Search Results (319)

Background/Objectives: Deep learning models for digital mammography sensor data are increasingly deployed across hospitals using different X-ray detector technologies and patient populations. Whether models trained on one sensor platform and population maintain accuracy when transferred to another has not been tested for the latest generation of mammography-specific foundation models under one controlled protocol. Methods: We fine-tuned five backbone architectures (ResNet-50, DINOv2-B14, Rad-DINO, Mammo-CLIP B5, and Mammo-FM) on CBIS-DDSM (film-digitized, USA, n = 714 validation) with three seeds, ablated a density-aware focal loss across three auxiliary weights, and evaluated transfer to three external sensor cohorts: CMMD (full-field digital, China, n = 1032), DMID (mixed digital, India, n = 509), and MIAS (film-digitized, UK, n = 322). Significance used paired DeLong z-tests with Benjamini–Hochberg FDR correction; temperature scaling tested post hoc recalibration at all transfer targets. Results: Within this single-source three-seed evaluation, ResNet-50 outperformed all four foundation models on CBIS-DDSM (AUC 0.867 vs. 0.847, 0.846, 0.813, and 0.703; all gaps p_adj < 0.05). The density-aware focal loss degraded both AUC and calibration at every weight tested. At transfer, every model lost 0.165 to 0.320 AUC points relative to in-distribution performance, with sensitivity at 95% specificity collapsing from 0.31 to 0.47 in-distribution to 0.11 to 0.22 across the three external targets. A per-seed Stouffer meta-analysis confirms that Mammo-CLIP B5 and Mammo-FM significantly outperformed ResNet-50 on DMID and Mammo-CLIP on CMMD, after BH-FDR; MIAS comparisons remained directional only. In the extremely dense subgroup (BI-RADS D4), Mammo-FM reached AUC 0.870 versus ResNet-50 at 0.842, a directional observation whose 95% CIs overlap heavily at the n = 140 sample size and which we do not interpret as a statistically supported advantage. Conclusions: In this single training-source, three-seed protocol, mammography-specific pretraining did not deliver the in-distribution AUC premium reported in the originating papers, and no architecture reached a level at which transfer deployment without local validation would be defensible. We frame these as observations specific to the present protocol rather than as broader conclusions about foundation models for mammography classification. The findings argue for sensor-stratified and population-stratified external validation and for local recalibration as practical prerequisites before clinical use. Code and weights are released under MIT license. Full article

This article describes the development of a compact and affordable variable-temperature NMR instrument designed primarily to measure dynamic molecular motion in solids and liquids. The instrument consists of Lab-Tools’ Mk4 palm-top time-domain NMR spectrometer fitted with a Peltier-cooled variable-temperature probe inside a shimmed Halbach magnet. Measurement of NMR relaxation times T₁, T₂, and ${\mathrm{T}}_1\rho$ is possible over the temperature range −20 °C to 70 °C with cooling and heating rates, and data acquisition is controlled from an integrated mini-PC. The overall footprint of the instrument is roughly that of a shoe box, making both in-the-field and bench-top measurements possible. Applications of this instrument include measuring pore-size distribution in porous rocks, the viscosity of oils and tars trapped in porous rock, the properties of polymers, and the viscosity of the liquid components of foods (e.g. fruits, vegetables and seeds). Results of test measurements for calibrated oils and olive oil are presented together with measurements of molecular mobility in a solid polymer. Full article

The sustainability, crop production, and food safety of agriculture are increasingly challenged by microplastic pollution, as agricultural soils are the largest reservoirs and may serve as points of contact for plastic particles in the food chain. This review provides a comprehensive overview of plant materials, fate and uptake pathways, detection techniques, and the possible risks of microplastics in agriculture. Agroecosystems are also a source of microplastics, such as plastic mulch films, sewage sludge, compost and manure additives, wastewater irrigation, polymer-coated fertilizers, greenhouse materials, atmospheric deposition, and decomposition of discarded agricultural plastics. Their distribution and mobility in soil are controlled by polymer composition, particle size, morphology, density, surface ageing, soil texture, organic matter content, tillage practices, runoff, leaching, and soil biota. Recent data show that microplastics, especially smaller microplastics and nanoplastics, can attach to root surfaces, penetrate plants via cracks in roots, areas of lateral root development, and apoplastic pathways, and eventually move to tissues aboveground. Plant tissue detection is often accomplished by digestion of the sample, density separation, visual and fluorescence microscopy, Fourier-transform infrared spectroscopy, Raman spectroscopy, pyrolysis–gas chromatography mass spectrometry, and electron microscopy, but standardization of these methods remains a significant challenge. Microplastics can disrupt seed germination, root structure, nutrient absorption, photosynthesis, oxidative homeostasis, biomass buildup, yield development, and quality. Further, their capacity to transport additives, plasticizers, heavy metals, and persistent organic pollutants raises concerns about the transfer of contaminants to edible plant parts and their potential transfer to human diets. Further studies are needed focusing on field-realistic exposure conditions, long-term crop–soil interactions, nanoplastics behaviour, standardised analysis procedures, uptake and translocation pathways, edible crop risk assessments, and sustainable mitigation approaches to reduce microplastics in agroecosystems. Full article

Postharvest seed cleaning is critical for ensuring high-quality rice seeds suitable for storage and planting. Traditional cleaning systems, which are often limited to one or two sieves, are insufficient for removing all impurities, resulting in reduced seed purity and potential germination issues. This study was designed to enhance the rice seed cleaning system by integrating a high-efficiency blower with a triple-sieving mechanism. The system utilized three sieves with progressively smaller mesh sizes to systematically separate contaminants such as dust, broken grains, husks, and other foreign particles. A controlled airflow from the blower distributes rice seeds uniformly across the sieves, optimizing separation while minimizing mechanical damage. Compared to existing conventional systems, the proposed design demonstrated significantly improved cleaning performance, resulting in higher seed purity levels and overall enhanced seed quality. The triple-sieve configuration, coupled with precise airflow control, led to more effective impurity removal and uniform seed handling. The improved seed-cleaning system offers several agronomic benefits, including reduced postharvest losses, increased seed germination rates, and improved crop establishment. By producing cleaner, higher-quality seeds, this system has the potential to support more efficient and productive rice cultivation. Full article

Physical uniaxial compressive tests were conducted on high porosity vesicular basalt specimens in the lab. The main experimental mechanical parameters (i.e., peak strength and elastic constants) were used to calibrate numerical models in the 2-D PFC. Two different contact bond models were applied during the numerical analysis, namely, the linear parallel bond model and the flat-joint model. Also, different seed values were tested to generate distinct two-dimensional pore structures. Further, two grain size distributions were tested: a coarser sized one and a finer sized one. The effects of these bond models and parameters on the fracturing response of the rock were studied. Two simple mathematical criteria were also proposed for the accurate determination of the cracking thresholds from the numerically derived crack count curve. The numerical results were compared with the laboratory derived ones, and the differences were acceptable. Ultimately, via the coupled experimental and numerical approach, we were able to physically interpret the micro- and macrocracking response of the rocks. Full article

Fusarium fungi are known to infect table beet (Beta vulgaris subsp. vulgaris) plants at various stages of development worldwide. Fusarium root rot, which develops post-harvest during long-term storage, is of particular economic significance. In Russia, there is no up-to-date information about the species diversity of pathogens causing this disease of table beets, which determined the purpose of this study. A total of 28 Fusarium isolates were collected from affected beet roots grown in the Moscow region of the Russian Federation from 2018 to 2023 years. Molecular phylogeny based on the TEF-1α and RPB2 genes in combination with morphological characterization showed that five Fusarium species were involved in the pathogenesis of Fusarium root rot of table beet during storage: F. acuminatum (43% of the total number of isolates), F. avenaceum, F. campestre (FTSC); F. sporotrichioides (FSAMSC) and F. solani (FSSC). At the same time, the species F. acuminatum, F. campestre, and F. sporotrichioides were first discovered on beet root in the Russian Federation. Temperature sensitivity of the identified species was studied at 5 °C and 25 °C. According to the value of the cold sensitivity index (CTI) on the nutrient medium and native substrate, the isolates were distributed differently: F. campestre (0.32) > F. acuminatum (0.22) > F. avenaceum (0.21) > F. sporotrichioides (0.19) > F. solani (0.20) and F. acuminatum (0.32) > F. campestre (0.21) > F. solani (0.03) > F. avenaceum and F. sporotrichioides (0.01), respectively. This confirms the need to study the pathogenic properties of isolates on a natural substrate (host plant) under different temperature conditions. When infected with the dominant and most aggressive species F. acuminatum, there was a high variation in the size of the affected area, depending on the genotype of the lines, under both temperature conditions (Va = 2–8 mm³ at 5 °C and Va = 31–1760 mm³ at 25 °C). Therefore, this species can be considered to be the most objective differentiating factor in assessing the resistance of table beet roots to fusarium rot, which determines the need to include it in the breeding process for creating resistant varieties and hybrids for the Central region of Russia. The data obtained in this study are of great importance for developing strategies for managing Fusarium fungi associated with Fusarium rot of beet-root during storage. The research results will also be relevant for other vegetable crops that remain fresh for long periods of time or undergo vernalization in the case of seed production at low temperatures. Full article

Tomato leaf diseases substantially reduce tomato yields and quality and remain a persistent challenge for efficient crop management. Although deep learning-based detectors have achieved strong accuracy in controlled benchmarks, many existing solutions are still difficult to transfer to resource-constrained agricultural systems because they rely on high-end GPUs, consume considerable power, and often lose performance after deployment on embedded devices. To address this practical gap, this study proposes HGS-YOLO, a system-oriented deployable lightweight adaptation of YOLOv11 for leaf-level tomato disease detection, together with an end-to-end edge sensing pipeline for low-power agricultural deployment. The main contribution lies in the coordinated system-level co-design of model structure, optimization, and deployment rather than in a novel detector architecture. Specifically, YOLOv11 is adapted through three coordinated modifications: an HGNetV2 backbone for efficient feature extraction, an HS-FPN neck with channel attention for lightweight multi-scale fusion, and an MPDIoU loss function for more stable localization optimization. Beyond the model architecture, the study establishes a complete engineering pipeline that includes training, optimization, post-training quantization, and hardware deployment with BPU acceleration on a D-Robotics RDK X5 handheld platform. Comprehensive benchmark experiments indicate that HGS-YOLO achieves 93.6% ${\mathrm{mAP}}_{50}$ and 72.1% mAP@[0.5:0.95] with 86.5% recall, only 1.3 M parameters, and a 3.1 MB model size, substantially reducing the model complexity and storage cost relative to the YOLOv11 baseline. A three-seed retraining comparison shows that HGS-YOLO trades roughly 0.5 ${\mathrm{mAP}}_{50}$ points for this compactness (a statistically significant but small concession) and recovers the cost on the deployment side: on the RDK X5 chip, HGS-YOLO is the fastest, most memory-efficient, and lowest-power model among all compared detectors. Indoor deployment tests using separately collected tomato leaf samples further achieve 90.3% ${\mathrm{mAP}}_{50}$, 82.3% recall, 89.0% precision, 25.0 ± 0.4 ms end-to-end latency, 40.0 ± 0.6 FPS, and 9.8 ± 0.4 W average system power. After PTQ, the ${\mathrm{mAP}}_{50}$ drops from 93.6% to 93.0% on the same benchmark; because this figure was measured under controlled imaging conditions, it is presented as an in-distribution reference point rather than as evidence of robustness in the open field. We also took the handheld system into a working tomato greenhouse for a small outdoor field round, where it ran end-to-end and produced on-device disease detections under natural sunlight, specular highlights, partial occlusion, background clutter, and handheld motion blur. These results show that HGS-YOLO reaches a good balance of accuracy, efficiency, and deployability and that it works in the field on an independent small-scale test; validating it more widely across sites, seasons, and weather is left to future work. Full article

Tiger nuts are rich in various nutrients, with starch being a key component that holds potential for applications in foods with lower digestibility. However, the varietal dependence of their functional properties has not yet been comprehensively characterized. In this study, starches from different tiger nut varieties—Yunnan Large-seeded Tiger Nut (YLL), Yunnan Small-seeded Tiger Nut (YSL), and Zhongyousha No.1 (ZYS)—were systematically analyzed to investigate differences in their multi-scale structures, functional properties, and digestibility. The results revealed significant varietal differences in amylose content (22.82–26.99%), granule size (D₅₀ 8.66–10.23 μm), and short-range molecular order. All starches exhibited A-type crystalline structures, though their relative crystallinity (25.40–30.60%) differed significantly. In vitro digestion profiles demonstrated a two-phase hydrolysis pattern across all varieties, a rapid digestion phase (0–30 min) followed by a slow digestion phase (30–120 min), with resistant starch content ranging from 33.55% to 38.31%. Among the three varieties, higher amylose content and crystallinity were generally associated with enhanced digestive resistance and lower peak viscosity, while gelatinization temperature appeared to be more closely related to granule size than to crystallinity. Peak gelatinization temperature (T_P) ranged from 67.40 to 68.16 °C and peak viscosity (PV) from 7030.0 to 7749.5 mPa·s, with YSL, which had a relatively broad granule size distribution and the lowest crystallinity, exhibiting the highest T_P and PV. This study provides a reference for understanding the structure–property relationships of tiger nut starches across different varieties and their potential application in functional foods. Full article

The recycling of polyethylene terephthalate (PET) is gaining increasing importance, as it enables the conversion of plastic waste into valuable raw materials and contributes to a circular economy. Recent research has primarily focused on optimizing the depolymerization step of PET glycolysis, while downstream processes often overlook what are at least equally critical downstream steps in recovering the monomer bis(2-hydroxyethyl) terephthalate (BHET). The implementation of a water-free PET glycolysis process eliminates challenges related to internal solvent and homogeneous catalyst recycling that commonly occur in conventional processes. This study, therefore, focuses on BHET crystallization and filtration as key downstream unit operations. Two nucleation strategies, gassing and seeding, were investigated and compared with experiments without a nucleation strategy. The aim was to achieve reproducible process control during crystallization and to obtain crystals with good filterability, which can be critical for subsequent steps in the product purification process. Experiments without a nucleation strategy showed poor reproducibility. In contrast, gassing and seeding improved crystallization control, particularly regarding nucleation temperature and relative crystallization yield. However, these strategies also resulted in significantly prolonged filtration times due to differences in filter cake properties. The anisotropic crystals exhibited a broad particle size distribution with a high fraction of fine particles, leading to small and heterogeneous pores in the filter cake. Limited crystal growth was identified as the main cause of the unfavorable filtration behavior. Full article

The Automatic Identification System (AIS) generates massive volumes of real-world ship trajectory data, providing a critical foundation for maritime ship-type classification. However, existing methods often struggle to simultaneously capture long-range temporal dependencies, maintain computational efficiency, and ensure model interpretability, making accurate multi-class classification challenging in real-world maritime environments. To address these limitations, this study proposes a robust and efficient hybrid framework that integrates a Feature Transformer module for deep temporal feature extraction with a LightGBM model for ensemble classification. The multi-head self-attention within the Feature Transformer captures long-range dependencies in preprocessed AIS sequences to generate compact 64-dimensional trajectory fingerprints. These deep representations are concatenated with 103 carefully designed kinematic, geometric, statistical, frequency-domain, and segment-level features and fed into a LightGBM classifier for final ship-type identification. We evaluate the framework on a real-world AIS dataset of 2196 trajectories collected between 2019 and 2023, covering 14 ship types under a natural long-tail distribution. Across five random seeds, the proposed hybrid model achieves 78.06% ± 1.15% accuracy (95% CI) and 74.09% ± 1.82% Macro-F1 (95% CI), significantly outperforming Transformer-only (65.09% accuracy) and LightGBM-only (66.85%) baselines, with paired statistical tests confirming the improvement (McNemar χ² = 172.07, p < 10⁻³⁹ vs. Transformer; χ² = 92.24, p < 10⁻²¹ vs. LightGBM). The hybrid model offers ultra-fast inference at 0.051 ms per trajectory on GPU at batch size 128 (≈19,500 samples/s), and provides instance-level interpretability via SHapley Additive exPlanations (SHAP) analysis. These properties make the framework practical for near-real-time maritime traffic monitoring and decision support. Full article

The rapid growth of the Internet of Things (IoT) has introduced significant cybersecurity challenges due to the heterogeneity, scale, and limited protection capability of connected devices. Although machine learning has been widely adopted for IoT intrusion detection, many existing studies still rely primarily on closed-world evaluation settings, unequal baseline comparison budgets, fixed decision thresholds, and limited integration of explainability into model assessment. To address these issues, this paper proposes an explainable XGBoost-based framework for IoT attack detection with unseen attack family evaluation using the large-scale CICIoT2023 dataset. In the proposed framework, IoT traffic is formulated as a binary classification task that distinguishes benign from malicious flows. The study integrates two complementary evaluation protocols: (1) closed-world stratified 10-fold cross-validation for in-distribution performance assessment and (2) unseen attack family evaluation, in which one malicious family is excluded from training and used only for testing under a zero-day-like but single-dataset condition. A fair-budget experimental design is adopted to compare seven representative models under the same training budget, including default XGBoost, optimized XGBoost, Random Forest, LightGBM, CatBoost, Logistic Regression, and a simple multilayer perceptron. To improve reproducibility and operational validity, the revised framework further reports the sampling strategy, split-overlap audit, XGBoost hyperparameter search protocol, repeated unseen-family evaluation, validation-based threshold calibration under fixed-FAR constraints, cost-sensitive threshold analysis, and XGBoost-native SHapley Additive exPlanations (SHAP) compatible feature contribution analysis. The closed-world results show that tree-based ensemble methods clearly outperform the linear and shallow neural baselines. Random Forest achieves the highest closed-world macro-F1 of 0.9713, followed by LightGBM with 0.9602 and optimized XGBoost with 0.9566. In the fair-budget unseen-family setting under the default threshold, Random Forest again obtains the highest mean macro-F1 of 0.8433 and the lowest false negative rate (FNR) of 0.0712, but it also produces a substantially higher false alarm rate (FAR = 0.0536). By contrast, optimized XGBoost provides a lower-FAR default operating point, achieving a mean macro-F1 of 0.8194, Matthews correlation coefficient (MCC) of 0.7067, FAR of 0.0086, and FNR of 0.2996. Repeated unseen-family experiments over five random seeds confirm the same trade-off: Random Forest provides stronger recall-oriented detection, whereas optimized XGBoost provides a lower-FAR default operating point. After validation-based threshold calibration at an approximate FAR target of 0.01, Random Forest achieves the strongest calibrated recall-oriented performance, with macro-F1 of 0.8754, MCC of 0.7757, FNR of 0.2000, and attack recall of 0.8000. Optimized XGBoost remains competitive at the same FAR target, with macro-F1 of 0.8323, MCC of 0.7193, FNR of 0.2760, and attack recall of 0.7240. The explainability analysis indicates that the optimized XGBoost detector relies mainly on TCP control-flag, temporal, and packet-statistical features, with rst_count, IAT, urg_count, Tot size, Number, Header_Length, and Magnitude among the most influential variables. Local contribution tables for representative true-positive, false-positive, false-negative, and true-negative cases further improve the readability of the explanation results and confirm that native pred_contribs reconstructs the model margin with negligible numerical error. Overall, the results show that the most appropriate model depends on the deployment objective: Random Forest is preferable when minimizing missed attacks under a calibrated FAR constraint is prioritized, whereas optimized XGBoost remains a strong primary model for an explainable low-FAR XGBoost-based framework that emphasizes scalability, operational conservativeness, and native contribution-based interpretation. Full article

Pycnospatha is a small and poorly known genus of Araceae distributed in Indochina and currently comprising only two accepted species. During botanical surveys in Si Sa Ket Province, eastern Thailand, an unusual population of Pycnospatha was discovered in a dry dipterocarp forest and found to differ from both P. arietina and P. palmata. Here, we describe this plant as a new species, Pycnospatha phanomdongrakensis. The new species is distinguished by a combination of characters, including a slender habit, shorter petiole and peduncle, a medium-sized spathe, a short and dense spadix, a distinctly curved style directed toward the apex of the spadix, a geophilous and ovoid infructescence, obovate berries, and asymmetrically ovoid seeds. The new taxon is currently known only from a single population in the Phanom Dong Rak mountain range. A preliminary conservation assessment is provided, and the species is treated as Critically Endangered (CR) following IUCN guidelines. An identification key to all species of Pycnospatha is also presented. The discovery of this new species highlights the continuing importance of field-based taxonomy in revealing overlooked aroid diversity in the seasonally dry forests of eastern Thailand. Full article

A first study analyzed the effect of the presence of a highly frequented tourist trail on the size and reproductive capacity of Globularia alypum, a Mediterranean shrub of conservation interest. In Sardinia, this species is a typical example of a plant with Extremely Small Populations (PSESPs), restricted to a natural area embedded within an urban matrix, which makes it particularly vulnerable to ecological pressures. In this second contribution, the investigation expands to the entire population of the species distributed across different habitats. The possible correlations between vegetative and reproductive traits of the plant are examined, along with the influence exerted by both habitat type and varying levels of human disturbance. To evaluate potential drivers of its persistence, morphological (H, diameter and plant volume) and reproductive traits (number of flowers, number of fruits and number of seed per plant) were recorded at the individual level. Additionally, to assess human disturbance (consisting mainly of trampling), the presence of trails was used as a proxy and, accordingly, each plant was categorized following its relative position to the nearest path according to three categories: Near Trail (NT), Mid-Trail Distance (MTD), or Far from Trail (FT). A total of 114 individuals distributed across four habitat types were measured. Statistical analyses revealed only marginal associations between habitat type and vegetative or reproductive traits. While trail proximity did not influence flower and fruit production, plant volume tended to be greater in individuals located farther from trails, suggesting a potential, albeit limited, effect of reduced human pressure on plant growth. These findings highlight the importance of understanding subtle ecological interactions that shape the persistence of PSESPs in urban close to nature area and provide valuable insights for developing targeted conservation and management strategies. Full article

As demand for sustainable plant protein rises, pumpkin seeds emerge as a promising but underutilized source. Conventional alkaline extraction (ALK) often impairs protein functionality, prompting interest in non-thermal alternatives. This study systematically compared the functional, colloidal, and structural properties of pumpkin seed protein isolates obtained via ALK (conducted at 50 °C), ultrasound-assisted (UAE), and microwave-assisted extraction (MAE). UAE produced the highest extraction yield (50.07%), superior overall solubility, greatest water and fat absorption capacities, and lowest least gelation concentration (12%). Furthermore, UAE best preserved native protein secondary structure (retaining 43.45% alpha-helix), as quantified by FTIR peak deconvolution, and maintained an intact, flake-like morphology under scanning electron microscopy (SEM), yielding the most uniform particle size distribution. Conversely, MAE achieved the highest protein content (73.53%) and the most negative zeta potential, leading to the highest emulsifying and foaming capacities despite inducing a bimodal particle size and irregular, porous surface morphology. ALK performed the poorest across structural and functional metrics, severely denaturing the proteins due to combined alkaline and thermal stress. UAE is recommended for applications requiring optimal solubility and gelation, whereas MAE is highly effective for emulsion- and foam-based food systems, reinforcing pumpkin seeds as a viable sustainable protein ingredient. Full article

This paper explores the role of the Taurus Mountains in shaping species differentiation and biogeographical regions, including distinct ecological zones at various altitudes. The genus Aethionema is a monophyletic group in tribe Aethionemeae, subfamily Aethionemoideae (Brassicaceae), a sister group to the rest of the family. Aethionema has significant taxonomic complexity, particularly in Türkiye, where the genus has the highest species diversity. In this study, four new species (Aethionema kadriyeae, A. uysalii, A. beysehirense, and A. ermenekense) are described based on morphological, palynological and phylogenetic analyses. The diagnoses, detailed descriptions, distribution maps, and illustrations of the new species are provided. Pollen and seed morphology, including detailed measurements of size, ornamentation, and shape, is given. Phylogenetic analyses using DNA sequences from nuclear region (ITS) and chloroplast region (rpl32-trnL^UAG) were conducted to determine the evolutionary relationships within the genus. Overall, this research provides new insights into the biodiversity and evolutionary history of Aethionema in Türkiye, highlighting the significance of the Taurus Mountains in supporting rich ecological diversity. Full article