCC with steel fibers is significantly improved, reaching 84.6 MPa (compared to 27.6 MPa for unconfined specimens); The ultimate strain of UHTCC with PVA or PE increased significantly, with an increase of 107.6% and 243.2% respectively compared to unconfined specimens. The deviation between the calculated results of the Samaam model and the experimental results is less than 10%, indicating that the use of the Samaam model to propose a mathematical model for the stress-strain of CFRP-confined UHTCC is feasible.

Advanced gesture recognition in Indian sign language using a synergistic combination of YOLOv10 with Swin Transformer model

Abstract Communication between deaf or mute individuals and hearing persons is often hindered by the lack of mutual understanding of sign or vocal language. To bridge this gap, Indian Sign Language Recognition (ISLR) systems are essential. This paper proposes a real-time ISLR framework based on the YOLOv10-ST model, which integrates the Swin Transformer into the YOLOv10 architecture for enhanced feature extraction. The model also incorporates Mish activation to improve gradient flow and detection accuracy. A custom dataset comprising 15, 000 static images (1, 000 per sign for 15 signs) and 35 dynamic videos (covering 7 sign classes) was used for training and evaluation. Experimental results demonstrate high performance, with the model achieving 97.50% precision, 98.10% recall, and 96.58% F1-score for image-based sign recognition, and 95.24% precision, 96.00% recall, and 95.87% F1-score for video-based gestures. The model also achieves a mean Average Precision (mAP) of 97.62% and real-time inference speeds of 48.7 FPS. Ablation studies validate the contributions of Swin Transformer and Mish activation, while paired t-tests confirm statistical significance (p $$< 0.005$$ ). The experimental findings demonstrate that the YOLOv10-ST model efficiently recognizes static and dynamic ISL in real time with minimal computational overhead.

The impact of heterogeneous interpersonal relationships on promoting cooperation under the reputation mechanisms in public goods game

Although the principle of “case-by-case analysis” is widely endorsed, achieving complete rationality in the real world continues to be fraught with difficulties. Interpersonal relationships are heterogeneous, and the influence of social relationships and worldly wisdom on reputation evaluation should not be overlooked. Therefore, based on indirect reciprocity theory, this paper constructs a public goods game model with strategy update rules driven by reputation mechanism, aiming to investigate the impact of heterogeneous interpersonal relationships on the promotion of cooperative public goods provision among residents. The paper categorizes interpersonal relationships into three types, and proposes three corresponding reputation evaluation rules. Simulation results demonstrate that varying intensities of interpersonal relationships result in different levels of cooperation. When conducting public activities, externalities and the organizational efficiency of managers must be considered. Meanwhile, cooperation is difficult to sustain if the reputation mechanism fails to function effectively.