This study presents a robust and scalable method for automatic layout detection in digitized newspapers to facilitate efficient knowledge extraction and information retrieval. A custom dataset comprising annotated newspaper images in English, Hindi, and other languages was developed, with layout regions categorized into five primary classes. An enhanced YOLOv12 object detection model was trained on this dataset and evaluated using the mean Average Precision (mAP) metric across various Intersection over Union (IoU) thresholds. The model achieved a mAP@50 of 0.88, demonstrating strong detection performance and outperforming several stateof-the-art object detection models in the same task. The findings validate the effectiveness of the proposed approach in handling multilingual, structurally diverse newspaper formats. This research provides a practical framework for integrating automated layout analysis into digital archiving systems, OCR pipelines, and media monitoring applications. It also supports broader efforts to digitize historical print media and improve accessibility to regional content, thereby enabling enhanced research, journalism, and public engagement.

A cucurbituril-based supramolecular fluorescent probe for the visual detection of glyphosate.

Drugs play an indispensable role in the fields of medicine， agriculture， and animal husbandry. However， their long-term and improper use may lead to drug residues in food， the environment and organisms， posing a potentially serious threat to human health and the ecological environment. For instance， antibiotic residues may induce bacterial resistance， pesticide residues may cause neurotoxicity， and hormone drugs may interfere with the endocrine system. Therefore， developing sensitive and accurate detection methods for drug residues has become an important prerequisite and current hot topic in drug research. Meanwhile， the complicated matrices and low contents of the residues make it necessary for the widely used chromatography/mass spectrometry （MS） determination technologies to be coupled with efficient sample pretreatment procedures. Molecularly imprinted solid-phase microextraction （MI-SPME） technology combines the rapidity， high efficiency and solvent-free characteristics of SPME， and the specific recognition and selective adsorption capabilities of molecularly imprinted polymers （MIPs）， and shows significant advantages in the highly selective separation and enrichment of drug residues in complex samples. In recent years， the MI-SPME technology has become a research hotspot in the field of drug residue detection.This work systematically reviews the research progress since 2019 on the application of MI-SPME coupled with chromatography/MS in drug residue detection across food safety， environmental monitoring and biomedical fields. First， this work introduces in detail on the working principle and operation process of SPME technology. SPME achieves efficient enrichment of target analytes through the selective adsorption of the stationary phase-coated fibers， offering simplicity， speed， minimal solvent use， and compatibility with analytical instruments such as chromatography/MS.Next， the review focuses on elaborating the preparation methods and new technologies and strategies of MIPs. The traditional methods for preparing MIPs mainly include free radical polymerization， in-situ polymerization and sol-gel methods. However， traditional MIPs have defects such as template leakage risk， limited binding ability， and irregular material morphology， which restrict the application range. To this end， researchers have developed a series of novel preparation technologies and strategies， such as surface imprinting， nanoimprinting， dummy template imprinting， multi-template imprinting， multifunctional monomer imprinting and stimulus-response imprinting. These technologies and strategies have significantly enhanced the recognition and enrichment ability of MIPs for trace drug residues in complex samples by optimizing their structures and performances.To meet the requirements of different sample types and analytical instruments， MI-SPME media need to be designed into specific technical configurations through chemical or physical methods. This review summarizes six different MI-SPME device modes： MIPs-coated fiber SPME， MIPs in-tube SPME （IT-SPME）， MIPs stir bar sorptive extraction （SBSE）， MIPs dispersive SPME （DSPME）， MIPs thin-film SPME （TFME）， and MIPs in-tip SPME. Each mode offers unique advantages for the separation， enrichment and determination of drug residues in real samples. For example， the coated fiber SPME is simple to operate and suitable for direct immersion or headspace extraction of liquid samples； IT-SPME features miniaturization and automation， with excellent compatibility with chromatographic and mass spectrometric systems； DSPME achieves efficient separation and enrichment by dispersing adsorbents directly into sample solutions.Then， the applications of MI-SPME in the fields of food safety， environmental monitoring and biological medicine are summarized， highlighting typical research examples. In the field of food safety， MI-SPME can be used to detect pesticide residues， veterinary drug residues， and drugs for human use in fruits， vegetables， animal meats and dairy products. In environmental monitoring， it can be used for the detection of drug residues in aqueous environments and soil. In the field of biological medicine， it can be used for the analysis of drug residues in biological samples such as plasma， urine， and serum.Although the MI-SPME technology has shown great potential in drug residue detection， it still faces some challenges. For example， the preparation process of MIPs needs to be further optimized to improve their selectivity and stability； the development and application of new materials （such as graphene， metal-organic frameworks） for composite MIPs still need to solve problems such as high cost and complex processes； the integration of MI-SPME technology and automated equipment is also a bottleneck and important direction for future development. Looking ahead， with the advancement of green chemistry principles and point-of-care testing technologies， MI-SPME is expected to play an even greater role in drug residue detection. It will provide more efficient and precise technical support for food safety， environmental monitoring， and biomedical research.

Hepatocellular carcinoma (HCC), the predominant form of primary liver cancer, significantly threatens to global health. Despite considerable advances in diagnostic and therapeutic approaches in recent years, the prognosis for patients with HCC remains unsatisfactory. The emergence of artificial intelligence (AI), particularly deep learning technologies, offers new hope for improving the diagnosis and treatment of HCC. Researchers have extensively explored ways to integrate deep learning models into the clinical management of HCC patients, which provides a valuable foundation for developing more personalized treatment strategies. Compared with other detection methods, computed tomography (CT) has attracted significant research interest because of its comprehensive advantages, including wide availability and high resolution, making it well suited for AI-powered analysis. This review systematically integrates deep learning technologies for HCC based on CT imaging, while focusing primarily on tumor diagnosis, segmentation, treatment response prediction, and patient prognosis prediction. Moreover, we review popular deep learning networks in various fields and describe the advantages of these prevalent deep learning models for different applications. Furthermore, we discuss the outstanding challenges in applying deep learning to extract information from CT images for the diagnosis and treatment of HCC patients. These insights could provide guidance for subsequent studies.

Abstract Federated Learning (FL) offers a privacy-preserving framework suitable for sensitive domains such as healthcare. This study aims to investigate how different types and intensities of anomalous data introduced by individual clients affect overall model performance in cross-silo FL environments. Additionally, it explores whether models analyzing gradient representations can detect such anomalies during training. We conduct systematic experiments injecting six types of anomalies at varying strengths into training data from two distinct datasets. Performance degradation is measured and statistically analyzed. Furthermore, we develop a Variational Autoencoder (VAE) trained on clean gradient representations to detect deviations caused by anomalies. Our findings indicate that the impact of anomalies on model accuracy varies significantly across datasets and anomaly types. CIFAR-10 data shows higher sensitivity compared to the biological cellular data derived from Quantitative Phase Imaging (QPI). The VAE-based gradient anomaly detection successfully identifies subtle shifts in gradient distributions, but effective differentiation is observed primarily for the QPI data. The results emphasize the importance of tailoring FL robustness and anomaly detection strategies to specific datasets and anomaly characteristics. Gradient-based detection methods show promise for enhancing FL security, but require further refinement. This work contributes critical insights for designing more reliable and secure FL systems, particularly in sensitive domains like healthcare.

Non-dispersive infrared (NDIR) gas sensors have the disadvantage of high limit of detection (LOD), and the signal-to-noise ratio (SNR) has a significant impact on LOD. This paper proposes the application of matched filtering to NDIR gas sensors, a weak signal detection method that maximizes the SNR, improving the LOD. Experimental results show that matched filtering outperforms traditional methods, achieving an LOD of 1.131 ppm, while the LOD based on lock-in amplification is 7.350 ppm. The full-scale relative error and relative standard deviation (RSD) of the sensor in detecting different gas concentrations were optimized and improved. This approach provides a significant algorithmic improvement in the detection accuracy of NDIR gas sensors, offering potential for the detection of gases with less prominent absorption peaks, such as hydrogen.

ABSTRACT The rising incidence of amphetamine misuse, particularly in the context of attention deficit hyperactivity disorder treatment, underscores the urgent need for sensitive and effective detection methods. This review examines innovative nanomaterial‐based approaches for AMP detection, emphasizing their advantages over conventional analytical techniques. Various nanomaterials, including carbon nanotubes, graphene and metal nanoparticles, have been utilized to enhance the sensitivity and selectivity of detection methods such as electrochemical sensors, surface‐enhanced Raman spectroscopy and fluorescence‐based assays. The distinctive properties of nanomaterials, including high surface area, conductivity and biocompatibility, enable the development of rapid and reliable detection systems. This paper discusses recent advancements in nanomaterial synthesis, functionalization and integration into detection platforms, along with the challenges and future directions in this field. By harnessing the potential of nanotechnology, these innovative approaches aim to improve the accuracy and efficiency of amphetamine detection, thereby enhancing the monitoring and management of substance abuse, particularly in individuals with attention deficit hyperactivity disorder.

Potato early blight, caused by Alternaria solani, presents a significant threat to the potato industry. Existing detection methods for A. solani often fail to simultaneously achieve simplicity and accuracy. A gold-platinum (AuPt) nanozyme-assisted CRISPR/Cas12a system, termed the nanoparticle enzyme-assisted CRISPR detection (NACD assay) was developed. By integrating the precise target recognition of CRISPR with the enzyme-like activity of AuPt nanozymes, this system achieves simple, sensitive, and visual detection of A. solani. The NACD assay provided visual results through a distinct color change produced by the substrate catalyzed by the AuPt nanozyme. It can detect 100 copies/μL of the target dsDNA (A. solani 5.8S rRNA gene) and 10⁻3ng/μL A. solani genomic DNA. This detection method demonstrates high specificity, with no cross-reactivity observed with three other pathogens. Moreover, the incorporation of a filter paper-based readout enables straightforward visual detection by the naked eye, making it particularly suitable for on-site testing. Overall, these features make it an effective on-site diagnostic tool, allowing the potato industry to manage early diseases more efficiently.

Abstract With the rapid progress of IoT applications, wearable indoor localization techniques combining Ultra-Wideband (UWB) with Inertial Measurement Units (IMUs) have gained increasing attention. However, the Non-Line-of-Sight (NLoS) obstruction and multipath (MP) effects of UWB, along with the inherent cumulative error of IMUs, remain critical challenges that limit the localization accuracy and constrain their practical applications. This investigation proposes an improved UWB/IMU integrated system with adaptive tight coupling data fusion for pedestrian wearable localization. It enhances UWB ranging accuracy and Pedestrian Dead Reckoning (PDR) step length estimation precision by leveraging UWB NLoS and multipath identification and dynamic error compensation, alongside an adaptive PDR calibration mechanism to improve the accuracy of step length estimation. Firstly, a lightweight SVM-based NLoS/MP detection method is proposed, which employs the maximum relevance minimum redundancy (mRMR) algorithm to select the optimal feature subset, and uses Radial Basis Function Support Vector Machines (RBF-SVM) for classification. Secondly, an adaptive PDR calibration mechanism based on UWB displacement reference is established to achieve step length estimation, which enhances the system's dynamic adaptability to user gait variations in different scenarios. Then, an Adaptive Federated Extended Kalman Filter (AFEKF) framework is constructed, which achieves a tight combination fusion of the corrected UWB and calibrated PDR by dynamically weighting the UWB NLoS/MP state using posterior probability. Experimental results demonstrate that the proposed AFEKF tightly-coupled localization algorithm achieves a significantly lower Root Mean Square Error (RMSE) in complex indoor environments compared to conventional Extended Kalman Filter (EKF) and Federated EKF (FEKF) tightly-coupled algorithms. By effectively suppressing the impact of raw anomalous measurements on the filter and adapting to user gait variations in real-time, the algorithm significantly enhances the localization accuracy, robustness, and environmental adaptability.

Multi-modal fake news detection is a technique designed to identify and classify fake news by integrating information from multiple modalities. However, existing multi-modal fake news detection models have significant limitations in capturing structural information when processing social context. The core issue stems from the reliance on simple linear aggregation or static attention mechanisms in existing graph detection methods, which are inadequate for effectively capturing complex long-distance propagation relationships and multi-layered social network structures. Furthermore, existing multi-modal detection approaches are limited by the feature representations within the respective semantic spaces of each modality. The semantic gaps between modalities lead to misalignment during information fusion, making it difficult to fully achieve modality complementarity. To address these issues, we propose GINMCL, a graph isomorphism network-driven modality enhancement and cross-modal consistency learning method for multi-modal fake news detection. This method builds on the extraction of text and image features by incorporating graph isomorphism networks (GIN) based on the Weisfeiler-Lehman (WL) injective aggregation mechanism to effectively capture both local dependencies and global relationships within social graphs. Modality consistency learning aligns text, image, and social graph information into a shared latent semantic space, enhancing modality correlations. Additionally, to overcome the limitations of traditional methods in modality fusion strategies, we leverage a hard negative contrastive learning mechanism, which softens the penalty on negative samples and optimizes contrastive loss, further improving the accuracy and robustness of the model. We conducted systematic evaluations of GINMCL on the Pheme and Weibo datasets, and experimental results demonstrate that GINMCL outperforms existing methods across all metrics, achieving state-of-the-art performance.

Heavy metal (HM) contamination by cadmium (Cd), chromium (Cr), arsenic (As), zinc (Zn), lead (Pb), mercury (Hg), and other toxic elements in the environment poses substantial threat to public health and different ecosystems. Originating from diverse anthropogenic and natural sources, these elements can induce several ecological disturbances and multi-organ toxicity in humans and wildlife. Conventional biological and physicochemical methods for the removal of HMs, though effective in some contexts, often have limitations such as being energy intensive, costly, and generation of secondary waste. As a result, there is growing interest in exploring cleaner, efficient, and more sustainable approaches like bioremediation. Bioremediation is progressively acknowledged as one of the cost effective and sustainable strategy for pollution abatement by employing plants, bacteria, and other microorganisms capable of eliminating, transforming, or immobilizing HMs. This work aims to provide an overview of the conventional and advanced methods for the remediation of HMs, weighing up their benefits and limitations. Various methods for detection of HMs are also reviewed highlighting suitability, sensitivity, cost, portability, and field applicability. Further, we have discussed about the synergistic advantages of combining biological and physicochemical methods over standalone approaches, highlighting the need of hybrid methods like integration of artificial intelligence (AI) and nanotechnology in bioremediation. Overall, this review highlights bioremediation as a pivotal strategy for achieving cleaner ecosystems and sustainability, while underscoring the need for further research to optimize bioremediation technologies for broader real-world environmental management applications.

Caseous lymphadenitis (CLA), an infectious disease caused by Corynebacterium pseudotuberculosis (C. pseudotuberculosis), poses a significant economic burden to the global small ruminant industry. This study aimed to investigate genetic variations in the CP40 gene of C. pseudotuberculosis and to develop a rapid detection assay for enhanced pathogen identification. Homology analysis was performed to compare the CP40 gene sequence of the FJ-PN strain with other Corynebacterium species. Specific primers targeting CP40 were designed, and a SYBR Green I-based real-time PCR protocol was optimized. The assay’s specificity, sensitivity, and reproducibility were subsequently validated. The FJ-PN strain exhibited ≥99.65% nucleotide identity and ≥98.94% amino acid identity with C. pseudotuberculosis biovar ovi reference strains, showing 90.18–91.84% nucleotide identity and 88.63–90.77% amino acid identity with C. pseudotuberculosis biovar equi, and ≤82.71% nucleotide identity and ≤78.63% amino acid identity with other Corynebacterium species. The established qPCR assay demonstrated high specificity, the limit of detection was 52 copies/μL, and it demonstrated good reproducibility (intra- and inter-assay CV < 1.0%). Clinical sample testing revealed 18.8% positivity rates in nasal swabs, which was higher than that detected by conventional PCR (16.3%). These results indicate that the CP40 gene is evolutionarily conserved and represents a specific molecular marker for the identification of C. pseudotuberculosis biovar ovis. The developed SYBR Green I real-time PCR assay enables the efficient detection of C. pseudotuberculosis and provides technical support for CLA surveillance and control.

Marek's disease, a highly contagious avian immunosuppressive disorder caused by the α-herpesvirus MDV-1, poses a significant threat to poultry health. The development of rapid visual detection methods capable of distinguishing epidemic MDV-1 strains from vaccine strains is crucial for early disease warning, vaccine efficacy evaluation, and precise disease control. We developed a novel isothermal detection system that integrates recombinase polymerase amplification (RPA) with CRISPR/Cas14a technology for the visual identification of epidemic MDV-1 strains. This method operates at a constant temperature of 37°C and allows for either real-time analysis or endpoint visual readout without the need for complex instrumentation. Our results showed no cross-reactivity with Newcastle disease virus, infectious bursal disease virus, MDV-1 vaccine strains, or herpesvirus of turkeys. Plasmid DNA standards were used to determine the sensitivity of the assay, and the detection limit was 24.6 copies/μL. Clinical evaluation using 24 field samples confirmed that the method successfully identified all Marek's disease virus-positive cases, demonstrating its diagnostic reliability. In conclusion, we have developed a rapid, highly specific nucleic acid detection platform for MDV-1 that enables visual readout without complex instrumentation by combining the sensitivity of RPA with the specificity of CRISPR/Cas14a technology, offering promising potential for field-based diagnostics and disease surveillance.IMPORTANCEMarek's disease virus (MDV-1) is a highly contagious and economically important avian pathogen. Existing diagnostic methods are unable to reliably distinguish between epidemic and vaccine strains in field settings, which hampers effective surveillance and evaluation of vaccination programs. To address this challenge, we developed a portable isothermal detection assay that combines recombinase polymerase amplification with CRISPR/Cas14a technology. This approach enables highly sensitive (24.6 copies/μL) and specific visual detection of epidemic MDV-1 strains without cross-reactivity with vaccine strains or related viruses. The assay demonstrated 100% agreement with reference methods when evaluated using clinical samples. As a cost-effective method that avoids the need for complex detection instruments, it offers a practical solution for rapid on-site diagnosis, facilitating enhanced outbreak control and improved poultry health management globally.

Road infrastructure maintenance in developing countries faces unique challenges due to resource constraints and diverse environmental factors. This study addresses the critical need for efficient, accurate, and locally-relevant pavement distress detection methods in these regions. We present a novel deep learning approach combining YOLO (You Only Look Once) object detection models with a Convolutional Block Attention Module (CBAM) to simultaneously detect and classify multiple pavement distress types. The model demonstrates robust performance in detecting and classifying potholes, longitudinal cracks, alligator cracks, and raveling, with confidence scores ranging from 0.46 to 0.93. While some misclassifications occur in complex scenarios, these provide insights into unique challenges of pavement assessment in developing countries. Additionally, we developed a web-based application for real-time distress detection from images and videos. This research advances automated pavement distress detection and provides a tailored solution for developing countries, potentially improving road safety, optimizing maintenance strategies, and contributing to sustainable transportation infrastructure development.

Abstract In chemical production processes, the nonlinear dynamic characteristics of data frequently result in suboptimal performance of process monitoring methods. To address this issue, a novel fault detection method integrating Slow Feature Analysis with Kullback-Leibler Divergence (SFA-KLD) is proposed. First, the dataset is decomposed using slow feature analysis to extract the corresponding slow features and fast features. Second, by integrating sliding window technique, the mean and variance of the fast features in each sample window are computed sequentially. Finally, the Kullback-Leibler Divergence is employed to develop novel statistical measures for fault detection. The proposed method is validated through two numerical examples and simulation experiments conducted on the Tennessee Eastman (TE) process. Experimental results demonstrate that, compared to traditional methods, SFA-KLD achieves better fault detection performance in nonlinear dynamic process monitoring.

ABSTRACT As the method of directly separate wax‐solid deposition from crude oil is difficult to achieve and its impact on the blockage mechanism of the permeation fluid channels cannot be independently explored, this study attempts to use indirect detection methods. When designing pressure parameters, factors such as rock mechanical properties should be considered. This article presents a detailed study of optimal immiscibility pressure and displacement processes. Through two sets of experiments, the minimum miscibility pressure of the target oil sample was determined to be 75 + MPa. For high‐pressure environments with pressures exceeding 30 MPa, the change in the productivity/injectivity index before gas breakthrough is relatively small, after and during gas breakthrough, fluctuations in the index increase significantly. Curve characteristics indicate that for this type of oil, when the pressure approaches a certain high‐pressure state, even before reaching the MMP, the permeate fluid channels may be blocked, which is not conducive to crude oil mining. This interpretation also supports and confirms the results of microscopic visualization simulation experiment. After comprehensively considering the rock fracture pressure value, end‐point reserve recovery trend line chart, and gas injection cost, 35 MPa was selected as the optimal immiscibility pressure.

In this study, we focus on surface defects in steel and conduct an analysis of various detection methods for steel surface defects. The detection of steel surface defects is a crucial analysis that ensures the quality of steel production. To address challenges such as low detection accuracy and inadequate feature extraction capability in steel surface defect detection, an enhanced YOLOv8-based steel defect detection algorithm, GS-YOLO, is proposed and implemented for the stated analysis. The network neck is augmented with an information aggregation-distribution mechanism module (GD) to strengthen cross-scale information recognition for steel surface defects. Furthermore, a scale sequence feature fusion module (ScalSeq) is employed to capture both high-dimensional and low-dimensional detail information from feature maps, enabling a more comprehensive integration of multi-scale features and enhancing the model’s performance in addressing multi-scale challenges. In the context of NEU-DET, several experiments have been conducted within the YOLO environment. The results of these experiments indicate that the modified/improved GS-YOLO model has reached an accuracy of 76.6%. This is a 2.5% improvement over the original method, and in general, it outperforms other standard object identification models.

In this work, a novel Ti-based MOF (NH2-MIL-125) bonded open-tubular (OT) column was firstly prepared via a one-step bonded growth method for capillary electrochromatography (CEC). The stationary phase was characterized by Fourier transform-infrared spectroscopy, scanning electron microscopy, X-ray diffraction spectroscopy, nitrogen adsorption-desorption isotherm measurements and zeta potential measurements. The results showed that the stationary phase exhibited a large specific surface area (1247.57 m2 g-1) with microporous and mesoporous structure and no obvious changes in the morphology/size inside and outside the column. Baseline separation of three β-adrenergic receptor blockers (Prop, Sot, and Lab), three β-adrenergic receptor agonists, and four sulfonamide antibiotics was obtained under the optimized CEC conditions with the shortest analysis time of 2.52 min and a maximum resolution of 11.52. The separation mechanisms were mainly attributed to the polarity and electrophoretic mobility of the analytes, as well as π-π interactions and hydrogen bonding interaction between the stationary phase and the analytes. A quantitative detection method for the three β-adrenergic receptor blockers in pork samples was established using the NH2-MIL-125 bonded OT column. Good linearity (R2 > 0.999) was obtained over the concentration range of 0.01-1.00 mg mL-1 with limits of detection of 0.0044-0.0084 mg mL-1 and recoveries of 90.24-106.74%. Thus, the developed method was simple, rapid and highly efficient, and could be applied for the simultaneous separation and detection of the three β-adrenergic receptor blockers in real samples.

Object detection, a cornerstone of computer vision, aims to localize and classify objects within images. This comprehensive survey reviews modern object detection methods, focusing on two dominant paradigms: Convolutional Neural Networks (CNNs) and Transformer-based architectures. This work provides a structured comparison of CNN-based and Transformer-based detection paradigms, highlighting their complementary strengths and trade-offs. CNNs demonstrate advantages in local feature extraction and computational efficiency, whereas Transformers excel at capturing global context through self-attention mechanisms. We also analyze multi-modal fusion techniques integrating Red-Green-Blue (RGB), Light Detection and Ranging (LiDAR), and language embeddings. Benchmark results from representative models include: Real-Time Detection Transformer (RT-DETR) achieves 53.1% mean Average Precision (mAP) at Intersection over Union (IoU) at 0.5:0.95, You Only Look Once version 8 (YOLOv8) achieves 50.2% mAP at 0.5:0.95, real-time detectors exceed 100 frames per second (FPS) with competitive accuracy, and specialized infrared methods achieve 92.45% F-measure on NUAA-SIRST dataset. The work introduces a novel taxonomy of multi-modal fusion strategies, documents field-wide and review-specific limitations, and synthesizes recent 2024 to 2025 benchmarks across diverse datasets. Despite these advances, significant challenges remain in handling scale variation, occlusion effects, and domain adaptation. This survey outlines these persistent obstacles and promising research directions, providing a structured reference for researchers and practitioners.

Development of a Rapid Detection Kit for Chicken Infectious Anemia Virus Nucleic Acid (RPA-LFD method).