搜索结果：Scientific data

Land use and land cover (LULC) classification is essential for environmental monitoring, urban planning, and resource management. This study explores the performance of three state-of-the-art deep learning architectures, MobileNetV3, ResNet34, and GoogleNet, which were enhanced with transfer learning, data augmentation, and adaptive learning rate scheduling. We evaluate these models on two benchmark datasets: EuroSAT, consisting of Sentinel-2 satellite imagery across 10 land cover classes, and PatternNet, a high-resolution aerial dataset with 38 diverse classes. The results demonstrate that MobileNetV3 achieved the highest overall accuracy (97.83% on EuroSAT and 99.23% on PatternNet) with minimal inference time, making it ideal for real-time applications. ResNet34 achieved 97.56% and 99.06% accuracy, respectively, excelling in classifying complex, visually similar classes due to its residual learning blocks. GoogleNet's balanced performance and efficiency achieved 97.36% and 99.58% accuracy across both datasets. An ablation study confirmed that data augmentation, transfer learning, and learning rate scheduling contributed to improvements in accuracy of 5-13%. This research highlights the effectiveness of modern deep learning architectures and optimized training pipelines for LULC classification across diverse datasets, providing a foundation for future advancements in cross-domain remote sensing applications.

The increasing application of time-series analysis in fields like biomedical engineering or telecommunications emphasizes the need for high-quality data to train and evaluate advanced machine learning models. Acquiring temporal data at suitable resolutions is often limited by ethical, economic, or practical constraints. We introduce CoSiBD (Complex Signal Benchmark Dataset for Super-Resolution), a synthetic dataset designed for reproducible time-series super-resolution research. CoSiBD provides 2,500 high-resolution signals (N = 5, 000 samples each over a reference domain τ ∈ [0, 4π]) with aligned low-resolution versions at four levels (150, 250, 500, and 1,000 samples) obtained via uniform decimation. Signals are generated with diverse non-stationary behaviors through piecewise frequency modulation and spline-based amplitude envelopes, and provides both clean and noisy variants. Signals are distributed as NumPy arrays, plain text, and JSON, with comprehensive metadata describing segment structure, generation parameters, and seeds for full reproducibility. Technical validation analyzes spectral properties and reports baseline SR benchmarking and transfer experiments on EEG and speech data.

In this study, we present a comprehensive set of experimental data aimed at uncovering the mechanisms and regularities governing the deformation behavior of composites reinforced with continuous carbon fibers (CF) based on thermoplastic polymers. This work describes data extraction techniques that can later be used to optimize the mechanical properties of such structures using neural network models. This paper examines the thermoplastic polymer polysulfone (PSU) of the Ultrason S 2010 brand, which was used as the matrix material for the composites, while high-strength Toray T700SC fibers were used as the reinforcing fibers. Composite samples in the form of rods with a diameter of 1 mm were obtained by impregnating the fibers with a solution of polysulfone in N-methyl-2-pyrrolidone, followed by solvent removal. The collected dataset contains more than 600 tensile test results, including load-strain diagrams for different test conditions, data on the failure mechanisms of the specimens, and SEM images of the specimen microstructure in cross and longitudinal sections. This dataset will be useful for ML model development.

Protecting and improving surface water quality is contingent upon understanding the trends and spatial patterns in physical, biological, and chemical conditions and their underlying drivers. This requires observational data, spanning a diverse range of water quality constituents, coupled with contextual environmental data. Here we present the first global-scale integration of stream water quality into large-sample hydrology (named Caravan-Qual), combining ~96 million observations of 100 constituents with streamflow measurements, meteorological forcing and catchment attributes covering the period 1980-2025. We envisage that the dataset can facilitate a diverse range of empirical analyses (e.g. spatio-temporal analysis across diverse regions, quantification of pollutant loadings and exports, concentration-discharge analysis), in addition to supporting development and evaluation of process-based and data-driven models for water quality prediction and management.

We compiled and verified a comprehensive inventory dataset of communication tower infrastructure across the range of the greater sage-grouse (Centrocercus urophasianus) and Gunnison sage-grouse (Centrocercus minimus), two species of conservation concern that are viewed as ecosystem health indicators for the entire sagebrush biome within the United States. Our dataset includes all known towers with emphasis on validating construction year and month for towers built between 1990-2023. The annual spatial time series format of the data allows users to visualize, assess, and download tower locations and duration (including dates of construction and dismantlement) across the sagebrush biome of the western U.S. Tower data were acquired from publicly available infrastructure databases and records were filtered to include communication tower structures within the area of interest. Data records were validated and checked for accuracy with high resolution aerial and satellite imagery, and a subset were verified during field visits. The final filtered dataset comprises 4,322 tower sites, of which 3,528 tower site locations were verified via satellite imagery or field visits, and 794 were unverified tower sites (tower presence could not be confirmed via satellite imagery). Each tower record includes geographic coordinates, structure height, estimated date of construction, number of towers at each site, and, if applicable, date of dismantlement. The data product closes spatiotemporal gaps and resolves discrepancies present in other public versions of similar data and can be used in ecological research, infrastructure planning/siting/permitting, decision support tools for biological or landscape management, environmental assessments, or general use pertaining to the historic and current locations of communication infrastructure across sagebrush ecosystems.

With rising demand for remote work and education, smartphones and other portable photographic devices are increasingly used to capture physical documents, which are then shared as electronic files. However, shadows in such images hinder reading. Currently available shadow removal datasets exhibit certain limitations. This paper creates a semi-synthetic dataset (SSD-DIS) with 12,224 image sets. Using Blender for shadow masks, multi-source shadow-free images, and adjusted shadow intensity/color, it simulates real-world shadow scenarios. Experiments show SSD-DIS enhances neural networks' learning of document shadow features; models trained on it outperform those using traditional datasets, supporting document shadow removal algorithm research.

Optical Chemical Structure Recognition (OCSR) aims to convert two-dimensional molecular images into machine-readable formats such as SMILES strings. Deep learning has substantially improved OCSR performance, yet most methods rely on synthetic training data and struggle to generalize to real-world inputs, especially hand-drawn diagrams, where stroke width, geometry, and drawing conventions vary widely across individuals. In this work, we propose an image-to-graph model AdaptMol that enables effective transfer from synthetic to real-world data without requiring manual graph annotations in the target domains. AdaptMol is an integrated pipeline that starts with training a base model on synthetic data, and then refines model representations through unsupervised domain adaptation and self-training. Our key insight is that bond features are domain-invariant in nature; they encode structural relationships between atoms that are independent of visual variations across domains. Thus, during domain adaptation, we align bond-level feature distributions via class-conditional Maximum Mean Discrepancy (MMD) to enforce cross-domain consistency. We also design a comprehensive data augmentation strategy to enhance the robustness of the base model, facilitating stable self-training on unlabeled target samples. On hand-drawn molecular images, our model achieves 82.6% accuracy and outperforms the best prior method by 10.7 points, while maintaining competitive performance across four benchmarks comprising molecular images from scientific literature and patent documents.Scientific contributionWe propose AdaptMol, an image-to-graph model that predicts molecular structures as graphs of atoms and bonds, achieving effective transfer from synthetic to hand-drawn molecular images without requiring target domain graph annotations. We combine class-conditional Maximum Mean Discrepancy to align bond features across domains with comprehensive data augmentation to increase training data variation, jointly improving base model accuracy sufficiently for self-training and addressing the critical failure mode of prior approaches that begin with insufficient accuracy. We further introduce a dual position representation that supervises atom positions through both discrete coordinate tokens and continuous spatial heatmaps to reduce false positives in atom localization.

The integration of Artificial Intelligence (AI) into medicine has progressed from discriminative models to Generative AI (GenAI), which can synthesize novel content. For orthopaedic surgeons, scientific publication remains a vital marker of academic success but is often constrained by clinical workload. This review proposes a structured, practical framework to help orthopaedists effectively harness AI tools, transitioning from opaque, "black box" generation to grounded, verifiable research assistance through Retrieval-Augmented Generation (RAG). A PubMed search was conducted to explore the application of GenAI in the context of orthopaedic scientific research. An interactive review with experts in GenAI was also conducted, from which the proposed structure was developed. From this synthesis, a three-phase workflow is proposed: (1) Evidence selection using semantic discovery systems to identify and map relevant literature beyond keyword matching; (2) Data extraction and synthesis employing RAG-based systems to anchor AI responses to verified PDF sources, thereby minimizing hallucinations; and (3) Drafting and refining using Large Language Models (LLMs) for structured composition, linguistic clarity, and iterative manuscript improvement. The workflow integrates platform features to enhance efficiency, accuracy, and accessibility in orthopaedic research. When applied within a controlled, evidence-grounded environment, these systems can automate literature synthesis, expedite data extraction, and assist with scientific writing, while preserving authorial intent and accountability. However, challenges remain. Risks include algorithmic bias, "hallucinations", privacy concerns, and ethical issues related to authorship. Despite these limitations, AI represents a paradigm shift in orthopaedic scholarship, functioning as a cognitive exoskeleton that augments rather than replaces human expertise. With vigilant human oversight and adherence to journal ethics, orthopaedic surgeons can leverage AI to enhance research productivity, reproducibility, and quality while upholding the highest standards of scientific integrity.

The increasing frequency of freshwater cyanobacterial blooms has emerged as a critical ecological and environmental concern, yet long-term time series data documenting such blooms remain scarce. This study presents a 13-year dataset (2010-2022) from two adjacent subtropical reservoirs (Shidou and Bantou) in Xiamen, Fujian Province, Southeast China. It provides a monthly and quarterly overview of 20 physicochemical parameters (348 samples), microscope-based phytoplankton (348 samples), and DNA sequence-based data for bacteria (342 samples) and microeukaryotes (348 samples). The dataset highlights recurrent cyanobacterial blooms dominated by Raphidiopsis raciborskii (basionym Cylindrospermopsis raciborskii). This long-term dataset serves as a valuable resource for investigating, predicting, and controlling cyanobacterial blooms, and will support efforts in biodiversity forecasting, ecological restoration, and targeted management of freshwater ecosystems.

The COVID-19 pandemic highlighted the importance of human behavior in mitigating the spread of disease. Nonetheless, human behavior is often overlooked in models of disease spread, particularly by underutilizing real-world data. We address this by estimating probabilities that individuals engage in behaviors that influence SARS-CoV-2 transmission risk during the COVID-19 pandemic, between September 2020 and June 2022. These behaviors include wearing a mask, using public transportation, spending time with others, avoiding contact with others, and going to work. Our estimates account for the age and sex of individuals and are generated for every county in the United States. We utilized multiple open-source datasets and United States Census data to produce these estimates. Multiple datasets were used for validation, showing our estimates demonstrated comparable accuracy and robustness. Our estimates aid in understanding human behavior dynamics during the COVID-19 pandemic and could be used to inform monthly or longer-term behavior in simulations of COVID-19. Moreover, the methods presented can be applied to other behaviors and features for future simulations of infectious disease.

Pancreatic cancer is characterized by prolonged subclinical progression, molecular heterogeneity, and late clinical presentation, resulting in diagnosis predominantly at advanced stages. Current screening approaches lack sufficient sensitivity and scalability, underscoring the need for risk-adapted early detection strategies. Artificial intelligence (AI) offers a shift from reactive diagnosis toward proactive, precision-oriented screening. This review synthesizes recent advances in AI for the early screening and diagnosis of pancreatic cancer. We focus on how AI enables population-level and high-risk prediction, augments diagnostic assessment in patients with suspicious clinical, imaging, or molecular findings, and supports precision stratification through multimodal integration of radiologic imaging, circulating biomarkers, and longitudinal electronic health records (EHRs). Advances span three domains. In imaging, deep learning models-including convolutional neural networks, transformer architectures, and self-configuring segmentation frameworks-improve pancreas segmentation, lesion detection, and classification, with several systems demonstrating radiologist-level performance in retrospective multicenter studies. In biomarker discovery, machine learning approaches such as LASSO, random forest, and XGBoost facilitate high-dimensional feature selection from transcriptomic, metabolomic, and exosomal data, enabling composite diagnostic signatures beyond CA19-9. In longitudinal EHR analysis, temporal deep learning models identify latent disease trajectories and predict pancreatic cancer risk months to years before clinical diagnosis. Despite these advances, most models remain retrospectively validated and face limitations related to data heterogeneity, interpretability, and cross-population generalizability. AI strengthens early detection through multimodal integration, risk-adapted stratification, and data-driven clinical support aligned with precision medicine. Its near-term value lies in augmenting detection among high-risk populations rather than enabling universal screening or autonomous diagnosis. Prospective multicenter validation and improved model transparency are critical for translation into routine practice.

Fifty years ago, Werner Irnich presented the concept of an optimal pacemaker capable of responding appropriately to various cardiac arrhythmias and perceptual disturbances, and intended to be used in 85% of patients. With this concept, Irnich was far ahead of his time. His proposed circuitry for AV block and atrial fibrillation, as well as his suggestions for antitachycardia pacing and interference detection, were visionary. In the field of rate-adaptive pacing, he introduced AV-time control, the first closed-loop system. Werner Irnich represents the close connection between engineers and physicians in the field of cardiac electrotherapy. His theoretical work on the chronaxie rheobase and the electrode surface, confirmed by experimental data, still forms the basis of modern electrical stimulation today. The most extensive data on the interference immunity of electronic implants comes from his laboratory. In addition to his membership in numerous scientific societies, Werner Irnich served as Senior Editor of the international journal Pacing and Clinical Electrophysiology (PACE) from 1978 to 2013. He passed away on December 2, 2023, at the age of 89, leaving behind his wife Hanni, five children, and twelve grandchildren. We will always remember him with gratitude and deep appreciation for his contributions to cardiac electrostimulation.

Wilson disease (WD) is a rare autosomal recessive disorder of copper metabolism presenting with acute liver failure, cirrhosis, or neurologic involvement. Liver transplantation (LT) is the definitive treatment; however, data remain limited, particularly from regions reliant on living donor LT (LDLT). We retrospectively analyzed a prospectively collected transplant database, identifying all patients (≥ 14 years) who underwent LT for WD between January 2001 and December 2023. Data on demographics, LT indications, disease characteristics, pre-transplant therapy, complications, and outcomes were collected. Survival was assessed using Kaplan-Meier methods, and neurologic outcomes from clinical documentation. Forty-one patients underwent LT for WD (median age: 23 years; 51.2% female). Ascites was present in 68.4%, encephalopathy in 32.4%, and hepatocellular carcinoma in 5.1%. Acute liver failure was the initial presentation in 17.9%. LDLT comprised 53.7%. Acute cellular rejection occurred in 29.7% but was manageable; no patient required re-transplantation. Neurologic involvement was present in 17.1%, with 71% improving post-LT. One-, five-, and ten-year survival rates were 94%, 94%, and 82%. LT for WD yields excellent long-term survival. Neurologic improvement occurred in most Neuro-Wilson patients, supporting LT even in neurologically affected cases. LDLT plays a crucial role in regions with limited deceased donors.

Trivalent chromium is an essential trace element involved in carbohydrate and lipid metabolism. The widespread global prevalence of metabolic syndrome and its close association with cardiovascular diseases and type 2 diabetes mellitus have increased scientific interest in the potential metabolic effects of chromium. However, currently available evidence regarding its clinical significance remains inconsistent. This narrative review describes the role of trivalent chromium in the context of metabolic syndrome. A systematic literature search was conducted in the Scopus and Web of Science databases for studies published between 2015 and 2025. The review included randomized controlled trials, observational studies, experimental studies, systematic reviews, and meta-analyses that investigated chromium intake, supplementation, or the association between chromium levels and components of metabolic syndrome. The reviewed studies reported heterogeneous findings regarding the effects of trivalent chromium on components of metabolic syndrome. While some studies demonstrated improvements in glucose metabolism, insulin sensitivity, and lipid profiles, other studies reported no clear or statistically significant effects. The inconsistency of results has been attributed to differences in study design, studied populations, types and dosages of chromium supplementation, and duration of interventions. The lack of uniform research methodologies, limited sample sizes, and the absence of standardized protocols for chromium supplementation hinder the comparability of results. In addition, the heterogeneity of the studied populations limits the reliability of the available data. Available evidence does not support the widespread clinical use of trivalent chromium. Therefore, further large-scale studies are required to determine its efficacy and safety.

DNA methylation-based age prediction has become a reliable method for individual identification. While current models have achieved high accuracy when targeting a single type of biological fluid, crime scenes often contain multiple body fluids. Applying single fluid models to other fluid types may result in significant prediction errors, potentially misleading investigations. Therefore, age prediction models applicable to multiple biological fluids are of critical importance. In this study, we screened three age-related methylation sites (cg05940966, cg10501210, and cg10528482) from blood, saliva, and semen samples by analysis of public databases. These sites are associated with age. We then quantified methylation levels in peripheral blood samples from 101 healthy individuals via pyrosequencing. Based on these data, machine learning algorithms were applied to construct multiple age prediction models. These models were evaluated for their prediction accuracy, applicability to other body fluids, sensitivity, inhibitor tolerance, and utility with aged forensic samples. A multiple linear regression model constructed using the pyrosequencing results displayed the highest prediction accuracy, with mean absolute deviation (MAD) values of 2.717, 3.506, and 4.154 years in blood, saliva, and semen, respectively. Pyrosequencing also demonstrated high sensitivity, with MAD remaining within 4 years, even in trace samples (0.5 ng of unconverted genomic DNA). However, when the concentration of oxidized heme exceeds 1 ng/µL and the concentration of humic acid exceeds 2 ng/µL, pyrosequencing cannot accurately measure the methylation values at the CpG sites. In summary, this study provides an accurate and reliable tool for age prediction based on multiple bodily fluids. From a practical standpoint, this technology enables rapid age estimation without requiring prior identification of fluid type, thereby conserving valuable biological samples at crime scenes. This breakthrough will significantly enhance the efficiency of criminal investigations, strengthen the reliability of biological evidence interpretation in complex scenarios, and provide scientific support for judicial processes.

Deep learning models for medical image analysis often rely on large-scale parameterization, which may limit their practical use in resource-constrained settings. This study aims to design a structurally compact multi-source framework capable of delivering competitive diagnostic performance with reduced computational overhead. We propose ML-ConvNet, a lightweight architecture comprising approximately 4.2 K parameters and 924 M FLOPs at 512×512 input resolution. The network incorporates Multi-Branch Re-parameterized Convolutions for scale-aware feature extraction, Hierarchical Dual-Path Attention for feature localization, Feature Self- Transformation for cross-feature interaction, and a Local Variance Weighted optimization strategy to address class imbalance. The framework is evaluated independently on three publicly available benchmark datasets representing heterogeneous imaging modalities: brain MRI, lung CT, and chest X-ray. Ablation studies, precision-recall analysis, cross-modality validation, and computational benchmarking are conducted to assess performance, stability, and efficiency under controlled experimental conditions. Within the evaluated settings, results indicate competitive diagnostic accuracy relative to established lightweight baselines, including EfficientNet and MobileNet variants, while substantially reducing parameter count. Class-wise F1-scores and PR-AUC values suggest relatively stable minority-class performance under repeated cross-validation sampling. Attention visualizations show activations concentrated over regions broadly associated with pathological findings, though these observations are qualitative in nature. Inference latency measurements on CPU and mobile hardware suggest feasibility for low-latency deployment under the tested single-image batch configurations, though real-world throughput may differ depending on hardware and operational conditions. These findings suggest that careful architectural design and domain-informed inductive biases may support competitive medical image classification on public benchmark datasets without extensive parameter scaling. The framework was evaluated exclusively under controlled conditions on publicly available data, and multi-institutional external validation is required before conclusions regarding generalizability or clinical applicability can be drawn.

Forensic epigenetics is emerging as a powerful extension of traditional forensic genetics, offering the capacity to infer age, lifestyle, and environmental exposures from epigenetic marks. Yet its promise is shadowed by significant ethical, legal, and social questions. This article analyzes the scientific foundations and practical applications of forensic epigenetic techniques while interrogating their implications for privacy, discrimination, and human rights. It argues that the promise of enhanced investigative capability must be balanced against risks of misuse, stigmatization, and function creep. Drawing on comparative perspectives in law and bioethics, the authors emphasize the importance of proportional governance frameworks that uphold transparency, accountability, and respect for persons. Suggestions are made for the responsible integration of epigenetic data in forensic contexts, if and when, it meets sufficiently rigorous standards.

Peritoneal fibrosis, driven by M2 macrophage polarization, limits the long-term application of peritoneal dialysis (PD). Although ADAM19 is known to mediate fibrosis in other organs, its specific role in PD-associated peritoneal fibrosis remains unclear. PD patients were enrolled in a single center and divided into three groups depending on the PD time. Demographic and clinical data were collected. We detected the expressions of ADAM19, Notch1, Fibrosis-associated protein, chemokines and inflammatory factors in the peritoneum dialysis effluent by real-time PCR and western-blot assays. Macrophages were identified through flow cytometry. Then we analysis the relationship between ADAM19 and clinical data in PD patients. Furthermore, we established mouse models for peritoneal fibrosis to verify the biological function of ADAM19 in regulating macrophage polarization. In the long-term group, the fibrotic proteins (Fibronectin, &#x3b1;-SMA) and inflammatory factors (IL-6, IL-10) and chemokines (CCL5, CCL2, CXCL16) were higher than short-term group and more macrophages polarized towards M2. ADAM19 expression was linearly correlated with dialysis time and Kt/v. The AUROC of ADAM19 was 0.738 to identify the predictive value for peritoneal dialysis adequacy. The cut-off of ADAM19 RNA level was 7.84. In logistic regression models, higher ADAM19 (&#x2265;&#x2009;7.84) was also independently associated with lower Kt/v (<&#x2009;1.67). Additionally, the results revealed a moderate increment of M1 macrophage (CD86+) and enormous rise of M2 macrophage (CD206+) with high-glucose dialysis fluid in mice model. Furthermore, the 8-week G4.25% group showed significant growth of M2 macrophage compared to the 4-week G4.25% group, indicating that prolonged dialysis duration has a more pronounced effect on promoting M2 polarization of macrophages via ADAM19/Notch1 signaling pathway. Through stimulating chemokines and inflammatory factors, ADAM19 regulated macrophage polarization and was correlated to the progression of peritoneal fibrosis. ADAM19 is expected to be a novel indicator for detecting peritoneal ultrafiltration function in PD patients.

A renewed reforming of the higher education system is taking place in the conditions of the Covid-19 pandemic, as well as the perception of the essence and content of the pedagogical profession is changing in the changing conditions of today, as well as new requirements are being put forward to the personality of the teacher. The purpose of the article was to study the dynamics of perceptions and evaluation of the components of the image of an ideal teacher by students in the process of traditional (full-time) and remote (online) learning in a higher education institution. Diagnostic methods were used in the research: free description method on the topic "Portrait of a teacher whom I respect", content analysis, "Educational-cognitive interaction between a student and a teacher of the university" method (author I.I. Snyadanko), methods of statistical data processing. The authors conducted two experimental sections: the first section was conducted during traditional full-time learning before the Covid-19 pandemic, the second section was conducted during the remote learning period during the Covid-19 pandemic. As a result of the research, it was determined that, in general, first-year students prefer strict and demanding teachers, but at the same time value such teacher qualities as kindness and sacrifice. The personal characteristics of the teacher and the ability to perceive the student as a person are more important for second-year students. A comparison of the data of the two sections made it possible to conclude that the ideal teacher should meet much greater characteristics in the process of remote learning than in the process of traditional learning. The results of the article can be used to optimize the educational and cognitive interaction between students and teachers of the university, to improve the professional training of pedagogical personnel.