Search Results (1,290)

Background/Objectives: Rotator cuff tears are a leading cause of shoulder disability. While multi-sequence MRI is standard, the optimal deep learning integration of heterogeneous image series and clinical metadata remains unresolved. This study evaluated a hierarchical, sequence-aware multimodal framework for patient-level binary rotator cuff tear classification. Methods: A single-center cohort of 199 patients (100 tears, 99 controls) was analyzed across four MRI sequences (T1 coronal, T2 fat-suppressed sagittal, and proton density [PD] fat-suppressed coronal and transverse/axial) and nine demographic features. Under a patient-level stratified three-fold cross-validation scheme preventing data leakage, we evaluated ResNet50 and Vision Transformer baselines (Study 0), full-protocol fusion topologies (Study 1), and systematically mapped sequence-subset combinations with or without metadata (Study 2). Results: In Study 0, the PD coronal ResNet50 model was the top baseline (AUC = 0.9834, F1 = 0.9515). In Study 1, late decision fusion yielded the highest AUC (0.9909), while feature concatenation optimized threshold balance (F1 = 0.9502). In Study 2, a streamlined three-sequence subset with metadata (C14M: T2 + PDc + PDt) achieved peak performance (AUC = 0.9961, 95% CI: 0.9823–0.9987, F1 = 0.9618, MCC = 0.9238), outperforming the full protocol (AUC = 0.9909, F1 = 0.9355). Metadata utility was configuration-dependent, assisting only fluid-sensitive combinations. Conclusions: Rather than indiscriminately aggregating entire clinical protocols, multimodal fusion is optimized by selecting complementary imaging series. For binary classification, excluding non-fat-suppressed T1 images in favor of a streamlined T2 and PD set stabilized by clinical demographics maximized classification performance in this internally validated, single-center cohort. Full article

To address the high non-stationarity of COVID-19 pandemic time-series data and the severe error accumulation issue in long-horizon forecasting, a spatiotemporal two-branch multi-step forecasting model named CFC-FLCA is proposed. This model integrates closed-form continuous-time neural networks (CFC), a hybrid decomposition strategy combining variational mode decomposition (VMD) and jump plus AM-FM mode decomposition (JMD), and a cross-attention (CA) mechanism. First, the VMD-JMD hybrid mode decomposition method is applied to preprocess raw new case sequences. By leveraging the complementary advantages of the two decomposition algorithms, non-stationary sequences are adaptively decomposed into high-frequency noise components and low-to-mid-frequency trend-periodic components, eliminating random disturbance interference at the data source. On this basis, a time–frequency dual-branch feature extraction network is constructed. CFC provides ultra-long-range temporal dependency modeling capability; the time-domain branch adopts Legendre projection units (LPU) to extract robust temporal evolution features, while the frequency-domain branch employs frequency-enhanced units (FEU) to uncover latent periodic patterns that are difficult to capture using traditional time-domain methods. A cross-attention mechanism is introduced to dynamically learn the importance weights of time–frequency-domain features, enabling the adaptive deep integration of complementary information and effectively mitigating error accumulation in long-horizon forecasting. Multi-step forecasting experiments are conducted on real-world COVID-19 datasets from Belgium, the Czech Republic, and Ireland, with comprehensive comparisons against mainstream time series forecasting models. The experimental results demonstrate that the CFC-FLCA model outperforms all comparison models across all evaluation metrics for all prediction horizons. Full article

Background: Accurate segmentation of brain tumors from multi-modal magnetic resonance imaging (MRI) is essential for diagnosis, treatment planning, and therapy monitoring. However, this task remains challenging due to tumor heterogeneity, irregular boundaries, and the complex anatomical structure of surrounding tissues. In particular, precise delineation of tumor sub-regions—including whole tumor, tumor core, and enhancing tumor—continues to be a major limitation of existing automated methods. Methods: In this study, we propose a novel hybrid CNN–Transformer framework that integrates local feature extraction with global contextual modeling for improved brain tumor segmentation. The architecture consists of three main components: a dual-pathway encoder for capturing fine-grained and contextual features, a multi-scale feature fusion module based on spatial pyramid pooling with dense connections, and a boundary-aware attention decoder designed to enhance segmentation accuracy around tumor edges. The model utilizes four MRI modalities (T1, T1ce, T2, and FLAIR) to capture complementary tumor characteristics. In addition, a hybrid loss function combining Dice, focal Tversky, and boundary losses is employed to address class imbalance and improve boundary precision. Results: Experimental results on the BraTS 2023 dataset demonstrate superior performance, achieving Dice scores of 92.3%, 88.7%, and 84.5% for whole tumor, tumor core, and enhancing tumor, respectively, while maintaining high computational efficiency. Conclusion: The proposed framework achieves accurate and robust brain tumor segmentation by effectively integrating local and global features, demonstrating its potential for automated multi-modal MRI analysis in clinical practice. Full article

As a critical process in the manufacturing of core components for equipment, milling directly influences workpiece quality and production efficiency through the process of tool wear. However, current monitoring methods face limitations in extracting comprehensive wear features from multi-source signals, while data-driven approaches are constrained by scarce wear samples and imbalanced stage distribution in practical applications. To address these challenges, this paper presents an unsupervised tool wear state classification method based on multi-modal feature fusion. The proposed approach establishes a dual-channel feature extraction model that processes vibration signals and workpiece surface texture images separately, enabling complementary-information integration through a fusion strategy. An autoencoder architecture performs unsupervised dimensionality reduction and reconstruction of the fused high-dimensional features, extracting low-dimensional representations strongly correlated with wear states. By optimizing feature discriminability through reconstruction error minimization, the method effectively mitigates the impact of limited labeled data on model performance. Experimental validation using cemented carbide milling cutters demonstrates that the proposed method achieves a classification accuracy of 93.33% under unsupervised conditions, significantly outperforming conventional unsupervised approaches. This research offers a viable approach for intelligent perception and precise analysis of cutting processes, presenting practical value for advancing equipment condition monitoring, process optimization, and predictive maintenance within intelligent manufacturing systems. Full article

Reliable identity preservation is essential in retail multi-object tracking because an identity switch may assign dwell time, shelf interactions, or demographic statistics to the wrong consumer trajectory. This challenge is amplified in crowded indoor scenes, where occlusions and appearance ambiguity weaken conventional association cues. This work presents a demographic-aware multi-object tracking framework for retail environments that exploits spatial, motion-based, appearance-based, and demographic cues within a modular tracking-by-detection pipeline. The proposed approach integrates IoU-based spatial association, LSTM-based motion forecasting, ReID appearance embeddings, and apparent demographic information derived from age-group and gender predictions into a common association cost. The core assumption is that demographic attributes should remain consistent across neighboring frames for the same tracked individual; therefore, demographic agreement can serve as a weak semantic cue during detection-to-tracklet association. Unlike conventional pipelines that treat multi-object tracking and demographic estimation as independent stages, the proposed method reuses the outputs of an existing Inception-based apparent demographic classifier to support identity preservation, without introducing a new demographic estimation model or requiring end-to-end retraining. The framework is evaluated on RGB retail tracking sequences from the Consumers dataset under both raw and privacy-aware anonymized settings. The proposed method achieves 84.8% MOTA and 87.8% IDF1 on raw data and 80.0% MOTA and 85.3% IDF1 under anonymization, improving the strongest evaluated baseline by 0.4 and 0.6 percentage points and by 1.0 and 1.3 percentage points, respectively, while preserving the same low ID-switch counts. Although incremental in absolute terms, these gains are consistent across both evaluation settings and are obtained over a strong baseline in challenging indoor retail sequences. The results indicate that even a lightweight demographic-consistency cue can provide measurable complementary information for improving consumer trajectory stability. Full article

In the digital transformation of the petroleum industry, massive multi-source heterogeneous tables are distributed across databases, Word documents, PDF reports, and engineering systems. Their unstructured format and weakly expressed inter-table relationships make it difficult for conventional keyword-based or single-table retrieval methods to locate the complete set of tables needed for complex queries. To address this problem, this paper proposes Relatab, a multi-table retrieval framework based on implicit association reasoning. Relatab first estimates query–table relevance through a dual-level semantic matching mechanism that combines table-level signals, including captions and column names, with value-level signals weighted by entropy and CRITIC criteria. It then constructs an implicit table graph using column-name and column-content similarity, and applies a max-product multi-hop propagation rule with decay and pruning to identify complementary tables that are not directly matched by the query. Finally, direct relevance and inter-table complementarity are fused to produce the retrieved table set. Experiments on Spider, Bird, and CementingTables show that Relatab achieves Top-2 recall rates of 79.21%, 61.26%, and 77.88%, respectively, outperforming DTR by 1.74, 2.33, and 1.85 percentage points. The results indicate that explicit modeling of implicit inter-table associations improves retrieval coverage in complex multi-table scenarios while remaining applicable to petroleum-domain documents. Full article

We propose a modular, retrieval-augmented pipeline for computational argumentation that integrates two complementary components: ArgStance, a multi-task model for argument and stance reasoning, and TargetMatch, a contrastive retrieval model that treats target identification as a first-class retrieval task. We further formulate stance detection as a framing-aware problem, recognizing that the polarity of a stance toward a target depends on how the proposition is framed. To support broad generalization, we construct a large dataset spanning Kialo discussions, Wikipedia, and curated news articles, and introduce a cross-source injection strategy that mitigates domain and style biases. Our compact models achieve F1 scores of 0.94 for argument detection (ModernBERT-base) and 0.84 for same-side stance detection (ModernBERT-large), while TargetMatch attains a top-10 retrieval accuracy of 0.75. Under controlled zero-shot comparisons with large language models, our models remain competitive while offering advantages in reproducibility, deployment cost, and controllable intermediate predictions. Full article

Fusing infrared and visible images can effectively compensate for the inherent limitations of each modality in different scenes, resulting in fused images that contain richer information. However, existing methods often struggle to balance global dependency modeling with local detail preservation and to effectively coordinate heterogeneous local and global features during fusion. To address these issues, this paper proposes a Significance-Preserving Progressive Fusion Network (SiPFusion). First, a progressive feature extraction framework was designed, which hierarchically extracts multi-scale local features using CNNs and then models long-range dependencies across scales via a Transformer-based global module. To adaptively integrate local-global complementary features, a significance-preserving fusion module was designed to obtain significance attention maps with a spatial selection mechanism, enabling dynamic fusion of multi-source features. Furthermore, we propose a significance similarity loss function that leverages intermediate feature guidance to enhance structural consistency and preserve salient-region information in the fused image. Extensive experiments on the MSRS, RoadScene, and TNO datasets demonstrate that SiPFusion achieves competitive visual quality and strong overall quantitative performance against 15 state-of-the-art fusion methods, obtaining leading results on most evaluated metrics. Full article

The large-scale integration of wind and solar power has significantly intensified peak-shaving pressure and operational risk in provincial power grids. Effectively leveraging the flexible regulation capability of hydropower to mitigate the uncertainty of wind and solar output is a promising approach to enhancing grid security and stability. To simultaneously improve the peak-shaving performance and risk resilience of hydro-wind-solar systems for a provincial power grid, this paper proposes a multi-objective short-term scheduling model that jointly minimizes the peak value of net load and the Conditional Value-at-Risk (CVaR) of flexibility shortage. Specifically, the residual peak load is used to quantify the system’s peak-shaving burden, while the average CVaR of upward/downward ramping deficits across all time periods characterizes the tail risk associated with insufficient flexibility. Historical wind and solar forecast error data are employed to generate representative uncertainty scenarios via Gaussian mixture model, and the Rockafellar–Uryasev formulation is adopted to accurately embed CVaR into a mixed-integer linear programming (MILP) framework. Furthermore, the normalized normal constraint (NNC) method is introduced to compute a well-distributed Pareto front. Numerical simulations based on a real-world hydro-wind-solar system in a provincial grid in Southwest China demonstrate that the proposed model can significantly reduce the peak load while effectively mitigating flexibility shortfall risk. The resulting Pareto front clearly reveals the trade-off between peak-shaving effectiveness and risk control, providing a scientific basis for day-ahead generation scheduling and coordinated dispatch of flexible resources. Full article

Accurate point-cloud registration is a fundamental task in intelligent 3D perception systems, including autonomous robotics, indoor digital twins, augmented reality, smart sensing, and scene-level spatial understanding. However, reliable registration in structured indoor environments remains challenging because repetitive architectural layouts, extensive planar regions, limited overlap, and rotational symmetries often weaken local geometric distinctiveness and lead to ambiguous correspondences. In addition, appearance inconsistency caused by illumination changes, exposure variation, and sensor-dependent color responses further degrades the reliability of cross-view matching. To address these issues, this paper presents MaCSE-Reg, an indoor registration method that combines explicit Manhattan-axis modeling with independent multi-color-space point-wise encoding. Rather than proposing a wholly new multi-modal paradigm, MaCSE-Reg builds on prior geometry–color fusion studies and focuses on two under-emphasized design choices for structured indoor scenes: (i) dominant orthogonal axes estimated from surface normals, regularized by an axis-consistency objective and gated by an axis-confidence score, and (ii) complementary RGB, Lab, and HSV representations learned through independent point-wise encoders for illumination-tolerant appearance matching. Geometric, structural, and color-aware features are adaptively fused, and a learnable multi-modal matching formulation is used to refine correspondences before weighted Procrustes pose estimation. Experiments on standard indoor registration benchmarks demonstrate that MaCSE-Reg improves registration robustness and pose accuracy under low-overlap conditions, repetitive structures, and illumination variations. The results support the value of this specific structural–appearance combination for intelligent indoor 3D perception while remaining complementary to existing geometry–color registration frameworks. Full article

High-resolution land cover classification is critical for monitoring environmental change and managing natural resources. This study presents an unsupervised framework with externally guided feature prioritization that integrates Sentinel-1 synthetic aperture radar (SAR) and Sentinel-2 optical imagery at 10 m spatial resolution. A cloud-native export protocol in Google Earth Engine (GEE) enables the generation of consistent, cloud-free, and snow-free seasonal composites across Ontario, Canada. A comprehensive feature engineering pipeline combines spectral indices, radar backscatter metrics, terrain derivatives from digital elevation models (DEMs), and temporal statistics to create a rich multi-sensor input space. Dimensionality reduction is performed using Sparse Principal Component Analysis (SparsePCA) and mutual-information-based feature selection. Clustering is conducted using three complementary algorithms: centroid-based K-means, density-based Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN), and reachability-based Ordering Points To Identify the Clustering Structure (OPTICS). Final land cover labels are assigned via a majority-voting ensemble, with prediction ties resolved deterministically using OPTICS. OPTICS is particularly effective for modeling heterogeneous landscapes due to its ability to detect clusters of varying density without requiring a global threshold. This study is designed as a pilot-site methodological demonstration using three representative 2 km × 2 km regions in Ontario, rather than a full provincial-scale land cover product. The resulting classification maps are validated against reference land cover data, demonstrating the effectiveness and potential scalability of the proposed external-label guided unsupervised mapping approach. Full article

Reliable obstacle detection and classification are essential capabilities for the safe and efficient navigation of Marine Autonomous Surface Ships. This paper introduces a decision-level multi-sensor fusion framework to enhance situational awareness for autonomous vessels by integrating RGB camera and LiDAR data. Visual information is processed using a pre-trained, open-source object detection model. At the same time, LiDAR measurements are analysed with a clustering-based algorithm, followed by a lightweight Random Forest classifier for semantic labelling. To support practical deployment in real maritime environments, the proposed approach relies on readily available perception modules, avoiding the need for training on proprietary datasets and limiting dependence on extensive task-specific tuning. The fusion of these complementary sources is employed to confirm and characterise dynamic obstacles, whose positions derived from LiDAR are continuously tracked using a Global Nearest Neighbour algorithm supported by a Kalman filter. Each stage of the proposed processing chain is thoroughly described and experimentally validated using real-world data collected in a representative marine environment, demonstrating the approach’s effectiveness in improving perception performance by reducing false positives from noisy measurements and achieving 92% track number accuracy in a complex scenario. Full article

The boost chopper (HS) is a core electrical component of the 440 V grid in high-speed maglev trains, providing reliable power for battery charging and auxiliary systems. Fault diagnosis of the HS is crucial for identifying operational faults and ensuring stable train operation. However, HS faults exhibit both long-period fluctuations and transient characteristics, which are difficult for a single network to capture synchronously. This paper proposes a multi-scale fault diagnosis method based on a TimesNet-CNN dual-branch architecture, constructing a parallel and complementary feature extraction mechanism. The TimesNet branch uses Fast Fourier Transform (FFT) to adaptively identify dominant periods, reshaping the 1D sequence into a 2D structure to explicitly model the global evolution of intra-period fluctuations and inter-period trends via Inception convolution. Meanwhile, the CNN branch employs stacked small convolutional kernels and hierarchical downsampling to extract local high-frequency anomaly features. After feature fusion, the method achieves synergistic discrimination of global periodicity and local transiency. Finally, experiments were conducted on a real-world dataset containing 11 system states (10 fault types and 1 normal state). Experimental results show that the proposed method outperforms TimesNet, CNN, ResNet and Informer models in precision, recall and F1-score. This validates the effectiveness of the dual-branch feature fusion mechanism in capturing multi-scale fault features, achieving high-precision identification of HS faults. Full article

Curling is a strategic sport in which shot decisions involve both expected rewards and inherent risks; expected-score distributions (ESDs)—probability distributions over possible final scores—capture this uncertainty as a risk-aware strategic indicator. Although unified frameworks predicting ESDs across all shots have been proposed, their internal behavior and learned output characteristics have not been systematically examined. We construct an all-shot ESD prediction framework grounded entirely in real-match board configurations drawn from World Curling Federation championship events and conduct a shot-wise analysis along two complementary axes: accuracy, the model’s reproduction of its training targets, and plausibility, the validity of those targets at aggregate and in-match scales. Accuracy degrades monotonically with shot number; only shot 16 admits comparison with an independent reference, while intermediate-shot labels are bootstrap rollouts of the next-shot model. An independent ground-truth probe at the first backward step (shot 15,500 boards × 95 contexts) bounds the consequence of this bootstrap structure: the resulting deviation is a small selection-induced hammer-underestimation bias common to both FiLM and concatenation chains, whose magnitude is more than an order of magnitude larger than the FiLM chain versus concatenation chain stage difference, and full multi-step verification beyond the first step is structurally out of reach. Within this scope the aggregate ESD shares the gross shape of the empirical end-score distribution and the qualitative hammer/non-hammer ordering, with a hammer-favorable offset attributed—after disentangling intent from execution on the shot-percentage 100% subset—to a label execution-noise envelope that is tighter than the realized play of top-tier matches. Within real matches the chain responds smoothly to each delivered stone and, at the directly validated terminal shot, assigns mean probability $\approx 0.71$–$0.72$ to the realized end score under intended execution (against $0.28$ for a marginal-frequency baseline), transferring from senior to junior populations. The framework is best read as an internally coherent chain anchored to a directly validated final-shot calibration, statistically plausible under intended execution with a clear boundary at execution failure. Two structural limitations remain: rare high-magnitude outcomes are scarce in real data and produce a heavy upper tail of accuracy errors, and a single fixed execution-noise envelope that is tighter than top-tier realized play accounts for the aggregate hammer-side offset and motivates recalibration against real-match execution statistics. Full article

Bioactive peptides constitute a highly diverse and therapeutically relevant molecular class, yet their systematic exploration remains challenging because of the vast size, heterogeneity, and fragmented annotation of peptide chemical space. In this context, complex networks have emerged as a complementary computational framework for organizing, analyzing, and exploiting peptide diversity. This methodological review examines the main components of graph-based peptide informatics, from graph-based data integration and curated repositories to descriptor-based representations, similarity-driven network construction, and topology-informed analysis. We describe how peptide sequences can be projected into multidimensional reference spaces using molecular descriptors, aggregation operators, and unsupervised feature selection, and how these representations support the construction of Chemical Space Networks, Half-Space Proximal Networks, and Metadata Networks. Special attention is given to topological analysis, including threshold selection, community detection, and centrality-based identification of representative peptides and scaffolds. We also review the development of Multi-query Similarity Searching Models as training-independent, topology-guided alternatives to conventional supervised predictors. Finally, we highlight the implementation of these methodologies in computational resources such as StarPepDB, StarPep Toolbox, and StarPepWeb, which illustrate the transition of peptide network science from conceptual workflows to accessible, scalable, and reproducible infrastructures. Overall, complex networks are presented as a mature and interpretable paradigm for the structured exploration, analysis, and discovery of bioactive peptides. Full article