Search Results (10,971)

Plant disease segmentation has achieved significant progress with the help of artificial intelligence. However, deploying high-accuracy segmentation models in resource-limited settings faces three key challenges, as follows: (A) Traditional dense attention mechanisms incur quadratic computational complexity growth (O($n^{2}d$)), rendering them ill-suited for low-power hardware. (B) Naturally sparse spatial distributions and large-scale variations in the lesions on leaves necessitate models that concurrently capture long-range dependencies and local details. (C) Complex backgrounds and variable lighting in field images often induce segmentation errors. To address these challenges, we propose Sparse-MoE-SAM, an efficient framework based on an enhanced Segment Anything Model (SAM). This deep learning framework integrates sparse attention mechanisms with a two-stage mixture of experts (MoE) decoder. The sparse attention dynamically activates key channels aligned with lesion sparsity patterns, reducing self-attention complexity while preserving long-range context. Stage 1 of the MoE decoder performs coarse-grained boundary localization; Stage 2 achieves fine-grained segmentation by leveraging specialized experts within the MoE, significantly enhancing edge discrimination accuracy. The expert repository—comprising standard convolutions, dilated convolutions, and depthwise separable convolutions—dynamically routes features through optimized processing paths based on input texture and lesion morphology. This enables robust segmentation across diverse leaf textures and plant developmental stages. Further, we design a sparse attention-enhanced Atrous Spatial Pyramid Pooling (ASPP) module to capture multi-scale contexts for both extensive lesions and small spots. Evaluations on three heterogeneous datasets (PlantVillage Extended, CVPPP, and our self-collected field images) show that Sparse-MoE-SAM achieves a mean Intersection-over-Union (mIoU) of 94.2%—surpassing standard SAM by 2.5 percentage points—while reducing computational costs by 23.7% compared to the original SAM baseline. The model also demonstrates balanced performance across disease classes and enhanced hardware compatibility. Our work validates that integrating sparse attention with MoE mechanisms sustains accuracy while drastically lowering computational demands, enabling the scalable deployment of plant disease segmentation models on mobile and edge devices. Full article

Background/Objectives: This study assessed how to best prepare pediatric and adolescent cancer patients for amputation and support them afterward. Methods: This pilot qualitative study explored pre- and post-amputation experiences from the perspectives of nine pediatric and adolescent survivors who underwent amputation. Hour-long audio-recorded semi-structured interviews were transcribed and analyzed using the Sort and Sift, Think and Shift qualitative approach. Results: Participants described the informational supports they received before surgery, including guidance on what to expect, contact with amputation-related organizations, and exposure to tangible tools, such as a physical model of a knee joint. Emotional support from fellow amputees and healthcare providers, particularly surgeons, was also found to be meaningful. Individuals also identified unmet needs and gaps in emotional care. These included clearer guidance on post-surgical adaptations (e.g., basic self-care and navigating physical limitations) and the need for information tailored to their learning styles. Many emphasized the importance of improved pain management resources, expanded access to mental health services for both them and their families, and support in adjusting to changes in body image and social relationships. Participants also shared advice for future patients, recommending strategies such as personalizing hospital rooms, connecting with other amputees through social media, and using art to process their experience and say goodbye to the lost limb. Conclusions: Interviews with nine cancer survivors provide guidance for improving holistic, patient-centered care throughout the amputation process. Informational and emotional support should be tailored to an individual’s learning style and specific needs, in addition to their age at the time of surgery. Full article

Fault diagnosis under variable operating conditions remains challenging due to the limited adaptability of traditional methods. This paper proposes a transfer learning-based approach for bearing fault diagnosis across different rotational speeds, addressing the critical need for reliable detection in changing industrial environments. The method trains a diagnostic model on labeled source-domain data and transfers them to unlabeled target domains through a two-stage adaptation strategy. First, only the source-domain data are labeled to reflect real-world scenarios where target-domain labels are unavailable. The model architecture combines a convolutional neural network (CNN) for feature extraction with a self-attention mechanism for classification. During source-domain training, the feature extractor parameters are frozen to focus on classifier optimization. When transferring to target domains, the classifier parameters are frozen instead, allowing the feature extractor to adapt to new speed conditions. Experimental validation on the Case Western Reserve University bearing dataset (CWRU), Jiangnan University bearing dataset (JNU), and Southeast University gear and bearing dataset (SEU) demonstrates the method’s effectiveness, achieving accuracies of 99.95%, 99.99%, and 100%, respectively. The proposed method achieves significant model size reduction compared to conventional TL approaches (e.g., DANN and CDAN), with reductions of up to 91.97% and 64%, respectively. Furthermore, we observed a maximum reduction of 61.86% in FLOPs consumption. The results show significant improvement over conventional approaches in maintaining diagnostic performance across varying operational conditions. This study provides a practical solution for industrial applications where equipment operates under non-stationary speeds, offering both computational efficiency and reliable fault detection capabilities. Full article

The growing use of unmanned aerial vehicles (UAVs) for real-time video surveillance in smart city and smart region infrastructures requires reliable and delay-aware data transmission models. In urban environments, UAV communication links are subject to stochastic variability, leading to jitter, packet loss, and unstable video delivery. This paper presents a novel approach based on the Graphical Evaluation and Review Technique (GERT) for modeling the transmission of video frames from UAVs over uncertain network paths with probabilistic feedback loops and lognormally distributed delays. The proposed model enables both analytical and numerical evaluation of key Quality-of-Service (QoS) metrics, including mean transmission time and jitter, under varying levels of channel variability. Additionally, the structure of the GERT-based framework allows integration with artificial intelligence mechanisms, particularly for adaptive routing and delay prediction in urban conditions. Spectral analysis of the system’s characteristic function is also performed to identify instability zones and guide buffer design. The results demonstrate that the approach supports flexible, parameterized modeling of UAV video transmission and can be extended to intelligent, learning-based control strategies in complex smart city environments. This makes it suitable for a wide range of applications, including traffic monitoring, infrastructure inspection, and emergency response. Beyond QoS optimization, the framework explicitly accommodates security and privacy preserving operations (e.g., encryption, authentication, on-board redaction), enabling secure UAV video transmission in urban networks. Full article

Background: Colorectal cancer (CRC) remains one of the leading causes of cancer-related mortality worldwide, emphasizing the urgent need for early, non-invasive, and accessible diagnostic tools. This study aimed to evaluate the effectiveness of a microelectromechanical systems (MEMS)-based electronic nose (E-nose) in combination with gas chromatography–mass spectrometry (GC-MS) for CRC detection through sweat volatile organic compounds (VOCs). Methods: A total of 136 sweat samples were collected from 68 volunteer participants. Samples were processed using solid-phase microextraction (SPME) and analyzed by GC-MS, while a custom-designed E-nose system comprising 14 gas sensors captured real-time VOC profiles. Data were analyzed using multivariate statistical techniques, including PCA and PLS-DA, and classified with machine learning algorithms (LDA, LR, SVM, k-NN). Results: GC-MS analysis revealed statistically significant differences between CRC patients and healthy controls (COs). Cross-validation showed that the highest classification accuracy for GC-MS data was 81% with the k-NN classifier, whereas E-nose data achieved up to 97% accuracy using the LDA classifier. Conclusions: Sweat volatilome analysis, supported by advanced data processing and complementary use of E-nose technology and GC-MS, demonstrates strong potential as a reliable, non-invasive approach for early CRC detection. Full article

Although rectal temperature is reliable, its measurement requires manual handling and causes stress to animals. IRT provides a non-contact alternative but often ignores bilateral eye temperature differences. This study presents an E-S-YOLO11n model for the automated detection of the binocular regions of sheep, which achieves remarkable performance with a precision of 98.2%, recall of 98.5%, mAP@0.5 of 99.40%, F₁ score of 98.35%, FPS of 322.58 frame/s, parameters of 7.27 M, model size of 3.97 MB, and GFLOPs of 1.38. Right and left eye temperatures exhibit a strong correlation (r = 0.8076, p < 0.0001), However, the eye temperatures show only very weak correlation with rectal temperature (right eye: r = 0.0852; left eye: r = −0.0359), and neither figure reaches statistical significance. Rectal temperature is 7.37% and 7.69% higher than the right and left eye temperatures, respectively. Additionally, the right eye temperature is slightly higher than the left eye (p < 0.01). The study demonstrates the feasibility of combining IRT and deep learning for non-invasive eye temperature monitoring, although environmental factors may limit it as a proxy for rectal temperature. These results support the development of efficient thermal monitoring tools for precision animal husbandry. Full article

Timely, objective pain recognition in dairy cattle is essential for welfare assurance, productivity, and ethical husbandry yet remains elusive because evolutionary pressure renders bovine distress signals brief and inconspicuous. Without verbal self-reporting, cows suppress overt cues, so automated vision is indispensable for on-farm triage. Although earlier systems tracked whole-body posture or static grimace scales, frame-level detection of facial micro-expressions has not been explored fully in livestock. We translate micro-expression analytics from automotive driver monitoring to the barn, linking modern computer vision with veterinary ethology. Our two-stage pipeline first detects faces and 30 landmarks using a custom You Only Look Once (YOLO) version 8-Pose network, achieving a 96.9% mean average precision ($mAP$) at an Intersection over the Union (IoU) threshold of 0.50 for detection and 83.8% Object Keypoint Similarity (OKS) for keypoint placement. Cropped eye, ear, and muzzle patches are encoded using a pretrained MobileNetV2, generating 3840-dimensional descriptors that capture millisecond muscle twitches. Sequences of five consecutive frames are fed into a 128-unit Long Short-Term Memory (LSTM) classifier that outputs pain probabilities. On a held-out validation set of 1700 frames, the system records 99.65% accuracy and an F1-score of 0.997, with only three false positives and three false negatives. Tested on 14 unseen barn videos, it attains 64.3% clip-level accuracy (i.e., overall accuracy for the whole video clip) and 83% precision for the pain class, using a hybrid aggregation rule that combines a 30% mean probability threshold with micro-burst counting to temper false alarms. As an early exploration from our proof-of-concept study on a subset of our custom dairy farm datasets, these results show that micro-expression mining can deliver scalable, non-invasive pain surveillance across variations in illumination, camera angle, background, and individual morphology. Future work will explore attention-based temporal pooling, curriculum learning for variable window lengths, domain-adaptive fine-tuning, and multimodal fusion with accelerometry on the complete datasets to elevate the performance toward clinical deployment. Full article

Scene text understanding, serving as a cornerstone technology for autonomous navigation, document digitization, and accessibility tools, has witnessed a paradigm shift from traditional methods relying on handcrafted features and multi-stage processing pipelines to contemporary deep learning frameworks capable of learning hierarchical representations directly from raw image inputs. This survey distinctly categorizes modern scene text recognition (STR) methodologies into three principal paradigms: two-stage detection frameworks that employ region proposal networks for precise text localization, single-stage detectors designed to optimize computational efficiency, and specialized architectures tailored to handle arbitrarily shaped text through geometric-aware modeling techniques. Concurrently, an in-depth analysis of text recognition paradigms elucidates the evolutionary trajectory from connectionist temporal classification (CTC) and sequence-to-sequence models to transformer-based architectures, which excel in contextual modeling and demonstrate superior performance. In contrast to prior surveys, this work uniquely emphasizes several key differences and contributions. Firstly, it provides a comprehensive and systematic taxonomy of STR methods, explicitly highlighting the trade-offs between detection accuracy, computational efficiency, and geometric adaptability across different paradigms. Secondly, it delves into the nuances of text recognition, illustrating how transformer-based models have revolutionized the field by capturing long-range dependencies and contextual information, thereby addressing challenges in recognizing complex text layouts and multilingual scripts. Furthermore, the survey pioneers the exploration of critical research frontiers, such as multilingual text adaptation, enhancing model robustness against environmental variations (e.g., lighting conditions, occlusions), and devising data-efficient learning strategies to mitigate the dependency on large-scale annotated datasets. By synthesizing insights from technical advancements across 28 benchmark datasets and standardized evaluation protocols, this study offers researchers a holistic perspective on the current state-of-the-art, persistent challenges, and promising avenues for future research, with the ultimate goal of achieving human-level scene text comprehension. Full article

The rapid expansion of bike-sharing systems has introduced significant management challenges related to spatial-temporal demand fluctuations and inefficient e-fence capacity allocation. This study proposes a Spatio-Temporal Graph Attention Transformer Network (STGATN), a novel hybrid deep learning framework for dynamic demand forecasting in bike-sharing e-fence systems. The model integrates Graph Convolutional Networks to capture complex spatial dependencies among urban functional zones, Bi-LSTM networks to model temporal patterns with periodic variations, and attention mechanisms to dynamically incorporate weather impacts. By constructing a city-level graph based on POI-derived e-fences and implementing multi-source feature fusion through Transformer architecture, the STGATN effectively addresses the limitations of static capacity allocation strategies. The experimental results from Shenzhen’s Nanshan District demonstrate the performance, with the STGATN model achieving an overall Mean Absolute Error (MAE) of 0.0992 and a Coefficient of Determination (R²) of 0.8426. This significantly outperforms baseline models such as LSTM (R²: 0.6215) and a GCN (R²: 0.5488). Ablation studies confirm the model’s key components are critical; removing the GCN module decreased R² by 12 percentage points to 0.7411, while removing the weather attention mechanism reduced R² by nearly 5 percentage points to 0.8034. The framework provides a scientific basis for dynamic e-fence capacity management, advancing spatio-temporal prediction methodologies for sustainable transportation. Full article

The accurate prediction of the pavement structural modulus is crucial for maintenance planning and life-cycle assessment. While recent deep learning models have improved predictive accuracy using Falling Weight Deflectometer data, challenges remain in effectively structuring time-series inputs and ensuring robustness against noise measurement. This paper presents an integrated framework that combines systematic time-step modeling with perturbation-based robustness evaluation. Five distinct input sequencing strategies (Plan A through Plan E) were developed to investigate the impact of temporal structure on model performance. A hybrid Wide & Deep ResRNN architecture incorporating SimpleRNN, GRU, and LSTM components was designed to jointly predict four-layer moduli. To simulate real-world sensor uncertainty, Gaussian noise with ±3% variance was injected into inputs, allowing the Monte-Carlo-style estimation of confidence intervals. Experimental results revealed that time-step design plays a critical role in both prediction accuracy and robustness, with Plan D consistently achieving the best balance between accuracy and stability. These findings offer a practical and generalizable approach for deploying deep sequence models in pavement modulus prediction tasks, particularly under uncertain field conditions. Full article

The escalating global threat of antimicrobial resistance (AMR) underscores the urgent need for innovative therapeutics. Bacteriophages (phages), natural bacterial predators, offer promising solutions, especially when harnessed through advances in artificial intelligence (AI). This review explores how AI-driven innovations are transforming phage biology, with an emphasis on three pivotal areas: (1) AI-enhanced structural prediction (e.g., AlphaFold); (2) deep learning functional annotation; (3) bioengineering strategies, including CRISPR-Cas. We further discuss applications extending to medical therapy, biosensing, agricultural biocontrol, and environmental remediation. Despite progress, critical challenges persist—including high false-positive rates, difficulties in modeling disordered protein regions, and biosafety concerns remain. Overcoming these requires experimental validation, robust computational frameworks, and global regulatory oversight. AI integration in phage research is accelerating the development of next-generation therapeutics to combat AMR and advance engineered living therapeutics. Full article

Multimodal imaging techniques have the potential to enhance the interpretation of histology by offering additional molecular and structural information beyond that accessible through hematoxylin and eosin (H&E) staining alone. Here, we present a quantitative approach for comparing the information content of different image modalities, such as H&E and multimodal imaging. We used a combination of deep learning and radiomics-based feature extraction with different information markers, implemented in Python 3.12, to compare the information content of the H&E stain, multimodal imaging, and the combined dataset. We also compared the information content of individual channels in the multimodal image and of different Coherent Anti-Stokes Raman Scattering (CARS) microscopy spectral channels. The quantitative measurements of information that we utilized were Shannon entropy, inverse area under the curve (1-AUC), the number of principal components describing 95% of the variance (PC95), and inverse power law fitting. For example, the combined dataset achieved an entropy value of 0.5740, compared to 0.5310 for H&E and 0.5385 for the multimodal dataset using MobileNetV2 features. The number of principal components required to explain 95 percent of the variance was also highest for the combined dataset, with 62 components, compared to 33 for H&E and 47 for the multimodal dataset. These measurements consistently showed that the combined datasets provide more information. These observations highlight the potential of multimodal combinations to enhance image-based analyses and provide a reproducible framework for comparing imaging approaches in digital pathology and biomedical image analysis. Full article

The Third Pole region, particularly the Hindu–Kush–Himalaya (HKH), is highly prone to lightning, causing thousands of fatalities annually. Skillful prediction and timely communication are essential for mitigating lightning-related losses in such observationally data-sparse regions. Therefore, this study evaluates kilometer-scale ICON-CLM-simulated atmospheric variables using six machine learning (ML) models to detect lightning activity over the Third Pole. Results from the ensemble boosting ML models show that ICON-CLM simulated variables such as relative humidity (RH), vorticity (vor), 2m temperature (t_2m), and surface pressure (sfc_pres) among a total of 25 variables allow better spatial and temporal prediction of lightning activities, achieving a Probability of Detection (POD) of ∼0.65. The Lightning Potential Index (LPI) and the product of convective available potential energy (CAPE) and precipitation (prec_con), referred to as CP (i.e., CP = CAPE × precipitation), serve as key physics aware predictors, maintaining a high Probability of Detection (POD) of ∼0.62 with a 1–2 h lead time. Sensitivity analyses additionally using climatological lightning data showed that while ML models maintain comparable accuracy and POD, climatology primarily supports broad spatial patterns rather than fine-scale prediction improvements. As LPI and CP reflect cloud microphysics and atmospheric stability, their inclusion, along with spatiotemporal averaging and climatology, offers slightly lower, yet comparable, predictive skill to that achieved by aggregating 25 atmospheric predictors. Model evaluation using the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) highlights XGBoost as the best-performing diagnostic classification (yes/no lightning) model across all six ML tested configurations. Full article

Fault diagnosis for embedded sensors in asphalt pavement faces significant challenges, including the scarcity of real-world fault data and the difficulty in identifying compound faults, which severely compromises the reliability of monitoring data. To address these issues, this study proposes an intelligent diagnostic framework that integrates a Discrete Wavelet Transform (DWT) with a staged, attention-based Long Short-Term Memory (LSTM) network. First, various fault modes were systematically defined, including short-term (i.e., bias, gain, and detachment), long-term (i.e., drift), and their compound forms. A fine-grained fault injection and labeling strategy was then developed to generate a comprehensive dataset. Second, a novel diagnostic model was designed based on a “Decomposition-Focus-Fusion” architecture. In this architecture, the DWT is employed to extract multi-scale features, and independent sub-models—a Bidirectional LSTM (Bi-LSTM) and a stacked LSTM—are subsequently utilized to specialize in learning short-term and long-term fault characteristics, respectively. Finally, an attention network intelligently weights and fuses the outputs from these sub-models to achieve precise classification of eight distinct sensor operational states. Validated through rigorous 5-fold cross-validation, experimental results demonstrate that the proposed framework achieves a mean diagnostic accuracy of 98.89% (±0.0040) on the comprehensive test set, significantly outperforming baseline models such as SVM, KNN, and a unified LSTM. A comprehensive ablation study confirmed that each component of the “Decomposition-Focus-Fusion” architecture—DWT features, staged training, and the attention mechanism—makes an indispensable contribution to the model’s superior performance. The model successfully distinguishes between “drift” and “normal” states—which severely confuse the baseline models—and accurately identifies various complex compound faults. Furthermore, simulated online diagnostic tests confirmed the framework’s rapid response capability to dynamic faults and its computational efficiency, meeting the demands of real-time monitoring. This study offers a precise and robust solution for the fault diagnosis of embedded sensors in asphalt pavement. Full article

Background: Differentiating acute kidney injury (AKI) from chronic kidney disease (CKD) in children remains a critical unmet need due to the limitations of current clinical and biochemical markers. Conventional ultrasound lacks the sensitivity to discern subtle parenchymal alterations. This study explores the application of ultrasound radiomics—a novel, non-invasive, and quantitative image analysis method—for distinguishing AKI from CKD in pediatric patients. Methods: In this retrospective cross-sectional pilot study, kidney ultrasound images were obtained from 31 pediatric subjects: 8 with oliguric AKI, 14 with CKD, and 9 healthy controls. Renal parenchyma was manually segmented, and 124 advanced texture features were extracted using the open-source ©PyFeats. Features encompassed multiple categories (e.g., GLCM, GLSZM, WP). Statistical comparisons evaluated intergroup differences. Principal Component Analysis identified the top 10 most informative features, which were used to train supervised machine learning models. Model performance used five-fold cross-validation. Results: Radiomic analysis revealed significant intergroup differences (p < 0.05). CKD cases exhibited increased echogenicity and heterogeneity, particularly in GLCM and GLSZM features, consistent with chronic fibrosis. AKI cases displayed more homogeneous texture, likely reflecting edema or acute inflammation. While echogenicity separated diseased from healthy kidneys, it lacked specificity between AKI and CKD. Among ML models, XGBoost achieved the highest macro-averaged F1 score (0.90), followed closely by SVM and Random Forest, demonstrating strong classification performance. Conclusions: Radiomics-based texture analysis of grayscale ultrasound images effectively differentiated AKI from CKD in this pilot study, offering a promising, non-invasive imaging biomarker for pediatric kidney disease. These preliminary findings justify prospective validation in larger, multicenter cohorts. Full article