Search Results (14,861)

Question answering from visually rich documents (VRDs) is the task of retrieving the correct answer to a natural language question by considering the content of textual and visual elements in the document, as well as the pages’ layout. To answer closed-ended questions that require a deep understanding of the hierarchical relationships between the elements, i.e., the full document-level understanding (FDU) task, state-of-the-art graph-based approaches to FDU model the pairwise element relationships in a graph model. Although they incorporate logical links (e.g., a caption refers to a figure) and spatial ones (e.g., a caption is placed below the figure), they currently disregard the semantic similarity among multimodal document elements, thus potentially yielding suboptimal scoring of the elements’ relevance to the input question. In this paper, we propose GRAS-FDU, a new graph attention network tailored to FDU. GATS-FDU is trained to jointly consider multiple document facets, i.e., the local, spatial, and semantic elements’ relationships. The results show that our approach achieves superior performance compared to several baseline methods. Full article

Recent advances in artificial intelligence (AI) have transformed ophthalmic diagnostics, particularly for retinal diseases. In this prospective, non-randomized study, we evaluated the performance of an AI-based software system against conventional clinical assessment—both quantitative and qualitative—of optical coherence tomography (OCT) images for diagnosing diabetic macular edema (DME). A total of 700 OCT exams were analyzed across 26 features, including demographic data (age, sex), eye laterality, visual acuity, and 21 quantitative OCT parameters (Macula Map A X-Y). We tested two classification scenarios: binary (DME presence vs. absence) and multiclass (six distinct DME phenotypes). To streamline feature selection, we applied paraconsistent feature engineering (PFE), isolating the most diagnostically relevant variables. We then compared the diagnostic accuracies of logistic regression, support vector machines (SVM), K-nearest neighbors (KNN), and decision tree models. In the binary classification using all features, SVM and KNN achieved 92% accuracy, while logistic regression reached 91%. When restricted to the four PFE-selected features, accuracy modestly declined to 84% for both logistic regression and SVM. These findings underscore the potential of AI—and particularly PFE—as an efficient, accurate aid for DME screening and diagnosis. Full article

(1) Background: Quantitative EEG (QEEG) offers potential for identifying objective neurophysiological biomarkers in psychiatric disorders and guiding neurofeedback interventions. This study examined whether three nonlinear QEEG metrics—Lempel–Ziv Complexity, Tsallis Entropy, and Renyi Entropy—can distinguish children with autism spectrum disorder (ASD) from typically developing (TD) peers, and assessed their relevance for neurofeedback targeting. (2) Methods: EEG recordings from 19 scalp channels were analyzed in children with ASD and TD. The three nonlinear metrics were computed for each channel. Group differences were evaluated statistically, while machine learning classifiers assessed discriminative performance. Dimensionality reduction with t-distributed Stochastic Neighbor Embedding (t-SNE) was applied to visualize clustering. (3) Results: All metrics showed significant group differences across multiple channels. Machine learning classifiers achieved >90% accuracy, demonstrating robust discriminative power. t-SNE revealed distinct ASD and TD clustering, with nonlinear separability in specific channels. Visual processing–related channels were prominent contributors to both classifier predictions and t-SNE cluster boundaries. (4) Conclusions: Nonlinear QEEG metrics, particularly from visual processing regions, differentiate ASD from TD with high accuracy and may serve as objective biomarkers for neurofeedback. Combining complexity and entropy measures with machine learning and visualization techniques offers a relevant framework for ASD diagnosis and personalized intervention planning. Full article

Background and Clinical Significance: Patients with a neurodegenerative condition known as posterior cortical atrophy (PCA) can present with attention impairments across a variety of cognitive contexts, but the consequences of these are little explored in example of single word reading. Case Presentation: We present a detailed single-case study of KL, a local resident of South Wales, a patient diagnosed with posterior cortical atrophy (PCA) in 2018, whose reading and letter-naming abilities are selectively disrupted under non-canonical visual presentations. In particular, KL shows significantly impaired accuracy performance when reading words presented in tilted (rotated 90°) format. By contrast, his reading under conventional horizontal (canonical) presentation is nearly flawless. Whilst other presentation formats including, mixed-case text (e.g., TaBLe) and vertical (marquee) format led to only mild performance decrements—even though mixed-case formats are generally thought to increase attentional ‘crowding’ effects. Discussion: These findings indicate that impairments of word reading can emerge in PCA when visual-attentional demands are sufficiently high, and access to ‘top down’ orthographic information is severely attenuated. Next, we explored a cardinal feature of attentional dyslexia, namely the word–letter reading dissociation in which word reading is superior to letter-in-string naming. In KL, a similar dissociative pattern could be provoked by non-canonical formats. That is, conditions that similarly disrupted his word reading led to a pronounced disparity between word and letter-in-string naming performance. Moreover, different orientation formats revealed the availability (or otherwise) of distinct compensatory strategies. KL successfully relied on an oral (letter by letter) spelling strategy when reading vertically presented words or naming letters-in-strings, whereas he had no ability to engage compensatory mental rotation processes for tilted text. Thus, the observed impact of non-canonical presentations was moderated by the success or failure of alternative compensatory strategies. Conclusions: Importantly, our results suggest that an attentional ‘dyslexia-like’ profile can be unmasked in PCA under sufficiently taxing visual-attentional conditions. This approach may prove useful in clinical assessment, highlighting subtle reading impairments that conventional testing might overlook. Full article

Image super-resolution (SR) techniques can significantly enhance visual quality and information density. However, existing methods often rely on large amounts of paired low- and high-resolution (LR-HR) data, which limits their generalization and robustness when faced with data scarcity, distribution inconsistencies, and missing high-frequency details. To tackle the challenges of image reconstruction in data-scarce scenarios, this paper proposes a semi-supervised learning-driven multi-fidelity fusion (SSLMF) method, which integrates multi-fidelity data fusion (MFDF) and semi-supervised learning (SSL) to reduce reliance on high-fidelity data. More specifically, (1) an MFDF strategy is employed to leverage low-fidelity data for global structural constraints, enhancing information compensation; (2) an SSL mechanism is introduced to reduce data dependence by using only a small amount of labeled HR samples along with a large quantity of unlabeled multi-fidelity data. This framework significantly improves data efficiency and reconstruction quality. We first validate the reconstruction accuracy of SSLMF on benchmark functions and then apply it to image reconstruction tasks. The results demonstrate that SSLMF can effectively model both linear and nonlinear relationships among multi-fidelity data, maintaining high performance even with limited high-fidelity samples. Finally, its cross-disciplinary potential is illustrated through an audio restoration case study, offering a novel solution for efficient image reconstruction, especially in data-scarce scenarios where high-fidelity samples are limited. Full article

Citrus diseases severely impact fruit yield and quality. To facilitate in-field, non-destructive, and rapid detection of citrus leaf physiological and pathological conditions, this study proposes a classification method for citrus leaf physiological and pathological statuses that integrates visible/near-infrared multispectral technology with deep learning. First, a handheld spectrometer was employed to acquire spectral images of five sample categories—Healthy, Huanglongbing, Yellow Vein Disease, Magnesium Deficiency and Manganese Deficiency. Mean spectral data were extracted from regions of interest within the 350–2500 nm wavelength range, and various preprocessing techniques were evaluated. The Standard Normal Variate (SNV) transformation, which demonstrated optimal performance, was selected for data preprocessing. Next, we innovatively introduced an adaptive spectral positional encoding mechanism into the Transformer framework. A lightweight, learnable network dynamically optimizes positional biases, yielding the ASTransformer (Adaptive Spectral Transformer) model, which more effectively captures complex dependencies among spectral features and identifies critical wavelength bands, thereby significantly enhancing the model’s adaptive representation of discriminative bands. Finally, the preprocessed spectra were fed into three deep learning architectures (1D-CNN, 1D-ResNet, and ASTransformer) for comparative evaluation. The results indicate that ASTransformer achieves the best classification performance: an overall accuracy of 97.7%, underscoring its excellent global classification capability; a Macro Average of 97.5%, reflecting balanced performance across categories; a Weighted Average of 97.8%, indicating superior performance in classes with larger sample sizes; an average precision of 97.5%, demonstrating high predictive accuracy; an average recall of 97.7%, showing effective detection of most affected samples; and an average F1-score of 97.6%, confirming a well-balanced trade-off between precision and recall. Furthermore, interpretability analysis via Integrated Gradients quantitatively assesses the contribution of each wavelength to the classification decisions. These findings validate the feasibility of combining a handheld spectrometer with the ASTransformer model for effective citrus leaf physiological and pathological detection, enabling efficient classification and feature visualization, and offer a valuable reference for disease detection of physiological and pathological conditions in other fruit crops. Full article

Unmanned Surface Vehicles (USVs) play a critical role in maritime monitoring, environmental protection, and emergency response, necessitating accurate scene understanding in complex aquatic environments. Conventional semantic segmentation methods often fail to capture global context and lack physical boundary consistency, limiting real-world performance. This paper proposes USV-Seg, a unified segmentation framework integrating a vision-language model, the Segment Anything Model (SAM), DINOv2-based visual features, and a physically constrained refinement module. We design a task-specific <Describe> Token to enable fine-grained semantic reasoning of navigation scenes, considering USV-to-shore distance, landform complexity, and water surface texture. A mask selection algorithm based on multi-layer Intersection-over-Prediction (IoP) heads improves segmentation precision across sky, water, and obstacle regions. A boundary-aware correction module refines outputs using estimated sky-water and land-water boundaries, enhancing robustness and realism. Unlike prior works that simply apply vision-language or geometric post-processing in isolation, USV-Seg integrates structured scene reasoning and scene-aware boundary constraints into a unified and physically consistent framework. Experiments on a real-world USV dataset demonstrate that USV-Seg outperforms state-of-the-art methods, achieving 96.30% mIoU in challenging near-shore scenarios. Full article

This exploratory study investigates the usability of performance metrics for generative adversarial network (GAN)-based models for speech-driven facial animation. These models focus on the transfer of speech information from an audio file to a still image to generate talking-head videos in a small-scale “everyday usage” setting. Two models, LipGAN and a custom implementation of a Wasserstein GAN with gradient penalty (L1WGAN-GP), are examined for their visual performance and scoring according to commonly used metrics: Quantitative comparisons using FID, SSIM, and PSNR metrics on the GRIDTest dataset show mixed results, and metrics fail to capture local artifacts crucial for lip synchronization, pointing to limitations in their applicability for video animation tasks. The study points towards the inadequacy of current quantitative measures and emphasizes the continued necessity of human qualitative assessment for evaluating talking-head video quality. Full article

Aberrant internal carotid arteries (ICA) are congenital vascular anomalies that occur from involution of the cervical portion of the ICA, which leads to enlargement of the normally small collateral inferior tympanic and caroticotympanic arteries. The inferior tympanic artery is a branch of the external carotid artery, usually the ascending pharyngeal artery, which extends through the inferior tympanic canaliculus (ITC), a small foramen located along the cochlea promontory. Aberrant ICAs can also be associated with a persistent stapedial artery (PSA), which is an abnormal vessel that arises from the petrous ICA and passes through the obturator foramen of the stapes. An aberrant ICA is a very important anomaly to recognize on imaging. Accurately describing its presence is important to help prevent iatrogenic injury during intervention. It is also important to distinguish an aberrant ICA from a lateralized ICA. The improvement of spatial resolution with photon counting detector (PCD)-CT has been proven to provide higher performance in detection of sub-centimeter vascular lesions compared to conventional energy-integrated detector (EID)-CT. PCD-CT also provides superior visualization of small skull-based foramina such as the inferior tympanic canaliculus, which can aid in more accurately characterizing an aberrant ICA (variant course without ITC involvement). Full article

Objectives: The objectives of this study are to investigate the therapeutic targets and mechanisms of proanthocyanidins in alleviating fluoride-induced liver injury through network pharmacology and animal experimental validation and to explore the medicinal value of grape seed proanthocyanidins. Methods: Potential targets of proanthocyanidins were predicted using databases such as PubChem, SwissTargetPrediction, and GeneCards, and disease-related targets of fluoride-induced liver injury were retrieved to identify common targets between proanthocyanidins and fluoride-induced liver injury. The STRING database was utilized to construct a protein–protein interaction network, and key targets were analyzed for network topology using Cytoscape software. GO and KEGG enrichment analyses were performed on core target genes to explore the potential molecular mechanisms by which proanthocyanidins alleviate fluoride-induced liver injury. The Genes-miRNA interaction network was generated using Networkanalyst, and the molecular docking results between active components and key targets were validated using the CB-Dock2 visualization tool. In the academic context, a rat model of chronic fluoride poisoning was successfully established by means of intragastric administration of sodium fluoride. The protein expression levels of p-mTOR, p-p70s6, p62, LC3-II, and PARP1 in rat liver tissues were detected via Western blot analysis. Results: Network pharmacological analysis successfully identified 96 key genes, through which proanthocyanidins mitigate fluoride-induced liver injury. KEGG enrichment analysis predicted that proanthocyanidins mainly exert their therapeutic effects through the mTOR signaling pathway. The molecular docking results further demonstrated strong binding affinities between proanthocyanidins and key targets, including mTOR and PARP1. The in vivo experimental results indicate that, compared with the control group, the protein expression levels of p-mTOR, p-p70s6k, and p62 in the liver tissues of rats exposed to sodium fluoride significantly increase. Conversely, the protein expression levels of LC3-II and PARP1 significantly decrease (p < 0.05). The outcome of liver intervention with proanthocyanidins is exactly the opposite. Conclusions: Proanthocyanidins can effectively alleviate fluoride-induced liver injury, potentially by regulating the mTOR signaling pathway, autophagy, and apoptosis mechanisms. This study provides valuable insights into the protective effects of proanthocyanidins against fluoride-induced hepatic damage and offers a theoretical basis for further research in this field. Full article

To improve fracturing support efficiency of terrestrial shale oil reservoirs with uneven proppant placement, this study used complex mesh flat-plate simulations and ANSYS FLUENT (2020) simulations to test four sand addition processes. Proppants were 70/140 mesh quartz sand with a density of 2650 kg/m³ and 40/70 mesh ceramic particles with a density of 2000 kg/m³, and the carrier was hydroxypropyl guar gum fracturing fluid with a viscosity of 4.46–13.4 mPa·s at 25 °C. Alternating sand addition performed best: sand-laying efficiency reached 52 percent, 10 percentage points higher than continuous sand addition and 12 percentage points higher than mixed sand addition; sand embankment void area was 1400 cm², 18.3 percent lower than continuous sand addition; proppant entry into secondary cracks increased 23.8 percent compared with reverse sand addition; at branch crack Position 2, 1.3 m from the inlet and at a 90-degree angle, its equilibrium height was 210 mm and paving rate 0.131. This study fills gaps of no systematic multi-process comparison and insufficient quantification of crack geometry–sand parameter coupling in existing research; its novelty lies in the unified visualization comparison of four processes, revealing geometry–parameter coupling and integrating experiment simulation; the optimal scheme also improves fracture support efficiency 21.5 percent compared with conventional continuous sand addition. Full article

In order to improve the localization accuracy and model interpretability of single-phase ground fault sections in distribution networks, a knowledge-integrated and data-driven fault localization model is proposed. The model transforms the transient zero-sequence currents into Markov Transition Field (MTF) images based on transfer entropy, and improves the two-channel feature expression with both causal and temporal structures. On this basis, a knowledge guidance mechanism based on a physical mechanism is introduced to focus on the waveform backpropagation characteristics of upstream and downstream nodes of the fault through the feature attention module, and a similarity weighting strategy is constructed by integrating the Hausdorff distance in the all-connectivity layer in order to enhance the model’s capability of discriminating between the key segments. The dataset is constructed in an improved IEEE 14-node simulation system, and the effectiveness of the proposed method is verified by t-SNE feature visualization, comparison experiments with different parameters, misclassification correction analysis, and anti-noise performance evaluation. For misclassified sample datasets, this method achieves an accuracy rate of 99.53%, indicating that it outperforms traditional convolutional neural network models in terms of fault section localization accuracy, generalization capability, and noise robustness. Research shows that the deep integration of knowledge and data can significantly enhance the model’s discriminative ability and engineering practicality, providing new insights for the construction of intelligent power systems with explainability. Full article

This study investigates reading performances and eye movements in individuals with eccentric fixation due to age-related macular degeneration (AMD). Overall, 17 individuals with bilateral AMD (7 males; mean age 77.47 ± 5.96 years) and 17 controls (10 males; mean age 72.18 ± 5.98 years) were assessed for reading visual acuity (VA), reading speed (Minnesota low vision reading chart in Slovene, MNREAD-SI), and near contrast sensitivity (Pelli-Robson). Microperimetry (NIDEK MP-3) was used to evaluate preferential retinal locus (PRL) location and fixation stability. Eye movements were recorded with Tobii Pro-glasses 2 and analyzed for reading duration, saccade amplitude, peak velocity, number of saccades, saccade duration, and fixation duration. Individuals with AMD exhibited significantly reduced reading indices (worse reading VA (p < 0.001), slower reading (p < 0.001), and lower near contrast sensitivity (p < 0.001)). Eye movement analysis revealed prolonged reading duration, longer fixation duration, and an increased number of saccades in individuals with AMD per paragraph. The number of saccades per paragraph was significantly correlated with all measured reading indices. These findings provide insights into reading adaptations in AMD. Simultaneously, the proposed approach in analyzing eye movements puts forward eye trackers as a prospective diagnostic tool in ophthalmology. Full article

In air–ground cooperative systems, identifying the identities of unmanned ground vehicles (UGVs) from an unmanned aerial vehicle (UAV) perspective is a critical step for downstream tasks. Traditional approaches involving attaching markers, like AprilTags on UGVs, fail under low-resolution or occlusion conditions, and the visually identical UGVs are hard to distinguish through similar visual features. This paper proposes a markerless method that associates UGV onboard sensor data with UAV visual detections to achieve identification. Our approach employs a Dempster–Shafer fused methodology integrating two proposed complementary association techniques: a projection-based method exploiting sequential motion patterns through reprojection error validation, and a topology-based method constructing distinctive topology using positional and orientation data. The association process is further integrated into a multi-object tracking framework to reduce ID switches during occlusions. Experiments demonstrate that under low-noise conditions, the projection-based method and the topology-based method achieves association precision at 89.5% and 87.6% respectively, which is superior to the previous methods. The fused approach enables robust association at 79.9% precision under high noise conditions, nearly 10% higher than original performance. Under false detection scenarios, our method achieves effective false-positive exclusion, and the integrated tracking process effectively mitigates occlusion-induced ID switches. Full article

Background/Objectives: There is increasing interest within cognitive neuro-surgery in preserving domains not traditionally assessed during awake surgery. The study aims at proposing a specific protocol to assist surgical resection in right temporal areas. Patients were not evaluated during direct cortical stimulation; instead, assessments occurred during the resection itself. The real-time neuropsychological testing (RTNT) protocol employed tasks evaluating visuospatial and social cognition, administered repeatedly throughout the resection using varied items. Methods: A consecutive series of 24 patients (median age 44) performed RTNT. The aim of RTNT is to maintain high accuracy through resection. Lesions in the right temporal cortex and the subcortical white matter beneath can cause deficits; accordingly, not all of our patients had pre-surgery performance within the normal range. In this case, the aim of RTNT is to maintain the not perfect pre-surgery level. Results: We found a statistically significant between-tasks difference in the patients’ median values (across RTNT runs), in their minimum score reached during resection, and in the delta between performance at the last vs. the first RTNT run. The tasks that varied belonged to visual–spatial attention (landmark task), face processing (recognition of famous faces), and social cognition (theory of mind). The outcome was measured by pre- vs. post-surgery neuropsychological score comparison. The number of patients scoring below the normal range did not significantly differ between post- vs. pre-intervention. Conclusions: Results demonstrated the feasibility of implementing a continuous monitoring protocol during the resection phase, and the potential of the selected tasks to assess visuospatial and social functions associated with the non-dominant (right) hemisphere. Full article