Search Results (5,403)

Background/Objectives: Otitis media (OM), including acute otitis media (AOM) and chronic otitis media (COM), is a common middle ear disease that can lead to significant morbidity if not accurately diagnosed. Otoscopic interpretation remains subjective and operator-dependent, underscoring the need for objective and reproducible diagnostic support. Recent advances in artificial intelligence (AI) offer promising solutions for automated otoscopic image analysis. Methods: We developed an AI-based diagnostic framework consisting of three sequential steps: (1) semi-supervised learning for automatic recognition and semantic segmentation of tympanic membrane structures, (2) region-based feature extraction, and (3) disease classification. A total of 607 clinical otoscopic images were retrospectively collected, including normal ears (n = 220), AOM (n = 157), and COM with tympanic membrane perforation (n = 230). Among these, 485 images were used for training and 122 for independent testing. Semantic segmentation of five anatomically relevant regions was performed using multiple convolutional neural network architectures, including U-Net, PSPNet, HRNet, and DeepLabV3+. Following segmentation, color and texture features were extracted from each region and used to train a neural network-based classifier to differentiate disease states. Results: Among the evaluated segmentation models, U-Net demonstrated superior performance, achieving an overall pixel accuracy of 96.76% and a mean Dice similarity coefficient of 71.68%. The segmented regions enabled reliable extraction of discriminative chromatic and texture features. In the final classification stage, the proposed framework achieved diagnostic accuracies of 100% for normal ears, 100% for AOM, and 91.3% for COM on the independent test set, with an overall accuracy of 96.72%. Conclusions: This study demonstrates that a semi-supervised, segmentation-driven AI pipeline integrating feature extraction and classification can achieve high diagnostic accuracy for otitis media. The proposed framework offers a clinically interpretable and fully automated approach that may enhance diagnostic consistency, support clinical decision-making, and facilitate scalable otoscopic assessment in diverse healthcare screening settings for disease prevention and health education. Full article

To address the power constraints of the emerging Internet of Things (IoT) era, we propose a compression-efficient feature extraction method for a CMOS image sensor that can extract binary feature data. This sensor outputs six-channel binary feature data, comprising three channels of binarized luminance signals and three channels of horizontal edge signals, compressed via a run length encoding (RLE) method. This approach significantly reduces data transmission volume while maintaining image recognition accuracy. The simulation results obtained using a YOLOv7-based model designed for edge GPUs demonstrate that our approach achieves a large object recognition accuracy (${\mathrm{AP}}_{\mathrm{L}}50$) of 60.7% on the COCO dataset while reducing the data size by 99.2% relative to conventional 8-bit RGB color images. Furthermore, the image classification results using MobileNetV3 tailored for mobile devices on the Visual Wake Words (VWW) dataset show that our approach reduces data size by 99.0% relative to conventional 8-bit RGB color images and achieves an image classification accuracy of 89.4%. These results are superior to the conventional trade-off between recognition accuracy and data size, thereby enabling the realization of low-power image recognition systems. Full article

Soft grippers exhibit excellent adaptability in handling objects of various shapes. However, due to the large deformation and high compliance of their constituent materials, the integration of sensing capabilities has long been a major research challenge. Based on the swallowing-type soft gripper proposed in previous work, this study explores the gripper’s capability to perceive object information by leveraging the characteristic that the sealed cavity undergoes volume change due to compression by the object during swallowing, thereby altering the pressure of the internal fluid medium. By establishing the geometric configuration of the sealed cavity composed of elastic membranes, the volume-pressure variation sensing model during the object swallowing process was derived. The performance of this sensing method was tested, and the application of the fluid pressure sensing strategy in closed-loop control was demonstrated, including the classification of objects by shape and sorting by size. This work provides a solution for the object shape-adaptive swallowing-type soft gripper to achieve sensory grasping functionality. Full article

Background/Objectives: Impacted third molars are frequent and may increase surgical complexity, particularly when the mandibular third molar is in close proximity to the inferior alveolar canal (IAC). This study aimed to estimate the prevalence and impaction patterns of third molars in a Portuguese population and to characterize, using a nested CBCT subsample, the three-dimensional relationship between mandibular third molars and the IAC, including cortical integrity and lingual plate thickness. Methods: A retrospective observational analysis of 1062 orthopantomograms (OPGs) was performed to determine the prevalence and panoramic patterns using Winter, Pell and Gregory classifications and Rood–Shehab signs. A consecutive CBCT subsample (n = 205) was assessed for IAC position, contact status (no contact; contact with cortical bone; contact without cortical bone), cortical integrity, and lingual plate thickness. Descriptive statistics were complemented by effect sizes to support clinical interpretability. Results: The prevalence of impacted third molars was 34.9%, occurring predominantly in the mandible. Vertical angulation was the most prevalent pattern in both jaws. In the CBCT subsample, IAC position and contact patterns varied widely, and loss of cortical integrity was more often observed when panoramic high-risk signs were present. No clinically meaningful left–right asymmetry was identified across key anatomical risk indicators. Conclusions: In this Portuguese cohort, impacted third molars showed consistent panoramic patterns, while CBCT provided clinically relevant three-dimensional risk descriptors—particularly IAC contact type and cortical integrity—supporting selective CBCT use based on anatomical risk indicators rather than routine imaging. Full article

This study aims to identify an effective neural network architecture for the task of semantic segmentation of the surface of beer wort at the stage of primary fermentation, using deep learning methodologies. Four contemporary architectures were evaluated and contrasted. The following networks are presented in both baseline and optimized forms: U-Net, DeepLabV3+, Fast-SCNN, and SegNet. The models were trained on a dataset of images depicting real beer surfaces at the primary fermentation stage. This was followed by the validation of the models using key metrics, including pixel classification accuracy, Mean Intersection over Union (mIoU), Dice Coefficient, inference time per image, and Graphics Processing Unit (GPU) resource utilization. Results indicate that the optimized U-Net achieved the optimal balance between performance and efficiency, attaining a validation accuracy of 88.85%, mIoU of 76.72%, and a Dice score of 86.71%. With an inference time of 49.5 milliseconds per image, coupled with minimal GPU utilization (18%), the model proves suitable for real-time deployment in production environments. Conversely, complex architectures, such as DeepLabV3+, did not yield the anticipated benefits, thereby underscoring the viability of utilizing compact models for highly specialized industrial tasks. This study establishes a novel quantitative metric for the assessment of fermentation. This is based on the characteristics of the foam surface and thus offers an objective alternative to traditional subjective inspections. The findings emphasize the potential of adapting optimized deep learning architectures to quality control tasks within the food industry, particularly in the brewing sector, and they pave the way for further integration into automated computer vision systems. Full article

Background/Objectives: Teacher burnout represents a complex cognitive-emotional syndrome characterized by the interplay between mental exhaustion and emotional dysregulation, threatening educational sustainability during crisis periods. This study employed comprehensive behavioral data analysis to investigate burnout syndrome patterns among Greek teachers during the COVID-19 educational crisis, aiming to identify risk factors and resilience patterns through multiple analytical approaches that capture the syndrome’s multidimensional nature. Methods: A cross-sectional study examined primary and secondary school teachers in Western Greece during the autumn of 2021. Stratified random sampling ensured representativeness across school levels, geographic locations, and employment types. Participants completed the Greek-adapted Maslach Burnout Inventory for Educators, which measured emotional exhaustion, depersonalization, and personal accomplishment. Behavioral data analysis integrated traditional statistical methods with advanced pattern recognition techniques, including classification trees for non-linear relationships, association analysis for behavioral patterns, and cluster analysis for profile identification. Results: The majority of teachers experienced high stress with inadequate coping capabilities. Classification analysis achieved high accuracy in predicting burnout severity, identifying emotional exhaustion as the primary predictor. Deputy teachers demonstrated severe cognitive-emotional strain compared to permanent colleagues across all dimensions, with dramatically reduced personal accomplishment and minimal resources. Association analysis revealed that combined low support and high workload more than doubled burnout risk. Three distinct profiles emerged: Resilient teachers, characterized by older age and permanent employment; At-Risk teachers, showing early warning signs; and Burned Out teachers, predominantly young and in precarious employment. Remote teaching, exceeding half of the workload, significantly increased strain. Multiple regression confirmed emotional exhaustion as the dominant syndrome predictor. Conclusions: Behavioral data analysis revealed complex cognitive-emotional patterns constituting burnout syndrome during educational crisis. Employment precarity emerged as the fundamental vulnerability factor, with young deputy teachers facing dramatically higher syndrome probability compared to supported senior permanent teachers. The syndrome manifests through cascading processes where cognitive overload triggers emotional exhaustion, subsequently reducing personal accomplishment. These findings provide an evidence-based framework for early syndrome identification and targeted interventions addressing both cognitive and emotional dimensions of teacher burnout. Full article

In modern power distribution networks, robust and intelligent fault management techniques are increasingly important as system complexity grows with the integration of distributed energy resources. This article reviews the use of deep learning methods for short-circuit fault detection, classification, and localization in power distribution systems, including symmetrical, asymmetrical, and high-impedance faults. The approaches examined include convolutional neural networks, recurrent neural networks, deep reinforcement learning, graph neural networks, and hybrid architectures. A comprehensive taxonomy of these models is presented, followed by an analysis of their application across the stages of fault diagnostics. Major contributions to the field are highlighted, and research gaps are identified in relation to data scarcity, model interpretability, real-time responsiveness, and deployment scalability. The paper provides an in-depth technical and performance comparison of deep learning approaches based on current research trends, and it also outlines the limitations of previous review studies. The objective of this work is to support researchers in selecting and implementing appropriate deep learning techniques for fault analytics in complex smart electricity grids with significant penetration of distributed energy resources. The review is intended to serve as an initial foundation for continued research and development in intelligent fault analytics for reliable and sustainable power distribution systems. Full article

Background/Objectives: Accurate and reliable automated dermoscopic lesion classification remains challenging. This is due to pronounced dataset bias, limited expert-annotated data, and poor cross-dataset generalization of conventional supervised deep learning models. In clinical dermatology, these limitations restrict the deployment of data-driven diagnostic systems across diverse acquisition settings and patient populations. Methods: Motivated by these challenges, this study proposes a transformer-based, dermatology-specific foundation model. The model learns transferable visual representations from large collections of unlabeled dermoscopic images via self-supervised pretraining. It integrates large-scale dermatology-oriented self-supervised learning with a hierarchical vision transformer backbone. This enables effective capture of both fine-grained lesion textures and global morphological patterns. The evaluation is conducted across three publicly available dermoscopic datasets: ISIC 2018, HAM10000, and PH2. The study assesses in-dataset, cross-dataset, limited-label, ablation, and computational-efficiency settings. Results: The proposed approach achieves in-dataset classification accuracies of 94.87%, 97.32%, and 98.17% on ISIC 2018, HAM10000, and PH2, respectively. It outperforms strong transformer and hybrid baselines. Cross-dataset transfer experiments show consistent performance gains of 3.5–5.8% over supervised counterparts. This indicates improved robustness to domain shift. Furthermore, when fine-tuned with only 10% of the labeled training data, the model achieves performance comparable to fully supervised baselines. Conclusions: This highlights strong data efficiency. These results demonstrate that dermatology-specific foundation learning offers a principled and practical solution for robust dermoscopic lesion classification under realistic clinical constraints. Full article

Background/Objectives: Accurate prognostic assessment remains crucial in metastatic renal cell carcinoma (mRCC), especially as treatment options have expanded beyond vascular endothelial growth factor (VEGF)-targeted therapies to include immune checkpoint inhibitors (ICIs) and ICI–TKI combinations. The widely used IMDC classification shows important limitations in the modern therapeutic era, highlighting the need for complementary prognostic tools. In this context, the Meet-URO and CANLPH scores—incorporating clinical, inflammatory, and nutritional markers—have emerged as promising alternatives. To evaluate and compare the prognostic performance of the Meet-URO and CANLPH scoring systems in a real-world mRCC cohort predominantly treated with first-line tyrosine kinase inhibitor (TKI) monotherapy due to limited access to ICI-based combinations. Methods: This retrospective single-center study included 112 patients with mRCC. The Meet-URO score was calculated for all patients, while the CANLPH score was assessed in 56 patients with complete laboratory data. CAR, NLR, and PHR were computed using baseline pre-treatment measurements. Overall survival (OS) and progression-free survival (PFS), the latter defined exclusively for first-line therapy, were estimated using the Kaplan–Meier method. Correlations between inflammatory markers and survival outcomes were analyzed using Spearman’s rho. Results: Meet-URO demonstrated clear prognostic stratification across all five categories, with the most favorable outcomes in score group 2 and progressively poorer OS and PFS in higher-risk groups. CANLPH also showed meaningful survival discrimination, with the highest inflammatory group (score 3) exhibiting markedly reduced OS and PFS. CAR was the strongest individual predictor of survival, while NLR and PHR showed weaker associations. Conclusions: Both Meet-URO and CANLPH provide strong, complementary prognostic information in mRCC, even in a cohort largely treated with TKI monotherapy. Their integration into routine risk assessment may enhance clinical decision-making, particularly in resource-limited settings. Full article

Background/Objectives: Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition characterized by differences in social communications and restricted, repetitive patterns of behaviors and interests, affecting approximately 1% of children globally. While functional magnetic resonance imaging (fMRI) has provided insights into altered brain connectivity patterns in ASD, classification based on neuroimaging remains a challenging due to the heterogeneity of the disorder and variability in imaging data across sites. This study employs a network-based approach using large-scale, multi-site rs-fMRI dataset from the Autism Brain Imaging Data Exchange (ABIDE I and II) to classify ASD and healthy controls using machine learning. Methods: A semi-blind Independent Component Analysis method, specifically the spatial constraint reference ICA, is applied to identify functional brain networks, and the ComBat harmonization technique is used to address site-specific variability across 11 independent datasets, ensuring consistency in feature representation. Support Vector Machines (SVMs) are employed for classification, focusing on three key networks: the Default Mode Network (DMN), Sensorimotor Network (SMN), and Visual Sensory Network (VSN). Results: The results demonstrate high classification accuracy, with the VSN achieving the highest performance (83.23% accuracy, 87.90% AUC), followed by the DMN (81.43% accuracy, 84.53% AUC) and the SMN (80.52% accuracy, 84.96% AUC), positioned with their recognized roles in social cognition and sensory–motor processing, respectively. Conclusions: The integration of ICA-based feature extraction with ComBat harmonization significantly improved classification accuracy compared to previous studies. These findings point out the potential of network-based approaches in ASD classification and point out the importance of integrating multi-site neuroimaging data for identifying reproduceable network-level features. Full article

Background/Objectives: Acute mesenteric ischemia (AMI) is a time-dependent condition associated with exceptionally high in-hospital mortality, particularly among elderly and comorbid patients. Early identification of patients at high risk of death remains challenging and has important implications for clinical decision-making. The objective of this study was to derive and internally validate a prognostic score for in-hospital mortality of patients with AMI. Materials and Methods: We conducted a multicenter, observational, retrospective cohort study including patients with AMI from 10 participating hospitals. A descriptive and analytical approach was performed. A Classification and Regression Tree (CART) model was used to determine cut-off points for continuous variables and assess their association with mortality. Based on these thresholds, a univariate analysis was performed, and variables with statistical significance (p < 0.05) were incorporated into a multivariate logistic regression model. A score—the RADIAL score—was then derived from the beta coefficients. The discriminative ability of the score was evaluated using the receiver operating characteristic (ROC) curve. Results: A total of 693 patients were studied. Thee mean age was 81 years (IQR 73–86) and 54.2% were women. A history of cardiovascular disease was present in 75.3% of participants. Overall mortality was 62.4%. Most patients (74%) were managed conservatively. Significant variables in the bivariate analysis included hypotension, age > 65 years, pH < 7.3, creatinine > 1.7 mg/dL, and absence of rectal bleeding. These variables were incorporated into the multivariate model. The resulting score showed an area under the ROC curve of 0.78 (95% CI: 0.74–0.82). Conclusions: The RADIAL score demonstrated robust predictive performance and allowed the identification of three mortality-risk groups: 30–40% (low), 50–60% (intermediate), and 80% (high). This tool may support clinical decision-making in the management of patients with AMI. Full article

Seismic monitoring is a crucial step in ensuring the safety and resilience of building structures. The implementation of effective monitoring systems, particularly across large-scale, complex building clusters, is currently hindered by the limitations of traditional sensor placement methods, which suffer from low efficiency, high subjectivity, and difficulties in replication. This paper proposes an innovative AI-based Automated Layout Method for seismic monitoring devices, leveraging building geometric recognition to provide a scalable, quantifiable, and reproducible engineering solution. The core methodology achieves full automation and quantification by innovatively employing a dual-channel approach (images and vectors) to parse architectural floor plans. It first converts complex geometric features—including corner coordinates, effective angles, and concavity/convexity attributes—into quantifiable deployment scoring and density functions. The method implements a multi-objective balanced control system by introducing advanced engineering metrics such as key floor assurance, central area weighting, spatial dispersion, vertical continuity, and torsional restraint. This approach ensures the final sensor configuration is scientifically rigorous and highly representative of the structure’s critical dynamic responses. Validation on both simple and complex Reinforced Concrete (RC) frame structures consistently demonstrates that the system successfully achieves a rational sensor allocation under budget constraints. The placement strategy is physically informed, concentrating sensors at critical floors (base, top, and mid-level) and strategically utilizing external corner points to maximize the capture of torsional and shear responses. Compared with traditional methods, the proposed approach has distinct advantages in automation, quantification, and adaptability to complex geometries. It generates a reproducible installation manifest (including coordinates, sensor types, and angle classification) that directly meets engineering implementation needs. This work provides a new, efficient technical pathway for establishing a systematic and sustainable seismic risk monitoring platform. Full article

Background/Objectives: Accurate sleep stage classification is essential for evaluating sleep quality and diagnosing sleep disorders. Despite recent advances in deep learning, existing models inadequately represent complex brain dynamics, particularly the time-lag effects inherent in neural signal propagation and regional variations in cortical activation patterns. Methods: We propose the MFST-GCN, a graph-based deep learning framework that models these neurobiological phenomena through three complementary modules. The Dynamic Dual-Scale Functional Connectivity Modeling (DDFCM) module constructs time-varying adjacency matrices using Pearson correlation across 1 s and 5 s windows, capturing both transient signal transmission and sustained connectivity states. This dual-scale approach reflects the biological reality that neural information propagates with measurable delays across brain regions. The Multi-Scale Morphological Feature Extraction Network (MMFEN) employs parallel convolutional branches with varying kernel sizes to extract frequency-specific features corresponding to different EEG rhythms, addressing regional heterogeneity in neural activation. The Adaptive Spatio-Temporal Graph Convolutional Network (ASTGCN) integrates spatial and temporal features through Chebyshev graph convolutions with attention mechanisms, encoding evolving functional dependencies across sleep stages. Results: Evaluation on ISRUC-S1 and ISRUC-S3 datasets demonstrates F1-scores of 0.823 and 0.835, respectively, outperforming state-of-the-art methods. Conclusions: Ablation studies confirm that explicit time-lag modeling contributes substantially to performance gains, particularly in discriminating transitional sleep stages. Full article

Lung cancer remains a leading cause of cancer-related mortality. Although reliable multiclass classification of lung lesions from CT imaging is essential for early diagnosis, it remains challenging due to subtle inter-class differences, limited sample sizes, and class imbalance. We propose an Adaptive Attention-Augmented Convolutional Neural Network with Vision Transformer (AACNN-ViT), a hybrid framework that integrates local convolutional representations with global transformer embeddings through an adaptive attention-based fusion module. The CNN branch captures fine-grained spatial patterns, the ViT branch encodes long-range contextual dependencies, and the adaptive fusion mechanism learns to weight cross-representation interactions to improve discriminability. To reduce the impact of imbalance, a hybrid objective that combines focal loss with categorical cross-entropy is incorporated during training. Experiments on the IQ-OTH/NCCD dataset (benign, malignant, and normal) show consistent performance progression in an ablation-style evaluation: CNN-only, ViT-only, CNN-ViT concatenation, and AACNN-ViT. The proposed AACNN-ViT achieved 96.97% accuracy on the validation set with macro-averaged precision/recall/F1 of 0.9588/0.9352/0.9458 and weighted F1 of 0.9693, substantially improving minority-class recognition (Benign recall 0.8333) compared with CNN-ViT (accuracy 89.09%, macro-F1 0.7680). One-vs.-rest ROC analysis further indicates strong separability across all classes (micro-average AUC 0.992). These results suggest that adaptive attention-based fusion offers a robust and clinically relevant approach for computer-aided lung cancer screening and decision support. Full article

Wind turbine (WT) misalignment fault diagnosis is constrained by critical signal processing challenges: weak fault features, intense background noise, and poor generalization. This study proposes a lightweight method for high-precision fault diagnosis. A fixed-threshold wavelet denoising method with the scene-specific pre-optimized parameter a (0 < a ≤ 1.3) is proposed: the parameter a is determined via offline grid search using the feature retention rate (FRR) as the objective function for typical wind farm operating scenarios. A multi-scale depthwise separable CNN (MS-DSCNN) captures multi-scale spatial features via 3 × 1 and 5 × 1 kernels, reducing computational complexity by 73.4% versus standard CNNs. An attention-based minimal peephole LSTM (AttMPLSTM) enhances temporal feature measurement, using minimal peephole connections for long-term dependencies and channel attention to weight fault-relevant signals. Joint L1–L2 regularization mitigates overfitting and environmental interference, improving model robustness. Validated on a WT test bench, the Adams simulation dataset, and the CWRU benchmark, the model achieves a 90.2 ± 1.4% feature retention rate (FRR) in signal processing, an over 98% F1-score for fault classification, and over 99% accuracy. With 2.5 s single-epoch training and a 12.8 ± 0.5 ms single-sample inference time, the reduced parameters enable real-time deployment in embedded systems, advancing signal processing for rotating machinery fault diagnosis. Full article