Search Results (696)

Late blight in tomato caused by Phytophthora infestans can lead to severe economic losses. Early detection is essential for effective disease management. This study investigates the spectral characteristics of healthy and late blight-infected tomato leaves and plants using non-destructive hyperspectral imaging and proposes a cost-effective machine vision system for early disease detection. Hyperspectral images from seven batches of leaf sets and six batches of whole plant sets were taken hourly over a 96 h period under both controlled and artificially infected conditions. The hyperspectral data cubes were processed with an image analysis model that identified healthy vs. infected regions. Key wavelengths (77 from leaf datasets and 24 from plant datasets) were selected using recursive feature elimination and analysed with four machine learning classifiers: k-nearest neighbour, support vector machine, random forest, and artificial neural network. The models differentiated healthy and infected tissue with high accuracy (98–99%). The hyperspectral data were simplified into a multichannel image with most informative wavelengths, using a custom spectral index and binary decision rule. Experimental limitations were addressed, and a conceptual design of practical hardware was proposed: a monochrome camera combined with a multichannel light source and polariser mounted on mobile equipment. Although further trials will be needed, this proof-of-concept study and conceptual hardware design can be adapted in other crops facing similar disease challenges. Full article

Background/Objectives: Impacted mandibular third molars are associated with distal caries of the adjacent second molar and pericoronal follicle enlargement. This study evaluated the independent associations of both Winter angulation and Pell–Gregory classification with these complications on panoramic radiographs. Methods: This retrospective radiographic study evaluated 1491 impacted mandibular third molars from 1487 patients aged 18 years or older, reported in accordance with the STROBE guidelines for observational studies. Impaction patterns were classified using both the Winter and Pell–Gregory systems. Follicular space exceeding 2.5 mm was defined as follicle enlargement, measured digitally on panoramic images. Distal caries was defined as a radiolucent lesion on the distal surface of the adjacent second molar. Binary logistic regression (univariate and multiple models) was used to assess associations. Results: Among the 1487 patients (745 females, 742 males; mean age 25.21 ± 4.15 years), neither Winter angulation nor Pell–Gregory classification was associated with distal caries in univariate or multivariable analysis (all p > 0.05). Pericoronal follicle enlargement was significantly associated with deeper impaction: Pell–Gregory Class I carried lower risk than Class III (OR 0.296; 95% CI 0.184–0.478; p < 0.001), and Class A carried lower risk than Class C (OR 0.254; 95% CI 0.161–0.400; p < 0.001). Younger age independently predicted follicle enlargement (OR 0.964/year; p = 0.028). Winter angulation was not independently associated with follicle enlargement (all p > 0.05). Conclusions: Neither Winter nor Pell–Gregory classification predicted distal second molar caries on panoramic radiography. Deeper Pell–Gregory impaction positions and younger age were independently associated with pericoronal follicle enlargement, supporting early radiographic surveillance in young patients with deep impactions. Full article

Accurate and non-destructive assessment of drought stress is important for improving lettuce production and supporting timely crop management. This study presents a detection-guided deep learning framework for plant-level drought-stress assessment in hydroponically grown lettuce using bird’s-eye-view RGB images. The study further investigates whether canopy segmentation can improve classification performance by reducing irrelevant background information. The framework was evaluated using 2190 images collected across three independent cultivation cycles in which drought stress was induced by isolating the plant root zones from the nutrient solution. In the first stage, YOLO-based object detection was used to localize individual plants, with YOLO26m achieving the highest detection performance of 99.4% mAP@0.5. The detected regions were subsequently used as spatial prompts for zero-shot canopy segmentation using the Segment Anything Model (SAM), with SAM ViT-B achieving a mean IoU of 0.9864. Six convolutional, transformer-based, and hybrid classification architectures were then evaluated independently using YOLO-cropped and SAM-segmented plant images. Segmented inputs consistently improved classification performance, with MaxViT-S achieving the highest binary test accuracy of 96.3%. The framework further distinguished time-defined pre-stress, early-stress, and late-stress periods with an accuracy of 92.4%. Plant-level generalization was further assessed using six-fold leave-one-plant-out cross-validation, resulting in a mean test accuracy of 90.25 ± 1.78% on unseen plants. These findings demonstrate that RGB-based plant-level analysis can support non-destructive drought-stress assessment and that canopy segmentation improves classification by reducing background influence. Full article

Background: The rotator cable is thought to contribute to load transmission and preservation of shoulder function in rotator cuff tears; however, its detectability on arthroscopy and magnetic resonance imaging (MRI), agreement between these modalities, and clinical relevance remain incompletely defined. This study evaluated rotator cable visibility on arthroscopy and MRI, factors associated with arthroscopic cable visibility, and its relationship with preoperative active shoulder motion. Methods: This retrospective cross-sectional study included 128 patients who underwent shoulder arthroscopy for supraspinatus tears between January 2019 and February 2023. Arthroscopic video recordings were reviewed for rotator cable visibility. Standardized preoperative MRI review was available in 58 patients. Agreement between MRI and arthroscopic visualization was assessed using Cohen’s kappa and percentage agreement measures. Multivariable binary logistic regression was performed to identify factors independently associated with arthroscopic cable non-visibility. Results: The rotator cable was visible arthroscopically in 79 of 128 patients (61.7%) and on MRI in 38 of 58 patients (65.5%). Overall agreement between MRI and arthroscopy was 67.2% (95% CI, 53.7–79.0%), with a Cohen’s kappa of 0.315 (95% CI, 0.090–0.540; p = 0.011), indicating fair agreement. Positive and negative percent agreement were 78.8% and 52.0%, respectively. Although increasing age was associated with cable non-visibility in univariable analysis, this association did not remain statistically significant after multivariable adjustment. Higher Lafosse grade was independently associated with cable non-visibility in the overall cohort (adjusted OR, 1.43 per grade; 95% CI, 1.04–1.97; p = 0.028), whereas increasing tear size was independently associated with cable non-visibility among patients with full-thickness tears (adjusted OR, 2.78 per category; 95% CI, 1.19–6.52; p = 0.019). The unadjusted association between massive tear size and reduced MRI cable visibility did not remain significant after false discovery rate adjustment (q = 0.276). No statistically significant associations were detected between cable visibility or MRI-measured cable dimensions and preoperative active abduction or forward elevation. Conclusions: MRI and arthroscopy demonstrated fair agreement in the assessment of rotator cable visibility. After multivariable adjustment, cable non-visibility was more closely associated with tear-related characteristics than with patient age. No statistically significant associations were detected between cable characteristics and the assessed preoperative range-of-motion measures; however, smaller or moderate associations cannot be excluded, particularly within the MRI subgroup. Rotator cable visibility should primarily be interpreted as a marker of detectability and tear morphology rather than as a direct surrogate for structural integrity or shoulder function. Full article

Increasing the accuracy of age-at-death estimation using two skeletal-characteristic stages can enhance confidence in biological identification using forensic science. Transition analysis and the inverse prediction method with a Bayesian approach was proposed in this study to estimate age from skeletal characteristics measured as binary variables. The Bayesian approach with adaptive rejection sampling was employed to derive the posterior distributions of the transition model parameters and the age classification threshold in order to reverse the age-at-death estimation from a binary predictor. The posterior odds ratio was proposed to assess the value of observed evidence for the age estimation. Subsequently, the efficiency of our proposed method, measured by the percentage of correct classification, was evaluated by Monte Carlo simulation and compared with the Maximum Likelihood Estimation with the inverse prediction method. The simulation results supported the advantages of our proposed method, especially when using small sample sizes. In an application involving chest X-ray images with two chest plate ossification stages, the results showed that our method could identify suitable features of the chest plate, providing good age-prediction performance with a high percentage of accuracy. Full article

Fine-grained poisonous mushroom recognition in natural scenes is challenging because of complex backgrounds, subtle morphological differences, and the limited interpretability of model decisions. To address these challenges, this paper proposes Att-FiLM, an attention-enhanced multi-scale Feature-Wise Linear Modulation network for poisonous mushroom image recognition. The model adopts an asymmetric dual-backbone architecture in which a frozen ConvNeXt-Base branch provides global semantic priors, while a trainable EfficientNet-B0 branch learns local discriminative features. Rather than directly concatenating heterogeneous features, Att-FiLM generates scale and shift parameters from semantic features and performs channel-wise modulation on multi-scale EfficientNet features at Stage 2 and Stage 4. This mechanism enables global semantic information to guide local feature learning while reducing feature redundancy and semantic inconsistency. Experimental results show that Att-FiLM achieves an Accuracy of 95.58% and an F1-score of 0.9455 on the poisonous/edible binary classification task. On the 190-class species-level classification task, it achieves a Top-1 Accuracy of 93.63% and a Macro-F1 of 0.9347. Interpretability analysis further shows that decision-relevant responses are frequently associated with morphologically relevant regions, including gills, annuli, volvae, and cap textures. These results indicate that Att-FiLM provides effective recognition performance together with interpretable decision evidence for mushroom recognition in complex natural scenes. Full article

This article presents an empirical comparison of four automatic face recognition methods: Principal Component Analysis (PCA), Local Binary Patterns (LBP), the deep neural network FaceNet, and an original hybrid approach proposed by the authors, referred to as FLLF (Feature-Level Late Fusion). The experiments were conducted on a subset of the VGGFace2 database, comprising approximately 480 classes in the training set and 60 classes in the validation set. For closed-set identification, a new test set (20%) was extracted from the training set. Classification accuracy, training and inference times, and prediction confidence distributions were evaluated for each method. The results show a clear advantage of the deep learning approaches: FaceNet achieved an accuracy of approximately 98% with only five training images per person, whereas the classical methods—PCA and LBP—reached only approximately 7% and 22%, respectively. The proposed FLLF method, which fuses FaceNet embeddings with PCA-whitened LBP descriptors at the feature level and classifies them with a calibrated linear SVM, further improved accuracy to approximately 98.5% and produced the highest prediction confidence values of all tested methods. However, calibration quality was not directly assessed using standard metrics such as expected calibration error or reliability diagrams, so this observation should be interpreted as a confidence-distribution shift rather than a formal calibration improvement. The article also discusses the theoretical foundations of each algorithm, their respective advantages and limitations, and the architecture of the software system implemented for this study. Full article

In this paper, we present deep learning models for classifying ancient-coins using transfer learning applied to a custom dataset curated from a specialized collection. The dataset was meticulously developed through rigorous preparation and quality screening; the retained fine-grained classes are imbalanced. We initially performed binary classification before progressing to separate fine-grained assessments of DenseNet121, EfficientNetB0, EfficientNetV2S, and ResNet50. In confirmatory experiments using a physical-coin-aware 70/15/15 split and five training seeds, the coarse CITY-versus-NABATAEAN task remained approximately perfectly separable. Fine-grained performance was lower: ResNet50 achieved 68.28 ± 2.45% CITY accuracy, whereas DenseNet121 achieved 72.89 ± 2.79% NABATAEAN accuracy. These results show that near-perfect coarse classification does not by itself imply reliable fine-grained attribution. This analysis underscores the importance of meticulous dataset preparation and strategic model selection in optimizing performance. A matched DenseNet121 ablation further showed that ImageNet initialization substantially improved the difficult fine-grained tasks compared with random initialization. The proposed framework supports digital documentation and analysis of ancient coin collections, while the current single-collection evaluation still limits claims of cross-collection generalization. Full article

Background/Objectives: Histopathology-correlated evidence for preoperative magnetic resonance imaging (MRI) in penile cancer remains limited. We evaluated multiparametric MRI with artificially induced erection for local and nodal staging in surgically treated penile cancer. Methods: This prospective, single-centre diagnostic accuracy study included consecutive patients who underwent preoperative multiparametric MRI between 2017 and 2024, with postoperative histopathology as the reference standard. The primary endpoint was binary local T-staging (≤T1 vs. ≥T2). Secondary endpoints were three-category T-staging (≤T1/T2/≥T3) and binary nodal staging (N0 vs. N+) in patients with pathological nodal verification. Two radiologists independently reassessed all 38 examinations in a post hoc reproducibility analysis. Results: Thirty-eight patients were included; all tumours were glans-based and treated with organ-sparing surgery. Binary local staging showed 68.4% accuracy (95% CI, 52.5–80.9%), 88.2% sensitivity, 52.4% specificity, 60.0% positive predictive value, 84.6% negative predictive value, and balanced accuracy of 0.703 (95% CI, 0.568–0.838). Overstaging predominated (10 false-positive vs. 2 false-negative classifications; p = 0.043). Three-category exact agreement was 52.6% (κ = 0.257). In the pathologically verified nodal subset (n = 30), accuracy was 83.3% (95% CI, 66.4–92.7%), sensitivity 78.6%, specificity 87.5%, and balanced accuracy 0.830 (95% CI, 0.691–0.970). Inter-reader agreement was 92.1% for binary T staging (κ = 0.837) and 97.4% for binary N staging (κ = 0.943). Conclusions: mpMRI was sensitive but insufficiently specific for ≥T2 disease and should not independently trigger treatment escalation. Nodal performance was encouraging but arose from a guideline-selected verified subset. These findings apply primarily to glans-based, organ-preserving cohorts and do not establish an incremental benefit of erection induction. Full article

Multimodal large language models are increasingly used for dermatologic queries, but whether Fitzpatrick skin type (FST) labels affect diagnostic accuracy is unknown. We evaluated 656 biopsy-confirmed photographs from the Diverse Dermatology Images (DDI) dataset under no-FST, DDI-concordant FST, and two DDI-discordant FST conditions using ChatGPT 5.2 Edu and Gemini 3.1 Pro browser configurations (5248 evaluations). Confirmed malignant diagnosis omission from the top three differential diagnoses (“lethal miss”) was a prespecified exploratory outcome. Gemini had higher ordinal accuracy than ChatGPT (odds ratio 2.32; 95% confidence interval 2.01–2.67; false discovery rate-adjusted p < 0.001). No FST prompt condition significantly improved accuracy over no-FST prompting; an equal-image-weighted sensitivity analysis yielded similar results. Among malignant images eligible for extreme discordance, extreme DDI-discordant prompting was associated with more ChatGPT lethal misses than no-FST prompting (paired odds ratio 5.0; p = 0.043); Gemini showed no significant paired shift (p = 1.00). Post hoc binary malignancy detection showed low no-FST specificity for ChatGPT (42.5%) and Gemini (22.9%), indicating frequent overcalling. Gemini showed lower no-FST accuracy for malignant FST V–VI images than for FST I–II and III–IV images. FST prompting provided no measurable diagnostic benefit. Neither configuration supports autonomous dermatologic use. Full article

In operational terms, post-fire landscapes require monitoring priorities that reflect hydrogeomorphic susceptibility rather than burned extent alone. This study tests whether European Forest Fire Information System (EFFIS) burned area evidence, combined with open terrain, rainfall, drainage, soil, land cover, and settlement context data, can be translated into source-to-output traceable monitoring priority units across Mediterranean and Iberian–Atlantic regions. The backbone integrates Decision-Making Trial and Evaluation Laboratory (DEMATEL) driver structuring, cell criticality score (CCS) screening from the CCS-V0 conventional weighted linear GIS baseline to CCS-V5 staged refinement, upper-quartile (Q75) hotspot topology, multi-criteria decision analysis/cost penalty (MCDA/COST-PEN) ranking, and quadratic unconstrained binary optimization (QUBO)-ready selected/reserve organization. Across Vesuvius–Campania, Attica, Cyprus, Portugal, and Spain, the screening outputs show procedural portability without implying geomorphological equivalence, field-confirmed hydrogeomorphic damage, or cross-theater hazard comparability: hotspot geometry, dominance, fragmentation, and candidate composition remain theater-dependent. In Vesuvius–Campania, CCS-V1 rainfall conditioning reduced Q75 hotspot clusters from 58 to 18 and increased the dominant cluster ratio from 23.1% to 63.07%; in Portugal, CCS-V5 contracted hotspot support from 17,240 to 862 cells while increasing dominance to 66.13%. The final QUBO-ready layer retained 11 of 32 candidate units. The contribution links (i) EFFIS/Moderate-Resolution Imaging Spectroradiometer (MODIS) source sector traceability and valid support construction, (ii) DEMATEL-informed CCS refinement and hotspot topology interpretation, (iii) role-based transfer testing across non-equivalent theaters, and (iv) QUBO-ready prioritization for verification-oriented post-fire monitoring. Full article

Background: Transvenous embolisation (TVE) has emerged as a promising treatment option for selected brain arteriovenous malformations (bAVMs). Objective: To evaluate the safety and efficacy of TVE-based treatment strategies for bAVMs. Methods: This Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020-compliant systematic review and meta-analysis was registered in PROSPERO (CRD420261397818). PubMed, Embase, Cochrane CENTRAL, and Scopus were searched on 20 May 2026, supplemented by citation screening. Studies reporting TVE for bAVMs in ≥5 patients were included. Binary outcomes were pooled using random-effects binomial-normal generalised linear mixed models (GLMMs) with a logit link under an intention-to-treat framework. Results: Twelve studies comprising 239 patients were included; 57.7% underwent combined transarterial/transvenous embolisation. Complete angiographic occlusion was 92.2% (217/239; 95% CI, 84.5–96.2%; I² = 0%) overall and 97.1% (149/160; 95% CI, 76.6–99.7%; I² = 0%) among patients with ≥3 months of imaging follow-up. Functional independence (mRS 0–2) was 85.0% (155/180; 95% CI, 68.7–93.7%; I² = 55%), and technical success was 97.1% (219/227; 95% CI, 91.5–99.0%; I² = 0%). GLMM pooled rates were 0.7% for ischaemic complications, 12.3% for haemorrhagic complications, 3.3% for permanent neurological morbidity, 2.3% for overall mortality, 0.8% for procedure-related mortality, and 13.9% for any procedure-related complication or death. Conclusions: TVE-based strategies are associated with high angiographic occlusion and favourable functional outcomes in carefully selected bAVMs, but the evidence is predominantly observational and safety estimates remain uncertain. As most patients underwent combined transarterial/transvenous treatment, these findings should not be interpreted as estimates of standalone TVE. Full article

Background: Pediatric head trauma is a common reason for emergency department presentation. Although most children have minor injuries, a small proportion harbor clinically important traumatic brain injuries requiring urgent intervention. Artificial intelligence (AI) may offer structured support in computed tomography (CT) decision making, but evidence regarding the performance of general-purpose large language models in pediatric head trauma remains limited. Objective: To evaluate the association between AI-based cranial CT recommendations and clinically meaningful outcomes in pediatric patients with blunt head trauma and to assess the diagnostic performance and clinical utility of the model. Methods: This retrospective single-center observational study included pediatric patients younger than 18 years with blunt head trauma who underwent cranial CT imaging and had complete outcome data. A general-purpose large language model generated binary CT recommendations (“CT recommended” or “CT not recommended”) using structured clinical information available at the time of emergency department presentation. The primary outcome was a composite adverse clinical outcome defined as the occurrence of at least one of the following: emergency surgical intervention, intensive care unit admission, intubation, neurological sequelae or mortality. Diagnostic performance metrics, calibration analysis and decision curve analysis were performed. Results: A total of 819 pediatric patients were included, and the AI model recommended CT in 530 patients (64.7%). The primary outcome occurred in 143 patients (17.5%) and was significantly more frequent in the CT-recommended group than in the CT-not recommended group (24.5% vs. 4.5%; OR 6.90, 95% CI 3.82–12.45; p < 0.001). Abnormal CT findings, emergency surgery, intubation and neurological sequelae were also significantly more common in patients for whom CT was recommended by the AI system. For the primary outcome, the AI recommendation demonstrated a sensitivity of 90.9%, specificity of 40.8%, positive predictive value of 24.5% and negative predictive value of 95.5%. Calibration analysis showed acceptable agreement between predicted probabilities and observed event rates. Decision curve analysis demonstrated greater net benefit than both the “treat-all” and “treat-none” strategies across a range of threshold probabilities. Conclusions: In this clinically selected cohort of pediatric patients with blunt head trauma who underwent cranial CT imaging, AI-based CT recommendations were strongly associated with adverse clinical outcomes and demonstrated high sensitivity and negative predictive value for identifying children at risk of clinically important events. These findings suggest that, within a clinically selected cohort of children who underwent cranial CT imaging, AI-generated CT recommendations were associated with clinically meaningful outcomes. However, these results should not be interpreted as validation of CT decision making in the broader pediatric head trauma population and require prospective validation in unselected cohorts. Full article

This study examined how different forms of tourism promotional materials are associated with visual attention and factual recall under natural viewing conditions. A total of 87 participants completed eye-tracking and memory-test tasks using text-only, image-only, and mascot–text composite promotional materials across three destination contexts: Hangzhou, Chongqing, and Beijing. Visual attention was assessed using fixation duration and fixation count, and memory performance was assessed using post-viewing multiple-choice responses. In the revised analysis, attention measures were aggregated at the participant–stimulus level, and memory accuracy was analyzed using participant-item-level binary response data. The results indicated that mascot–text composite materials were associated with significantly longer total fixation duration, whereas material-type differences in total fixation count were not significant. Memory analysis indicated that text-containing materials were associated with higher factual recall accuracy than image-only materials, while text-only and mascot–text composite materials showed comparable recall performance. Rather than supporting a fixed mascot-first processing route, the study provides exploratory evidence for flexible attention allocation and attention–memory coordination in tourism promotional communication. Full article

Color texture classification is an important task in computer vision, with applications in medical imaging, industrial inspection, remote sensing, and material analysis. This paper presents a hybrid framework that integrates Principal Component Analysis (PCA), Local Binary Patterns (LBPs), Wavelet Scattering Transform, and the XGBoost classifier for color texture classification. The proposed pipeline first performs image pre-processing, including resizing and denoising, followed by channel-wise feature extraction using LBP and Wavelet Scattering Transform on the Red, Green, and Blue channels independently. Then, the obtained feature vectors were concatenated, and PCA was applied on the fused feature space for dimensionality reduction and redundancy elimination before proceeding to XGBoost classification. This method not only leverages complementary information of Chroma and texture information but also achieves reduced dimensionality and computational burden. The finally optimized features were input into the XGBoost classifier for color texture classification, which is good at fitting non-linear dependency and includes a regularization to generalize better. Our proposed framework was tested on three benchmark color texture datasets: KTH-TIPS, Outex_10, and VisTex. Experimental results have demonstrated that on these three datasets, the average performance reaches 98.0% accuracy, 0.981 precision, 0.981 recall, and 0.979 F1-score, respectively. It demonstrates that the two selected complementary feature extraction methods provide a compact yet effective representation for color texture classification on these datasets. It is expected that the proposed framework serves as an efficient combination of established methods and as a good competitive baseline for color texture analysis. Future works will consider applying it to larger color texture datasets for general verification, enhancing its computational efficiency and automating the parameter selection process. Full article