Search Results (2,499)

In low-altitude multi-object tracking (MOT), challenges such as frequent inter-object occlusion and complex non-linear motion disrupt the appearance of individual targets and the continuity of their trajectories, leading to frequent tracking failures. We posit that the relatively stable spatio-temporal relationships within object groups (e.g., pedestrians and vehicles) offer powerful contextual cues to resolve such ambiguities. We present NOWA-MOT (Neighbors Know Who We Are), a novel tracking-by-detection framework designed to systematically exploit this principle through a multi-stage association process. We make three primary contributions. First, we introduce a Low-Confidence Occlusion Recovery (LOR) module that dynamically adjusts detection scores by integrating IoU, a novel Recovery IoU (RIoU) metric, and location similarity to surrounding objects, enabling occluded targets to participate in high-priority matching. Second, for initial data association, we propose a Graph Cross-Attention (GCA) mechanism. In this module, separate graphs are constructed for detections and trajectories, and a cross-attention architecture is employed to propagate rich contextual information between them, yielding highly discriminative feature representations for robust matching. Third, to resolve the remaining ambiguities, we design a cascaded Matched Neighbor Guidance (MNG) module, which uniquely leverages the reliably matched pairs from the first stage as contextual anchors. Through MNG, star-shaped topological features are built for unmatched objects relative to their stable neighbors, enabling accurate association even when intrinsic features are weak. Our comprehensive experimental evaluation on the VisDrone2019 and UAVDT datasets confirms the superiority of our approach, achieving state-of-the-art HOTA scores of 51.34% and 62.69%, respectively, and drastically reducing identity switches compared to previous methods. Full article

Background/objectives: The continuous necessity to support biostrumental data with biolomecular data collected using non-invasive tools is influencing the world of ocular surface devices. The ocular imprint still represents a non-invasive and safety technique for collecting corneal and conjunctival epithelia in an easy way, as performed in human and veterinary clinics. Although used in clinical practice since 1977, operators might benefit from improvements in these techniques, especially in terms of handling and management. Methods: Herein, by reporting the design and characteristics of a patent of ocular surface sampling (the SurfAL pen and periocular-assisted SurfAL pen; PCT WO2016IB51474 20160316), we performed a validation and analysis of its value compared to gold standards. The level-headedness and advantages of this device were verified in 15 sclerocorneal specimens (sampling advantages) and tested in 25 volunteers (handling and operator efficiency, as well as frequency of discomfort in volunteers). Morphological as well as biomolecular analyses were used to compare SurfAL devices with conventional ones. Results: The easy management of SurfAL pens and the good detection of epithelial/goblet cells were confirmed. The SurfAL pen was found to be smart and suitable for routine analysis, as confirmed by quick and reproducible onsite sampling. Periocular-assisted SurfAL pen was comparable in terms of sampling quality but less comparable in terms of subject confidence due to its geometry. Conclusions: This study suggests that the SurfAL pen and periocular-assisted SurfAL pen might represent an additional and hands-on way of sampling ocular surface cells and improve the diagnostic route in ophthalmology. Full article

Background: Diagnosing facial melanoma, specifically lentigo maligna (LM) and lentigo maligna melanoma (LMM), is a daily clinical challenge, particularly for small or traumatized lesions. LM and LMM are part of the broader group of atypical pigmented facial lesions (aPFLs), which also includes benign look-alikes such as solar lentigo (SL), atypical nevi (AN), seborrheic keratosis (SK), and seborrheic-lichenoid keratosis (SLK), as well as pigmented actinic keratosis (PAK), a potentially premalignant keratinocytic lesion. Standard dermoscopy with handheld devices is the most widely used diagnostic tool in dermatology, but its accuracy heavily depends on the clinician’s experience and the perceived difficulty of the case. As a result, many benign aPFLs are excised for histological analysis, often leading to aesthetic concerns. Reflectance confocal microscopy (RCM) can reduce the need for biopsies, but it is limited to specialized centers and requires skilled operators. Aims: This study aimed to assess the impact of personal skill, diagnostic confidence, and perceived difficulty on the diagnostic accuracy and management in the differential dermoscopic diagnosis of aPFLs. Methods: A total of 1197 aPFLs dermoscopic images were examined on a teledermoscopic web platform by 155 dermatologists and residents with 4 skill levels (<1, 1–4, 5–8, >8 years). They were asked to give a diagnosis, to estimate their confidence and rate the case, and choose a management strategy: “follow-up”, “RCM” or “biopsy”. Diagnostic accuracy was examined according to the personal skill level, confidence level, and rating in three settings: (I) all seven diagnoses, (II) LM vs. PAK vs. fully benign aPFLs, (III) malignant vs benign aPFLs. The same analyses were performed for management decisions. Results: The diagnostic confidence has a certain impact on the diagnostic accuracy, both in terms of multi-class diagnosis of six aPFLs in diagnostic (setting 1) and in benign vs malignant (setting 3) or benign vs. malignant/premalignant discrimination (setting 2). The perceived difficulty influences the management of benign lesions, with easy ratings predominantly matching with “follow-up” decision in benign cases, but not that of malignant lesions assigned to “biopsy”. The experience level had an impact on the perception of the number of real easy cases and had no to minimal impact on the average diagnostic accuracy of aPFLs. It, however, has an impact on the management strategy and specifically in terms of error reduction, namely the lowest rates of missed malignant cases after 8 years of experience and the lowest rates of inappropriate biopsies of benign lesions after 1 year of experience. Conclusions: The noninvasive diagnosis and management of aPFLs rest on a daily challenge. Highlighting which specific subgroups of lesions need attention and second-level examination (RCM) or biopsy can help detect early malignant cases, and, in parallel, reduce the rate of unnecessary removal of benign lesions. Full article

Conventional spectrophotometric methods used for determining total phenolic content and antioxidant activity are typically time-consuming, labor-intensive, and require large amounts of reagents. In the context of sustainable development and green chemistry, minimizing the use of hazardous substances and reducing reagent consumption have become key priorities. The implementation of microplate-based methods offers significant advantages, including reduced reagent volumes and shorter analysis times compared with traditional methods. Therefore, the aim of this study was to validate the Folin–Ciocalteu and DPPH microplate methods and compare their performance with conventional protocols. The limits of detection (LOD) for the microplate methods were lower than those for the conventional approaches, being approximately 0.7 µg/mL and 4.1 µg/mL for TPC, and 0.015 µg/mL and 0.081 µg/mL for DPPH, respectively. The relative standard deviation (RSD) of repeatability and reproducibility for both microplate methods was ≤6%. The accuracy ranged from 95.0% to 97.7% for TPC and from 95.3% to 98.7% for DPPH. Overall, the results confirm that the microplate and conventional methods are statistically equivalent at the 95% confidence level, demonstrating that microplate assays represent a reliable and environmentally friendly alternative for assessing total phenolic content and antioxidant activity. Full article

Background: Broccoli contains diverse phytochemicals, including glucosinolates and their hydrolysis products, with potential nutritional and bioactive properties. Accurate metabolite profiling requires optimized sample preparation and comprehensive databases. Methods: A rapid enzymatic deactivation method with 70% methanol, implemented prior to cryogrinding, was evaluated for processing freeze-dried and fresh broccoli florets, which were compared as plant materials. A widely targeted, organ-resolved metabolite database was constructed by integrating over 612 reported phytochemicals with glucosinolate degradation products. LC-HRMS combined with MS-DIAL and GNPS was employed for metabolite detection and annotation. Results: Freeze-dried samples yielded nearly twice the number of glucosinolates, isothiocyanates, and nitriles compared with standard-processed fresh tissue. Methanol pre-treatment preserved metabolite integrity in fresh samples, achieving comparable sensitivity to freeze-dried material. Using the integrated database, 998 metabolites were identified or tentatively characterized, including amino acids, carboxylic acids, phenolics, alkaloids, terpenoids, and glucosinolate derivatives. Cross-platform reproducibility was improved and false positives reduced. Conclusions: Optimized sample preparation combined with a curated metabolite database enables high-confidence, comprehensive profiling of broccoli florets phytochemicals. The resulting dataset provides a valuable reference for studies on genotype–environment interactions, nutritional quality, and functional bioactivity of cruciferous vegetables. Full article

Tea is one of the most consumed beverages in the world. However, tea plants are often susceptible to various diseases, especially leaf diseases. Currently, most tea farms identify leaf diseases through manual inspection. Due to its time-consuming and resource-intensive nature, manual inspection is impractical for large-scale applications. This study proposes a novel convolutional neural network model designated as YOLOv5 Bi-HIC for detecting tea leaf diseases, including algal leaf spot, tea brown blight, and tea grey blight. The model enhances the conventional YOLOv5 object detection model by incorporating instance and context information to improve the detection performance. A total of 1091 raw images of tea leaves affected by algal leaf spots, tea brown blight, and tea grey blight were captured at Wenhua Tea Farm, Miaoli City, Taiwan. The results indicate that the proposed model achieves precision, recall, F1 Score, and mAP values of 0.977, 0.943, 0.968, and 0.96, respectively, during training. Furthermore, it exhibits a detection confidence score of 0.94, 0.98, and 0.92 for algal leaf spot, tea brown blight, and tea grey blight, respectively. Overall, the results indicate that YOLOv5 Bi-HIC provides an accurate approach for real-time detection of leaf diseases and can serve as a valuable tool for timely intervention and management in tea plantations. Full article

The integration of autonomous vehicles (AVs) into road transport requires robust experimental tools to analyze the human–machine interaction, particularly under conditions of system disengagement. This study presents the primary controls calibration and virtual scenario validation of the EVACH autonomous driving simulator, designed to reproduce the SAE Level 2 and Level 3 driving modes in rural road scenarios. The simulator was customized through hardware and software developments including a dedicated data acquisition system to ensure the accurate detection of braking, steering, and other critical control inputs. Calibration tests demonstrated high fidelity, with minor errors in brake and steering control measurements, consistent with values observed in production vehicles. To validate the virtual driving rural environment, comparative experiments were conducted between naturalistic road tests and simulator-based autonomous driving, where five volunteers participated in the preliminary pilot test. Results showed that average speeds in the simulation closely matched those recorded on real roads, with differences of less than 1 km/h with minimum standard deviation and confidence values. These findings confirm that the EVACH simulator provides a stable and faithful reproduction of autonomous driving conditions. The experimental platform offers valuable support for current and future research on the safe deployment of automated vehicles. Full article

Background: Stomatitis is a common adverse event associated with targeted therapies for hormone receptor-positive, HER2-negative (HR+/HER2–) breast cancer, particularly those inhibiting the PI3K/AKT/mTOR pathway. While mTOR-inhibitor-associated stomatitis is well established, less is known about its occurrence with other kinase inhibitors in real-world settings. We performed a pharmacovigilance disproportionality analysis of the FDA Adverse Event Reporting System (FAERS) to evaluate stomatitis reports for alpelisib, capivasertib, everolimus, and palbociclib. Methods: Events were identified using four term sets—Stomatitis, Original Trial Terms (OTT), Comprehensive Trial Terms (CTT), and Stomatitis-Associated Main Terms (SAMT)—which reflect varying definitions and medical terminologies. Disproportionality analyses using reporting odds ratio (ROR), proportional reporting ratio (PRR), and Information Component (IC) were calculated with 95% confidence intervals. Results: All agents showed ROR and PRR >1, indicating higher odds and reporting proportions of stomatitis compared with other drugs. These findings were confirmed by IC analysis. Everolimus demonstrated the strongest association (ROR: 30.72 [29.61–31.88]), followed by alpelisib (ROR: 13.11 [11.79–14.58]) and palbociclib (ROR: 11.73 [11.35–12.11]). Capivasertib had the lowest reporting odds (ROR: 3.14 [1.81–5.43]), though limited by fewer reports. Differences between CTT and SAMT were minimal (~2%). Conclusions: These results support the use of the SAMT as an efficient screening tool. Furthermore, these findings underscore the need for optimized stomatitis detection and continued monitoring, particularly for PI3K and mTOR inhibitors, in both clinical trials and postmarketing surveillance. Full article

Background: Sotorasib, a KRAS G12C inhibitor, is approved for treating non-small cell lung cancer (NSCLC) and has shown a distinct safety profile in randomized clinical trials (RCTs). However, post-marketing pharmacovigilance is crucial to identify real-world safety signals including sex-specific differences that may not be evident in controlled trial settings. Methods: This analysis reviewed 845 individual case safety reports (ICSRs) from the EudraVigilance (EV) database between 1 January 2021, and 8 April 2025, involving NSCLC patients treated with sotorasib. Adverse drug reactions (ADRs) were assessed by sex, seriousness, outcome, and system organ class (SOC). Disproportionality analyses were conducted to detect sex-specific safety signals, and results were compared with data from the CodeBreaK200 RCT by using a two-proportion z-test. Results: Among the ICSRs, 49.2% involved male and 40.1% female patients. Serious ADRs accounted for 47.5% of cases, with females at higher risk (relative risk [RR] = 1.31; 95% confidence interval (CI): 1.22–1.40; p < 0.0001). The most frequently reported SOCs were neoplasms (15.8%), gastrointestinal disorders (15.3%), and hepatobiliary disorders (11.5%). Four sex-specific safety signals were identified: women had a significantly increased risk of cholestasis (RR = 3.37) and hepatotoxicity (RR = 3.01), while men were less likely to report decreased appetite (RR = 0.20) and rash (RR = 0.14). Real-world data showed lower reporting of diarrhea, fatigue, nausea, and liver enzyme elevations (p < 0.0001). Conclusions: Real-world pharmacovigilance supports the RCT findings and highlights sex-specific risks, thus emphasizing the importance of sex-aware monitoring and personalized toxicity management. Full article

On construction sites, it is crucial and and in most cases mandatory to wear safety equipment such as helmets, safety shoes, vests, and belts. The most important of these is the helmet, as it protects against head injuries and can also serve as a marker for detecting and tracking workers, since a helmet is typically visible to cameras on construction sites. Checking helmet usage, however, is a labor-intensive and time-consuming process. A lot of work has been conducted on detecting and tracking people. Some studies have involved hardware-based systems that require batteries and are often perceived as intrusive by workers, while others have focused on vision-based methods. The aim of this work is not only to detect workers and helmets, but also to identify workers through labeled helmets using symbol detection methods. Person and helmet detection tasks were handled by training existing datasets and gained accurate results. For symbol detection, 14 different shapes were selected and put on helmets in a triple format side by side. A total of 11,243 images have been annotated. YOLOv5 and YOLOv8 were used to train the dataset and obtain models. The results show that both methods achieved high precision and recall. However, YOLOv5 slightly outperformed YOLOv8 in real-time identification tests, correctly detecting the helmet symbols. A testing dataset containing different distances was generated in order to measure accuracy by distance. According to the results, accurate identification was achieved at distances of up to 10 meters. Also, a location-based symbol-ordering algorithm is proposed. Since symbol detection does not follow any order and works with confidence values in the inference mode, a left to right approach is followed. Full article

Background/Objectives: The COVID-19 pandemic highlighted how infodemics—an excessive amount of both accurate and misleading information—undermine health responses. Artificial intelligence (AI) and digital tools have been increasingly applied to monitor, detect, and counter health misinformation online. This scoping review aims to systematically map digital and AI-based interventions, describing their applications, outcomes, ethical and equity implications, and policy frameworks. Methods: This review followed the Joanna Briggs Institute methodology and was reported according to PRISMA-ScR. The protocol was preregistered on the Open Science Framework . Searches were conducted in PubMed/MEDLINE, Scopus, Web of Science, and CINAHL (January 2017–March 2025). Two reviewers independently screened titles/abstracts and full texts; disagreements were resolved by a third reviewer. Data extraction included study characteristics, populations, technologies, outcomes, thematic areas, and domains. Quantitative synthesis used descriptive statistics with 95% confidence intervals. Results: A total of 63 studies were included, most published between 2020 and 2024. The majority originated from the Americas (41.3%), followed by Europe (15.9%), the Western Pacific (9.5%), and other regions; 22.2% had a global scope. The most frequent thematic areas were monitoring/surveillance (54.0%) and health communication (42.9%), followed by education/training, AI/ML model development, and digital engagement tools. The domains most often addressed were applications (63.5%), responsiveness, policies/strategies, ethical concerns, and equity/accessibility. Conclusions: AI and digital tools provide significant contributions in detecting misinformation, strengthening surveillance, and promoting health literacy. However, evidence remains heterogeneous, with geographic imbalances, reliance on proxy outcomes, and limited focus on vulnerable groups. Scaling these interventions requires transparent governance, multilingual datasets, ethical safeguards, and integration into public health infrastructures. Full article

Background/Objectives: Ankle plantar flexor spasticity after stroke may limit mobility, especially during turning and multi-directional stepping. Evidence on performance differences and measurement properties across spasticity levels is limited. We examined whether performance on the Activities-specific Balance Confidence Scale (ABC Scale), Five Times Sit-to-Stand Test (5xSTS), Figure-of-8 Walk Test (F8WT), and Four-Square Step Test (FSST) differs by spasticity severity, and evaluated test–retest reliability, the intraclass correlation coefficient (ICC), the standard error of measurement (SEM), and the minimal detectable change (MDC). Methods: In an observational cross-sectional comparative study, 54 individuals more than 6 months post-stroke were classified into three groups by the Modified Ashworth Scale (MAS = 0, MAS = 1 − 1+, MAS ≥ 2). Participants completed the ABC Scale, 5xSTS, F8WT, and FSST. One-way analysis of variance with Bonferroni adjustment tested group differences. Reliability was quantified using ICC (2,1); SEM and MDC at the 95% confidence level indexed absolute reliability. Results: No significant differences were found for the ABC Scale or 5xSTS. F8WT and FSST differed by spasticity level (p < 0.05), with poorer performance in the highest-spasticity group versus no spasticity. ICCs were high across assessments. All SEMs were <20% of test–retest means, and all MDCs were <20% of maximum scores. Conclusion: Assessments that require directional change detected differences across spasticity levels, whereas balance confidence and repeated sit-to-stand did not. All measures showed acceptable relative and absolute reliability. Findings support selecting outcomes by spasticity severity and using SEM and MDC as reference values when interpreting change in stroke rehabilitation. Full article

Forest fires pose significant threats to ecosystems, human life, and the global climate, necessitating rapid and reliable detection systems. Traditional fire detection approaches, including sensor networks, satellite monitoring, and centralized image analysis, often suffer from delayed response, high false positives, and limited deployment in remote areas. Recent deep learning-based methods offer high classification accuracy but are typically computationally intensive and unsuitable for low-power, real-time edge devices. This study presents an autonomous, edge-based forest fire and smoke detection system using a lightweight MobileNetV2 convolutional neural network. The model is trained on a balanced dataset of fire, smoke, and non-fire images and optimized for deployment on resource-constrained edge devices. The system performs near real-time inference, achieving a test accuracy of 97.98% with an average end-to-end prediction latency of 0.77 s per frame (approximately 1.3 FPS) on the Raspberry Pi 5 edge device. Predictions include the class label, confidence score, and timestamp, all generated locally without reliance on cloud connectivity, thereby enhancing security and robustness against potential cyber threats. Experimental results demonstrate that the proposed solution maintains high predictive performance comparable to state-of-the-art methods while providing efficient, offline operation suitable for real-world environmental monitoring and early wildfire mitigation. This approach enables cost-effective, scalable deployment in remote forest regions, combining accuracy, speed, and autonomous edge processing for timely fire and smoke detection. Full article

Recent advances in hyperspectral imaging (HSI) and multimodal deep learning have opened new opportunities for crop health analysis; however, most existing models remain limited by dataset scope, lack of interpretability, and weak cross-domain generalization. To overcome these limitations, this study introduces Agri-DSSA, a novel Dual Self-Supervised Attention (DSSA) framework that simultaneously models spectral and spatial dependencies through two complementary self-attention branches. The proposed architecture enables robust and interpretable feature learning across heterogeneous data sources, facilitating the estimation of spectral proxies of chlorophyll content, plant vigor, and disease stress indicators rather than direct physiological measurements. Experiments were performed on seven publicly available benchmark datasets encompassing diverse spectral and visual domains: three hyperspectral datasets (Indian Pines with 16 classes and 10,366 labeled samples; Pavia University with 9 classes and 42,776 samples; and Kennedy Space Center with 13 classes and 5211 samples), two plant disease datasets (PlantVillage with 54,000 labeled leaf images covering 38 diseases across 14 crop species, and the New Plant Diseases dataset with over 30,000 field images captured under natural conditions), and two chlorophyll content datasets (the Global Leaf Chlorophyll Content Dataset (GLCC), derived from MERIS and OLCI satellite data between 2003–2020, and the Leaf Chlorophyll Content Dataset for Crops, which includes paired spectrophotometric and multispectral measurements collected from multiple crop species). To ensure statistical rigor and spatial independence, a block-based spatial cross-validation scheme was employed across five independent runs with fixed random seeds. Model performance was evaluated using $R^2$, RMSE, F1-score, AUC-ROC, and AUC-PR, each reported as mean ± standard deviation with 95% confidence intervals. Results show that Agri-DSSA consistently outperforms baseline models (PLSR, RF, 3D-CNN, and HybridSN), achieving up to $R^2=0.86$ for chlorophyll content estimation and F1-scores above 0.95 for plant disease detection. The attention distributions highlight physiologically meaningful spectral regions (550–710 nm) associated with chlorophyll absorption, confirming the interpretability of the model’s learned representations. This study serves as a methodological foundation for UAV-based and field-deployable crop monitoring systems. By unifying hyperspectral, chlorophyll, and visual disease datasets, Agri-DSSA provides an interpretable and generalizable framework for proxy-based vegetation stress estimation. Future work will extend the model to real UAV campaigns and in-field spectrophotometric validation to achieve full agronomic reliability. Full article

We investigated the diagnostic efficacy of the universal transport medium™ (UTM^®) as a transport medium for aqueous humor polymerase chain reaction (PCR) testing in patients with clinically suspected viral uveitis. This retrospective study included 31 patients (31 eyes) with presumed viral uveitis who underwent anterior chamber sampling and compared the viral detection rates between using UTM^® and conventional test tubes only. The positivity rate for any virus, including cytomegalovirus, herpes simplex virus, and varicella zoster virus, was significantly higher in the UTM^T® group than in the test tube group (64.3% vs. 23.5%, p = 0.033). Logistic regression analysis also revealed that the use of UTM^® significantly increased the PCR positivity rate (odds ratio, 5.850; 95% confidence interval, 1.222–27.994; p = 0.027). Thus, the use of UTM^® was associated with improved detection of causative pathogens in patients with presumed viral uveitis. Full article