Search Results (132)

Alzheimer’s disease (AD) involves the accumulation of amyloid-β (Aβ) plaques, whose quantification plays a central role in understanding disease progression. Automated segmentation of Aβ deposits in histopathological micrographs enables large-scale analyses but is hindered by the high cost of detailed pixel-level annotations. Weakly supervised learning offers a promising alternative by leveraging coarse or indirect labels to reduce the annotation burden. We evaluated a weakly supervised approach to segment and analyze thioflavin-S-positive parenchymal amyloid pathology in AD and age-matched brains. Our pipeline integrates three key components, each designed to operate under weak supervision. First, robust preprocessing (including retrospective multi-image illumination correction and gradient-based background estimation) was applied to enhance image fidelity and support training, as models rely more on image features. Second, class activation maps (CAMs), generated by a compact deep classifier SqueezeNet, were used to identify, and coarsely localize amyloid-rich parenchymal regions from patch-wise image labels, serving as spatial priors for subsequent refinement without requiring dense pixel-level annotations. Third, a patch-based convolutional neural network, U-Net, was trained on synthetic data generated from micrographs based on CAM-derived pseudo-labels via an extensive object-level augmentation strategy, enabling refined whole-image semantic segmentation and generalization across diverse spatial configurations. To ensure robustness and unbiased evaluation, we assessed the segmentation performance of the entire framework using patient-wise group k-fold cross-validation, explicitly modeling generalization across unseen individuals, critical in clinical scenarios. Despite relying on weak labels, the integrated pipeline achieved strong segmentation performance with an average Dice similarity coefficient (≈0.763) and Jaccard index (≈0.639), widely accepted metrics for assessing segmentation quality in medical image analysis. The resulting segmentations were also visually coherent, demonstrating that weakly supervised segmentation is a viable alternative in histopathology, where acquiring dense annotations is prohibitively labor-intensive and time-consuming. Subsequent morphometric analyses on automatically segmented Aβ deposits revealed size-, structural complexity-, and global geometry-related differences across brain regions and cognitive status. These findings confirm that deposit architecture exhibits region-specific patterns and reflects underlying neurodegenerative processes, thereby highlighting the biological relevance and practical applicability of the proposed image-processing pipeline for morphometric analysis. Full article

Creating high-fidelity digital twins (DTs) for Industry 4.0 applications, it is fundamentally reliant on the accurate 3D modeling of physical assets, a task complicated by the inherent imperfections of real-world point cloud data. This paper addresses the challenge of reconstructing accurate, watertight, and topologically sound 3D meshes from sparse, noisy, and incomplete point clouds acquired in complex industrial environments. We introduce a robust two-stage completion-to-reconstruction framework, C2R3D-Net, that systematically tackles this problem. The methodology first employs a pretrained, self-supervised point cloud completion network to infer a dense and structurally coherent geometric representation from degraded inputs. Subsequently, a novel adaptive surface reconstruction network generates the final high-fidelity mesh. This network features a hybrid encoder (FKAConv-LSA-DC), which integrates fixed-kernel and deformable convolutions with local self-attention to robustly capture both coarse geometry and fine details, and a boundary-aware multi-head interpolation decoder, which explicitly models sharp edges and thin structures to preserve geometric fidelity. Comprehensive experiments on the large-scale synthetic ShapeNet benchmark demonstrate state-of-the-art performance across all standard metrics. Crucially, we validate the framework’s strong zero-shot generalization capability by deploying the model—trained exclusively on synthetic data—to reconstruct complex assets from a custom-collected industrial dataset without any additional fine-tuning. The results confirm the method’s suitability as a robust and scalable approach for 3D asset modeling, a critical enabling step for creating high-fidelity DTs in demanding, unseen industrial settings. Full article

With growing reliance on hydraulic fracturing to develop tight oil and gas reservoirs characterized by low porosity and permeability, optimizing proppant transport and placement has become critical to sustaining fracture conductivity and production. However, how fracture geometry influences proppant distribution under varying field conditions remains insufficiently understood. This study employed computational fluid dynamics to investigate proppant transport and placement in hydraulic fractures of which the aperture tapers linearly along their length. Four taper rate models (δ = 0, 1/1500, 1/750, and 1/500) were analyzed under a range of operational parameters: injection velocities (1.38–3.24 m/s), sand concentrations (2–8%), proppant particle sizes (0.21–0.85 mm), and proppant densities (1760–3200 kg/m³). Equilibrium proppant pack height was adopted as the key metric for pack morphology. The results show that increasing injection rate and taper rate both serve to lower pack heights and enhance downstream transport, while a higher sand concentration, larger particle size, and greater density tend to raise pack heights and promote more stable pack geometries. In tapering fractures, higher δ values amplify flow acceleration and turbulence, yielding flatter, “table-top” proppant distributions and extended placement lengths. Fine, low-density proppants more readily penetrate to the fracture tip, whereas coarse or dense particles form taller inlet packs but can still be carried farther under high taper conditions. These findings offer quantitative guidance for optimizing fracture geometry, injection parameters, and proppant design to improve conductivity and reduce sand-plugging risk in tight formations. These insights address the challenge of achieving effective proppant placement in complex fractures and provide quantitative guidance for tailoring fracture geometry, injection parameters, and proppant properties to improve conductivity and mitigate sand plugging risks in tight formations. Full article

Multi-modal remote sensing image (MRSI) matching suffers from severe nonlinear radiometric distortions and geometric deformations, and conventional feature-based techniques are generally ineffective. This study proposes a novel and robust MRSI matching method using the side window filter (MSWF). First, a novel side window scale space is constructed based on the side window filter (SWF), which can preserve shared image contours and facilitate the extraction of feature points within this newly defined scale space. Second, noise thresholds in phase congruency (PC) computation are adaptively refined with the Weibull distribution; weighted phase features are then exploited to determine the principal orientation of each point, from which a maximum index map (MIM) descriptor is constructed. Third, coarse position, orientation, and scale information obtained through global matching are employed to estimate image-pair geometry, after which descriptors are recalculated for precise correspondence search. MSWF is benchmarked against eight state-of-the-art multi-modal methods—six hand-crafted (PSO-SIFT, LGHD, RIFT, RIFT2, HAPCG, COFSM) and two learning-based (CMM-Net, RedFeat) methods—on three public datasets. Experiments demonstrate that MSWF consistently achieves the highest number of correct matches (NCM) and the highest rate of correct matches (RCM) while delivering the lowest root mean square error (RMSE), confirming its superiority for challenging MRSI registration tasks. Full article

The dynamic non-equilibrium effect (DNE) describes the non-unique character of saturation–capillary pressure relationships observed under static, steady-state, or monotonic hydrodynamic conditions. Macroscopically, the DNE manifests as variations in soil hydraulic characteristic curves arising from varying hydrodynamic testing conditions and is fundamentally governed by soil matrix particle size distribution. Changes in the DNE across porous media with discrete particle size fractions are investigated via stepwise drying experiments. Through quantification of saturation–capillary pressure hysteresis and DNE metrics, three critical signatures are identified: (1) the temporal lag between peak capillary pressure and minimum water saturation; (2) the pressure gap between transient and equilibrium states; and (3) residual water saturation. In the four experimental sets, with the finest material (Test 1), the peak capillary pressure consistently precedes the minimum water saturation by up to 60 s. Conversely, with the coarsest material (Test 4), peak capillary pressure does not consistently precede minimum saturation, with a maximum lag of only 30 s. The pressure gap between transient and equilibrium states reached 14.04 cm H₂O in the finest sand, compared to only 2.65 cm H₂O in the coarsest sand. Simultaneously, residual water saturation was significantly higher in the finest sand (0.364) than in the coarsest sand (0.086). The results further reveal that the intensity of the DNE scales inversely with particle size and linearly with wetting phase saturation (Sw), exhibiting systematic decay as Sw decreases. Coarse media exhibit negligible hysteresis due to suppressed capillary retention; this is in stark contrast with fine sands, in which the DNE is observed to persist in advanced drying stages. These results establish pore geometry and capillary dominance as fundamental factors controlling non-equilibrium fluid dynamics, providing a mechanistic framework for the refinement of multi-phase flow models in heterogeneous porous systems. Full article

Detecting small targets in infrared imagery remains highly challenging due to sub-pixel target sizes, low signal-to-noise ratios, and complex background clutter. This paper proposes PSHNet, a hybrid deep-learning framework that combines dense spatial heatmap supervision with geometry-aware regression for accurate infrared small-target detection. The network generates position–scale heatmaps to guide coarse localization, which are further refined through sub-pixel offset and size regression. A Complete IoU (CIoU) loss is introduced as a geometric regularization term to improve alignment between predicted and ground-truth bounding boxes. To better preserve fine spatial details essential for identifying small thermal signatures, an Enhanced Low-level Feature Module (ELFM) is incorporated using multi-scale dilated convolutions and channel attention. Experiments on the NUDT-SIRST and IRSTD-1k datasets demonstrate that PSHNet outperforms existing methods in IoU, detection probability, and false alarm rate, achieving IoU improvement and robust performance under low-SNR conditions. Full article

Reconfigurable Intelligent Surfaces (RISs) are among the key technologies envisaged for sixth-generation (6G) multiple-input multiple-output (MIMO)–orthogonal frequency division multiplexing (OFDM) wireless systems. However, their passive nature and the frequent absence of a line-of-sight (LoS) path in dense urban environments make uplink channel estimation and localization challenging tasks. Therefore, to achieve channel estimation and localization, this study models the RIS-mobile station (MS) channel as a double-sparse angular structure and proposes a hybrid channel parameter estimation framework for RIS-assisted MIMO-OFDM systems. In the hybrid framework, Simultaneous Orthogonal Matching Pursuit (SOMP) first estimates coarse angular supports. The coarse estimates are refined by a novel refinement stage employing a Variational Bayesian Expectation Maximization (VBEM)-based Off-Grid Sparse Bayesian Learning (OG-SBL) algorithm, which jointly updates azimuth and elevation offsets via Newton-style iterations. An Angle of Arrival (AoA)–Angle of Departure (AoD) matching algorithm is introduced to associate angular components, followed by a 3D localization procedure based on non-LoS (NLoS) multipath geometry. Simulation results show that the proposed framework achieves high angular resolution; high localization accuracy, with 97% of the results within 0.01 m; and a channel estimation error of 0.0046% at 40 dB signal-to-noise ratio (SNR). Full article

This paper addresses the challenge of robust pose estimation for spacecraft under rapid inter-frame motion, proposing a two-stage point cloud registration framework. The first stage computes coarse pose estimation by leveraging Fast Point Feature Histogram (FPFH) descriptors with random sample and consensus (RANSAC) for correspondence matching, effectively handling significant positional displacements. The second stage refines the solution through geometry-aware fine registration using raw point cloud data, enhancing precision through a multi-scale iterative ICP-like framework. To validate the approach, we simulate time-of-flight (ToF) sensor measurements by rendering NASA’s public 3D spacecraft models and obtain 3D point clouds by back-projecting the depth measurements to 3D space. Comprehensive experiments demonstrate superior performance over several state-of-the-art methods in both accuracy and robustness under rapid inter-frame motion scenarios. The dual-stage architecture proves effective in maintaining tracking continuity while mitigating error accumulation from fast relative motion, showing promise for autonomous spacecraft proximity operations. Full article

This study presents a comparative numerical investigation of Reynolds-averaged Navier–Stokes (RANS) and partially averaged Navier–Stokes (PANS) turbulence models applied to the Joubert BB2 submarine geometry under steady, calm-water conditions. To assess the influence of turbulence resolution and grid density on hydrodynamic performance prediction, simulations were conducted using three mesh resolutions—coarse, medium, and fine—based on unstructured hexahedral grids. The results were validated against international benchmark data, with emphasis placed on total resistance, pressure and shear stress distributions, wake development, and vortex structure. The PANS model consistently outperformed RANS in accurately predicting total resistance and resolving wake asymmetry, especially at medium grid resolution, due to its ability to partially resolve turbulence without full reliance on eddy viscosity assumptions. It demonstrated superior capability in capturing coherent vortex structures and preserving axial momentum in the stern region, resulting in more realistic surface pressure recovery and delayed boundary layer separation. Cross-sectional and circumferential velocity distributions in the propeller plane further highlighted PANS’s enhanced turbulence fidelity, which is essential for downstream propeller performance evaluation. Overall, the findings support the suitability of the PANS model as a practical and computationally efficient alternative to RANS for high-fidelity submarine flow simulations, particularly in wake-sensitive applications where LES remains computationally prohibitive. Full article

Godunov-based finite volume (FV) methods are widely employed to numerically solve the Shallow-Water Equations (SWEs) with application to simulate flood inundation over irregular geometries and real-field, where unstructured triangular meshing is favored. Second-order extensions have been devised, mostly on the MUSCL reconstruction and the discontinuous Galerkin (DG) approaches. In this paper, we introduce a novel second-order Runge–Kutta discontinuous Galerkin (RKDG) solver for flood modeling, specifically addressing positivity preservation and wetting and drying on unstructured triangular meshes. To enhance the RKDG model, we adapt and refine positivity-preserving and wetting and drying techniques originally developed for the MUSCL-based finite volume (FV) scheme, ensuring its effective integration within the RKDG framework. Two analytical test problems are considered first to validate the proposed model and assess its performance in comparison with the MUSCL formulation. The performance of the model is further explored in real flooding scenarios involving irregular topographies. Our findings indicate that the added complexity of the RKDG model is justified, as it delivers higher-quality results even on very coarse meshes. This reveals that there is a promise in deploying RKDG-based flood models in real-scale applications, in particular when field data are sparse or of limited resolution. Full article

Sparse autoencoders have recently produced dictionaries of high-dimensional vectors corresponding to the universe of concepts represented by large language models. We find that this concept universe has interesting structure at three levels: (1) The “atomic” small-scale structure contains “crystals” whose faces are parallelograms or trapezoids, generalizing well-known examples such as (man:woman::king:queen). We find that the quality of such parallelograms and associated function vectors improves greatly when projecting out global distractor directions such as word length, which is efficiently performed with linear discriminant analysis. (2) The “brain” intermediate-scale structure has significant spatial modularity; for example, math and code features form a “lobe” akin to functional lobes seen in neural fMRI images. We quantify the spatial locality of these lobes with multiple metrics and find that clusters of co-occurring features, at coarse enough scale, also cluster together spatially far more than one would expect if feature geometry were random. (3) The “galaxy”-scale large-scale structure of the feature point cloud is not isotropic, but instead has a power law of eigenvalues with steepest slope in middle layers. We also quantify how the clustering entropy depends on the layer. Full article

Controlling the geometry and functionality of multi-component self-assembled superstructures on surfaces is a complex task that requires numerous experimental tests. In this contribution, we demonstrate how computer modeling can be utilized to preselect functional tectons capable of forming low-dimensional architectures with tailored features. To this end, coarse-grained Monte Carlo simulations were conducted for a mixture of tripod and tetrapod units, each equipped with discrete centers for short-range directional interactions, and adsorbed onto a (111) crystalline substrate. The calculations conducted for various isomers of the tetrapod molecule revealed qualitatively distinct self-assembly scenarios, including mixing and segregation, depending on the directionality of interactions assigned to this tecton. The resulting superstructures were classified, and their formation was monitored using temperature-dependent metrics, such as coordination functions. The findings of this study contribute to a better understanding of the on-surface self-assembly of molecules with differing symmetries and can aid in the design of bicomponent overlayers for specific applications. Full article

The increasing reliance on 3D scanning and meshless methods highlights the need for algorithms optimized for point-cloud geometry representations in CAE simulations. While voxel-based binning methods are simple, they often compromise geometry and topology, particularly with coarse voxelizations. We propose an algorithm based on a Series of Local Triangulations (SOLT) as an intermediate representation for point clouds, enabling efficient upsampling and downsampling. This robust and straightforward approach preserves the integrity of point clouds, ensuring resampling without feature loss or topological distortions. The proposed techniques integrate seamlessly into existing engineering workflows, avoiding complex optimization or machine learning methods while delivering reliable, high-quality results for a large number of examples. Resampled point clouds produced by our method can be directly used for solving PDEs or as input for surface reconstruction algorithms. We demonstrate the effectiveness of this approach with examples from mechanically sampled point clouds and real-world 3D scans. Full article

In three-dimensional (3D) measurement, the motion of objects inevitably introduces errors, posing significant challenges to high-precision 3D reconstruction. Most existing algorithms for compensating motion-induced phase errors are tailored for object motion along the camera’s principal axis (Z direction), limiting their applicability in real-world scenarios where objects often experience complex combined motions in the X/Y and Z directions. To address these challenges, we propose a universal motion error compensation algorithm that effectively corrects both pixel mismatch and phase-shift errors, ensuring accurate 3D measurements under dynamic conditions. The method involves two key steps: first, pixel mismatch errors in the camera subsystem are corrected using adjacent coarse 3D point cloud data, aligning the captured data with the actual spatial geometry. Subsequently, motion-induced phase errors, observed as sinusoidal waveforms with a frequency twice that of the projection fringe pattern, are eliminated by applying the Hilbert transform to shift the fringes by π/2. Unlike conventional approaches that address these errors separately, our method provides a systematic solution by simultaneously compensating for camera-pixel mismatch and phase-shift errors within the 3D coordinate space. This integrated approach enhances the reliability and precision of 3D reconstruction, particularly in scenarios with dynamic and multidirectional object motions. The algorithm has been experimentally validated, demonstrating its robustness and broad applicability in fields such as industrial inspection, biomedical imaging, and real-time robotics. By addressing longstanding challenges in dynamic 3D measurement, our method represents a significant advancement in achieving high-accuracy reconstructions under complex motion environments. Full article

This study examines the static bending behavior of functionally graded beams using a newly developed modified Timoshenko beam element. The mixed finite element formulation and Timoshenko beam theory serve as the foundation for the formulation of the proposed beam element. There are two nodes and two degrees of freedom in each node of the new beam element. The suggested element is free of shear locking, without the need for reduced or selective integrations, because of the mixed finite element formulation. Comparative results demonstrate high accuracy in computations, even with both regular and irregular meshes, as well as coarse and fine discretization. Because of its rapid convergence rate, the proposed element is an excellent tool for analyzing beam structures with complex geometries and load conditions. Several examples are provided to demonstrate the accuracy and high convergence of the proposed beam element. Additionally, the effects of various parameters, such as the power-law index and thickness-to-length ratio, on the bending behavior of functionally graded beams are investigated. The findings highlight the robustness and versatility of the developed beam element, which makes it a useful contribution to research into the computational mechanics of beam structures. Full article