Search Results (30,999)

Deep learning models based on supervised learning rely heavily on large annotated datasets and particularly in the context of medical image segmentation, the requirement for pixel-level annotations makes the labeling process labor-intensive, time-consuming and expensive. To overcome these limitations, self-supervised learning (SSL) has emerged as a promising alternative that learns generalizable representations from unlabeled data; however, existing SSL frameworks often employ highly parameterized encoders that are computationally expensive and may lack robustness in label-scarce settings. In this work, we propose a scattering-based SSL framework that integrates Wavelet Scattering Networks (WSNs) and Parametric Scattering Networks (PSNs) into a Bootstrap Your Own Latent (BYOL) pretraining pipeline. By replacing the initial stages of the BYOL encoder with fixed or learnable scattering-based front-ends, the proposed method reduces the number of learnable parameters while embedding translation-invariant and small deformation-stable representations into the SSL pipeline. The pretrained encoders are transferred to a U-Net and fine-tuned for cardiac image segmentation on two datasets with different imaging modalities, namely, cardiac cine MRI (ACDC) and cardiac CT (CHD), under varying amounts of labeled data. Experimental results show that scattering-based SSL pretraining consistently improves segmentation performance over random initialization and ImageNet pretraining in low-label regimes, with particularly pronounced gains when only a few labeled patients are available. Notably, the PSN variant achieves improvements of 4.66% and 2.11% in average Dice score over standard BYOL with only 5 and 10 labeled patients, respectively, on the ACDC dataset. These results demonstrate that integrating mathematically grounded scattering representations into SSL pipelines provides a robust and data-efficient initialization strategy for cardiac image segmentation, particularly under limited annotation and domain shift. Full article

Aquatic vegetation (AV) forms the structural and functional basis of lake ecosystems, providing irreplaceable ecological functions such as water self-purification and the sustenance of biodiversity. Under the “Yangtze River’s Great Protection Strategy”, the action of returning nets to the lake has significantly improved water-quality in the middle and lower reaches of the Yangtze River (MLRYR) basin. However, its ecological benefits for key biotic components, particularly AV communities, remain unclear. To address this knowledge gap, this study utilized Landsat and Sentinel-1 satellite imagery to analyze the dynamic evolution of enclosure aquaculture (EA) and AV in 25 lakes (>10 km²) within the MLRYR basin from 1989 to 2023. A U-Net deep learning model was employed to extract EA data (2016–2023), and a vegetation and bloom extraction algorithm was applied to map different AV groups (1989–2023). Results indicate that by 2023, 88% (22/25) of the lakes had completed EA removal. Over the 34-year period, floating/emergent aquatic vegetation (FEAV) exhibited fluctuating trends, while submerged aquatic vegetation (SAV) demonstrated a significant decline, particularly during the EA demolition phase (2016–2023), when its area sharply decreased from 804.8 km² to 247.3 km²—a reduction of 69.3%. Spatial comparative analysis further confirmed that SAV degradation was substantially more severe in EA removal areas than in EA retention areas. This study demonstrates that EA demolition, while beneficial for improving water quality, exerts significant short-term negative impacts on AV. These findings highlight the urgent need for lake governance policies to shift from single-objective management toward integrated strategies that equally prioritize water-quality improvement and ecological restoration. Future efforts should enhance targeted restoration in EA removal areas through active vegetation recovery and habitat reconstruction, thereby preventing catastrophic regime shifts to phytoplankton-dominated turbid-water states in lake ecosystems. Full article

This study aims to investigate the predictive value of pre-treatment multi-phasic contrast-enhanced computed tomography (CECT) radiomic features for treatment resistance in patients with rat sarcoma virus (RAS)-mutated colorectal liver metastases (CRLMs) receiving bevacizumab-based chemotherapy. Seventy-three samples with RAS-mutated CRLMs receiving bevacizumab-combined chemotherapy regimens were evaluated. Radiomic features were extracted from arterial phase (AP), portal venous phase (PVP), AP-PVP subtraction image, and Delta phase (DeltaP, calculated as AP-to-PVP ratio) images. Three groups of radiomics features were extracted for each phase, including peritumor, core tumor, and whole-tumor regions. For each of the four phases, a two-sided independent Mann–Whitney U test with the Bonferroni correction and K-means clustering was applied to the remnant features for each phase. Subsequently, the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm was then applied for further feature selection. Six machine learning algorithms were then used for model development and validated on the independent testing cohort. Results showed peritumoral radiomic features and features derived from Laplacian of Gaussian (LoG) filtered images were dominant in all the compared machine learning algorithms; NB models yielded the best-performing prediction (Avg. training AUC: 0.731, Avg. testing AUC: 0.717) when combining all features from different phases of CECT images. This study demonstrates that peritumoral radiomic features and LoG-filtered pre-treatment multi-phasic CECT images were more predictive of treatment response to bevacizumab-based chemotherapy in RAS-mutated CRLMs compared to core tumor features. Full article

Traffic state estimation (TSE) is a core task in intelligent transportation systems (ITSs) that seeks to infer key operational parameters—such as speed, flow, and density—from limited observational data. Existing methods often face challenges in practical deployment, including limited estimation accuracy, insufficient physical consistency, and weak generalization capability. To address these issues, this paper proposes a hybrid estimation framework that integrates multi-agent reinforcement learning (MARL) with the Lighthill–Whitham–Richards (LWR) traffic flow model. In this framework, each roadside detector is modeled as an agent that adaptively learns fundamental diagram (FD) parameters—the free-flow speed and jam density—by fusing local detector measurements with global CAV trajectory sequences via an interactive attention mechanism. The learned parameters are then passed to an LWR solver to perform sequential (rolling) prediction of traffic states across the entire road segment. We design a reward function that jointly penalizes estimation error and violations of physical constraints, enabling the agents to learn accurate and physically consistent dynamic traffic state estimates through interaction with the physics-based LWR environment. Experiments on simulated and real-world datasets demonstrate that the proposed method outperforms existing models in estimation accuracy, real-time performance, and cross-scenario generalization. It faithfully reproduces dynamic traffic phenomena, such as shockwaves and queue waves, demonstrating robustness and practical potential for deployment in complex traffic environments. Full article

This study aims to assess hand function recovery in stroke patients during the mid-to-late Brunnstrom stages and to encourage active participation in rehabilitation exercises. To this end, a deep residual network (ResNet) integrated with Focal Loss is employed for gesture recognition, achieving a Macro F1 score of 91.0% and a validation accuracy of 90.9%. Leveraging the millimetre-level precision of Leap Motion 2 hand tracking, a mapping relationship for hand skeletal joint points was established, and a static assessment gesture data set containing 502,401 frames was collected through analysis of the FMA scale. The system implements an immersive augmented reality interaction through the Unity development platform; C# algorithms were designed for real-time motion range quantification. Finally, the paper designs a rehabilitation system framework tailored for home and community environments, including system module workflows, assessment modules, and game logic. Experimental results demonstrate the technical feasibility and high accuracy of the automated system for assessment and rehabilitation training. The system is designed to support stroke patients in home and community settings, with the potential to enhance rehabilitation motivation, interactivity, and self-efficacy. This work presents an integrated research framework encompassing hand modelling and deep learning-based recognition. It offers the possibility of feasible and economical solutions for stroke survivors, laying the foundation for future clinical applications. Full article

Foreign fiber (FF, plural: FFs) contamination has been demonstrated to have a substantial impact on the quality and profitability of cotton textiles. Machine vision technology, characterized by its non-contact approach and high efficiency, has emerged as the primary solution for detecting FFs in cotton. This paper commences with a precise definition and classification of FF and a concomitant analysis of the mechanisms of contamination. Subsequently, a systematic review of global research advancements in imaging technologies and the evolution of algorithms is conducted. This paper emphasizes the use of X-ray, ultraviolet fluorescence, line laser, polarized light, infrared imaging, and hyperspectral imaging techniques for FF detection. Through a comparative analysis, it reveals the applicable scope and effectiveness of various imaging schemes. Regarding the evolution of algorithms, this paper expounds on the technical development process from traditional image processing to machine learning (ML) and deep learning (DL). The study meticulously examines the strengths and weaknesses of each algorithmic stage. In conclusion, this paper synthesizes the prevailing technical challenges confronting machine vision detection of FFs in cotton and proffers recommendations for future research directions in this domain, emphasizing multi-technology integration, algorithm optimization, and hardware innovations. Full article

Accurate numerical simulation of debris flows is essential for hazard assessment and early-warning design, yet high-fidelity solvers remain computationally expensive, especially when large ensembles must be explored under epistemic uncertainty in rheology, initial conditions, and topography. At the same time, field observations are typically sparse and heterogeneous, limiting purely data-driven approaches. In this work, we develop a deep-learning Fourier Neural Operator (FNO) as a fast, physics-consistent surrogate for one-dimensional shallow-water debris-flow simulations and demonstrate its application to the Rendinara–Morino system in central Italy. A validated finite-volume solver, equipped with HLLC and Rusanov fluxes, hydrostatic reconstruction, Voellmy-type basal friction, and robust wet–dry treatment, is used to generate a large ensemble of synthetic simulations over longitudinal profiles representative of the study area. The parameter space of bulk density, initial flow thickness, and Voellmy friction coefficients is systematically sampled, and the resulting space–time fields of flow depth and velocity form the training dataset. A two-dimensional FNO in the $(x,t)$ domain is trained to learn the full solution operator, mapping topography, rheological parameters, and initial conditions directly to $h(x,t)$ and $u(x,t)$, thereby acting as a site-specific digital twin of the numerical solver. On a held-out validation set, the surrogate achieves mean relative $L^2$ errors of about 6–$7\%$ for flow depth and 10–$15\%$ for velocity, and it generalizes to an unseen longitudinal profile with comparable accuracy. We further show that targeted reweighting of the training objective significantly improves the prediction of the velocity field without degrading depth accuracy, reducing the velocity error on the unseen profile by more than a factor of two. Finally, the FNO provides speed-ups of approximately $36\times$ with respect to the reference solver at inference time. These results demonstrate that combining physics-based synthetic data with operator-learning architectures enables the construction of accurate, computationally efficient, and site-adapted surrogates for debris-flow hazard analysis in data-scarce environments. Full article

Water quality monitoring faces increasing challenges due to accelerating industrialization and urbanization, demanding accurate, real-time, and reliable prediction technologies. This study presents a novel ensemble learning framework integrating Gaussian Process Regression, Support Vector Regression, and Random Forest algorithms for high-precision water quality pH prediction. The research utilized a comprehensive spatiotemporal dataset, comprising 11 water quality parameters from 37 monitoring stations across Georgia, USA, spanning 705 days from January 2016 to January 2018. The ensemble model employed a dynamic weight allocation strategy based on cross-validation error performance, assigning optimal weights of 34.27% to Random Forest, 33.26% to Support Vector Regression, and 32.47% to Gaussian Process Regression. The integrated approach achieved superior predictive performance, with a mean absolute error of 0.0062 and coefficient of determination of 0.8533, outperforming individual base learners across multiple evaluation metrics. Statistical significance testing using Wilcoxon signed-rank tests with a Bonferroni correction confirmed that the ensemble significantly outperforms all individual models (p < 0.001). Comparison with state-of-the-art models (LightGBM, XGBoost, TabNet) demonstrated competitive or superior ensemble performance. Comprehensive ablation experiments revealed that Random Forest removal causes the largest performance degradation (+4.43% MAE increase). Feature importance analysis revealed the dissolved oxygen maximum and conductance mean as the most influential predictors, contributing 22.1% and 17.5%, respectively. Cross-validation results demonstrated robust model stability with a mean absolute error of 0.0053 ± 0.0002, while bootstrap confidence intervals confirmed narrow uncertainty bounds of 0.0060 to 0.0066. Spatiotemporal analysis identified station-specific performance variations ranging from 0.0036 to 0.0150 MAE. High-error stations (12, 29, 33) were analyzed to distinguish characteristics, including higher pH variability and potential upstream pollution influences. An integrated software platform was developed featuring intuitive interface, real-time prediction, and comprehensive visualization tools for environmental monitoring applications. Full article

Predicting specific RNA–protein interactions remains a challenging task. Despite the existence of numerous methods, a unified approach has yet to emerge. Additional difficulties emerge from the properties of in vivo IP experiments and their systematic biases, such as the overrepresentation of highly expressed RNAs. Here, we present the PLERIO machine learning framework, which utilizes eCLIP data for a single protein to reconstruct the full spectrum of its potential interactions with the cellular transcriptome (i.e., both highly expressed and lowly expressed RNAs). In an effort to extrapolate our methodology to a multi-protein paradigm for de novo prediction of RNA–protein interactions on proteins lacking available eCLIP data, we extended our approach to 220 cellular proteins. We then demonstrate that this approach might not be well tailored to the limitations of current in vivo immunoprecipitation data, and may only be meaningful for in vitro experiments such as RNAcompete. Full article

Mitochondria are essential organelles involved in metabolism, energy production, and cell signaling. Assessing mitochondrial morphology is key to tracking cell metabolic activity and function. Quantifying these structural changes may also provide critical insights into disease pathogenesis and therapeutic responses. This work details the development and validation of a novel, quantitative image analysis pipeline for the characterization and classification of dynamic mitochondrial morphologies. Utilizing high-resolution confocal microscopy, the pipeline integrates first-order statistics (FOS) and a comprehensive suite of gray-level texture analyses, including gray level co-occurrence matrix (GLCM), gray level run length matrix (GLRLM), gray level dependence matrix (GLDM), gray level size zone matrix (GLSZM), and neighboring gray tone difference matrix (NGTDM) with machine learning approaches. The method’s efficacy in objectively differentiating key mitochondrial structures—fibers, puncta, and rods—which are critical indicators of cellular metabolic and activation states is demonstrated. Our open-source pipeline provides robust quantitative metrics for characterizing mitochondrial variation. Full article

Machine learning (ML)-powered vision for robotic inspection has accelerated with smart manufacturing, enabling automated defect detection and classification and real-time process optimization. This review provides insight into the current landscape and state-of-the-art practices in smart manufacturing quality control (QC). More than 50 studies spanning across automotive, aerospace, assembly, and general manufacturing sectors demonstrate that ML-powered vision is technically viable for robotic inspection in manufacturing. The accuracy of defect detection and classification frequently exceeds 95%, with some vision systems achieving 98–100% accuracy in controlled environments. The vision systems use predominantly self-designed convolutional neural network (CNN) architectures, YOLO variants, or traditional ML vision models. However, 77% of implementations remain at the prototype or pilot scale, revealing systematic deployment barriers. A discussion is provided to address the specifics of the vision systems and the challenges that these technologies continue to face. Finally, recommendations for future directions in ML-powered vision for robotic inspection in manufacturing are provided. Full article

The aim of this article is to develop a machine learning (ML)-based predictive model for industrial accidents in the energy sector. The dataset used in this study was obtained from the Kaggle platform and consists of summaries derived from reports of occupational incidents resulting in injuries or deaths between 2015 and 2017. A total of 4739 accident cases were included, containing information on accident date, accident summary, degree and nature of injury, affected body part, event type, human factors, and environmental factors. Six supervised machine learning models—Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Gradient Boosting Decision Trees (GBDT), and Extreme Gradient Boosting (XGBoost)—were developed and compared to identify the most suitable model for the data. Model performance was evaluated using accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (AUC), which were selected to ensure reliable prediction in safety-critical accident scenarios. The results indicate that XGBoost and GBDT achieve superior performance in predicting human and environmental risk factors. These findings demonstrate the potential of machine learning for improving safety management in the energy sector by identifying risk mechanisms, enhancing safety awareness, and providing quantitative predictions of fatal and non-fatal accident occurrences for integration into safety management systems. Full article

Recent evidence has highlighted the pivotal roles of reactive oxygen species (ROS) and the SIRT1, AMPK, and mTOR signaling pathways in aging and longevity, making them attractive targets for studies of lifespan-extending interventions. We previously demonstrated that 1,1-diethoxyethane (1,1-DEE) could interact with mitochondrial complex I (NADH–ubiquinone oxidoreductase), leading to transient mitochondrial ROS (mtROS) production and activation of the AMPK pathway. This study further examined the effects of 1,1-DEE on longevity in model organisms. Treatment with 1,1-DEE decreased senescence in endothelial cell EA.hy926. In Caenorhabditis elegans (C. elegans), 1,1-DEE induced a hormetic response and extended the lifespan, whereas its structural isoform, 1,2-diethoxyethane (1,2-DEE), showed no such effect. In rat models, administration of 1,1-DEE markedly improved survival rate, mortality risk, restricted mean survival time (RMST), and median lifespan, associated with an accelerated body weight reduction. Additionally, 1,1-DEE could also enhance learning and memory, as assessed by the Morris water maze test in rats. These findings suggest that 1,1-DEE may serve as a novel small-molecule modulator of mitochondrial function and redox signaling, with potentials for promoting anti-aging and longevity. Full article

In this study, a hybrid deep learning model that combines the VGG16 and ResNet101V2 architectures is proposed for image-based fire detection. In addition, a balanced drone guidance algorithm is developed to efficiently assign tasks to available UAVs. In the fire detection phase, the hybrid model created by combining the VGG16 and ResNet101V2 architectures has been optimized with Global Average Pooling and layer merging techniques to increase classification success. The DeepFire dataset was used throughout the training process, achieving an extremely high accuracy rate of 99.72% and 100% precision. After fire detection, a task assignment algorithm was developed to assign existing drones to fire points at minimum cost and with balanced load distribution. This algorithm performs task assignments using the Hungarian (Kuhn–Munkres) method and cost optimization, and is adapted to direct approximately equal numbers of drones to each fire when the number of fires is less than the number of drones. The developed system was tested in a Python-based simulation environment and evaluated using performance metrics such as total intervention time, energy consumption, and task balance. The results demonstrate that the proposed hybrid model provides highly accurate fire detection and that the task assignment system creates balanced and efficient intervention scenarios. Full article

Current electrical self-sensing methods for composite structural health monitoring face significant limitations. Firstly, they often require complicated electrode layouts. Secondly, accurately determining both the location and amplitude of external loads remains a significant challenge. In this study, a deep learning-based self-sensing method is developed to identify the location and amplitude of external mechanical loads in resin-based conductive composites with a simple electrode layout. First, conductive filler-filled resin composites are prepared, and three-dimensional conductive networks are constructed within them. Subsequently, four electrodes are installed at the edges of the composite plate, and boundary electrical resistance responses are collected when applying mechanical loads at various positions on the composite plate. Finally, a residual learning-based CNN model is proposed for the accurate localization and amplitude identification of the applied loads. Research results demonstrate that the trained CNN model can accurately and effectively determine both the load amplitude and position. The obtained localization error and amplitude error are 0.91 mm and 0.13 N, respectively, surpassing the reported error values in previous studies. The research presented here opens a new avenue for achieving highly accurate and efficient prediction of load location and amplitude, which can be widely applied in composite structural health monitoring. Full article