Search Results (2,846)

Physically consistent estimation of soft-tissue mechanical properties is critical for surgical robotics, intraoperative safety monitoring, and simulator initialization, yet existing methods typically require force-sensing hardware or manual parameter tuning. This paper presents a physics-informed generative framework that estimates tissue stiffness ($k_s$), damping coefficient ($k_d$), and tool–tissue contact force magnitude ($F_{mag}$) from monocular laparoscopic video in a label-free manner with respect to mechanical parameters and interaction forces. The pipeline integrates three components: DepthPro, a multi-scale Vision Transformer (ViT) for zero-shot metric depth estimation; a 3D geometric contact detection pipeline; and a dual-mode conditional generative network trained via a five-term physics–adversarial loss. A differentiable Mass–Spring–Damper (MSD) simulator is embedded directly in the training loop. This enables gradient-based parameter learning without force-sensor, displacement, or boundary-condition supervision. Parameter identifiability is supported through dual observational grounding: MSD physics consistency against observed contact displacement, and next-frame depth map reconstruction. Validated on CholecSeg8k cholecystectomy sequences, Physics-Informed Neural Network (PINN)-estimated parameters significantly outperform static literature baselines (Wilcoxon p = 2.49 × 10⁻⁸, Cohen’s d = 0.374), with physically plausible viscoelastic settling dynamics recovered within 0.9 s of tool release. Since no force sensors were present at acquisition time, evaluation follows an indirect simulation-consistency protocol. Mechanical parameters are estimated at 1.6 ms/frame, a negligible addition to the monocular depth front end that sets the pipeline rate. Estimated parameters directly enable stiffness-aware haptic rendering, intraoperative safety monitoring, and scene-adapted surgical simulation initialization. Full article

Deep learning surrogate models trained with mean-squared-error loss produce statistically accurate but physically unconstrained flood predictions: water may flow uphill, appear spontaneously, or smooth over street-level corridors. In this work, a physics-informed training framework is developed for CNN-LSTM models that predict urban flood depths at 15 min intervals over a $128\times 128$ spatial grid. Three differentiable penalty terms are embedded directly into the loss function: (i) a gravity loss that penalizes depth increases against the water-surface-elevation gradient, (ii) a continuity loss enforcing local mass conservation with rainfall-adaptive thresholds, and (iii) a topography-aware false-alarm penalty modulated by the topographic wetness index (TWI). The framework is evaluated on the Norfolk, Virginia, flood dataset spanning two major storm events (August 2017 and September 2022) comprising 300 samples, with all variants trained on identical splits and robustness assessed over repeated random splits and leave-one-storm-out tests. A road-proximal evaluation restricted to a TWI-derived street mask quantifies street-level skill. The physics-constrained model achieves near-zero gravity violations (∼10⁻⁶) and the highest street-channel recall (0.77 ± 0.09 versus 0.44 ± 0.10 for the unconstrained baseline), the capability most relevant to downstream traffic routing, and its recall advantage more than doubles on a held-out storm, while a uniform false-alarm variant attains 16% lower mean absolute error but suppresses street recall to 0.25. The proposed TWI-modulated penalty reconciles this trade-off: it improves upon the uniform variant on every metric measured, recovering 60% higher street recall at the lowest MAE among all constrained variants and the best street-level F1 score. These results expose a fundamental tension between aggregate pixel-level error metrics and application-specific physical plausibility, and demonstrate that terrain-aware loss modulation offers a principled resolution. Full article

Liquid crystal (LC) is a key enabling technology for continuously phase-reconfigurable microwave devices, offering analogue-tuning capabilities distinct from discrete alternatives such as MEMS and p-i-n diodes. However, the insertion loss of LC-based phase shifters is inevitably influenced by conductor surface roughness—a factor often neglected in idealised simulations. This paper presents, for the first time, a rigorous numerical quantification of how metal surface roughness affects the insertion loss and phase shift of a 60 GHz LC-filled coaxial delay line (0–180° phase shifter) with radial conductor surfaces instead of conventional planar ones. Using full-wave finite-element simulations incorporating Groisse’s phenomenological model and Huray’s snowball model, four surface configurations are analysed at 54–66 GHz: perfectly smooth conductors, roughness on both inner and outer conductors simultaneously, and roughness applied to each conductor individually. Results show that roughness induces a measurable increase in insertion loss—worst when both conductors are rough—but its impact on differential phase shift remains minimal (<0.32°). Huray’s model predicts conductor losses 1.77 times higher than Groisse’s model, yielding more conservative metrics. For the insertion loss evaluation in Case 2 at 60 GHz under the reference isotropic LC state, Groisse’s model predicts 1.90921 dB, while Huray’s model predicts 2.16319 dB, a 0.25 dB discrepancy (12% uncertainty relative to the mean). The inner conductor dominates roughness-induced losses due to concentrated current density, suggesting prioritised surface finishing of the core line. This study isolates loss mechanisms in a coaxial LC structure, providing insights into low-loss reconfigurable devices. Practical PCB copper foil fabrication methods are also evaluated with quantitative analysis of non-ideal cylindrical geometries. Full article

Background: The timing and efficacy of Achilles tendon lengthening (ATL) in Duchenne muscular dystrophy (DMD) remain controversial because indiscriminate surgery can accelerate ambulation loss. Case Description: This report presents a 5-year longitudinal comparative analysis (2021–2026) of monozygotic twins with identical genetic backgrounds (DMD exon 30–43 deletion) who received comparable rehabilitation and pharmacological management, including concurrent gene therapy in 2025. Twin A was managed conservatively with orthosis and subsequent serial casting for progressive equinus contracture, whereas Twin B underwent early bilateral ATL during transition to the non-ambulatory phase. Despite a temporary postoperative decline, Twin B demonstrated a more stable longitudinal motor trajectory, outperforming Twin A in gross motor function (Gross Motor Function Measure-88: 38.60% vs. 32.85% in 2026) by preserving residual standing and crawling dimensions. Longitudinal KIDSCREEN-52 assessments revealed a clinically meaningful improvement in psychological well-being in Twin B (+10.9 points) relative to Twin A. Conclusions: This single-pair twin case serves as a hypothesis-generating observation, highlighting a potential longitudinal association between early surgery and attenuated progressive motor decline within a multi-disciplinary care regimen. Given highly variable psychosocial outcomes, these trends cannot be directly attributed to surgery alone. When evaluating such orthopedic interventions, clinical decisions may benefit from looking beyond immediate gait metrics toward broader, long-term functional trends, though the relative contributions of concurrent therapies remain to be elucidated. Full article

3D Gaussian Splatting (3DGS) SLAM provides an explicit scene representation for dense tracking and mapping, which is useful for underwater robotic perception. However, underwater monocular 3DGS SLAM lacks reliable metric depth cues: monocular depth estimation can provide dense structural priors, but its scale and reliability often degrade under underwater appearance changes. Forward-looking sonar (FLS) provides range–azimuth acoustic measurements whose range coordinate is related to physical distance, but raw sonar observations are sparse, noisy, and ambiguous. Our key insight is that FLS returns can serve as sparse metric depth anchors when they are associated with visually detected object regions. Based on this insight, we propose SonarReg-GS SLAM, an underwater visual–acoustic 3DGS SLAM framework with sparse sonar-guided depth regularization. Given synchronized RGB and sonar inputs, SonarReg-GS SLAM uses object masks to constrain the search space for acoustic range association. Filtered sonar responses are selected as sparse metric anchors through object-aware sampling, bearing-to-beam gating, and valid-pair checking. These anchors regularize the scale of monocular depth and generate metric depth priors for Gaussian initialization and tracking. An object-aware RGB mask loss further increases supervision on detected object regions while preserving full-scene mapping. Experiments on two public RGB–sonar underwater datasets show that SonarReg-GS SLAM improves tracking accuracy and mapping quality compared with representative classical SLAM and Gaussian Splatting SLAM baselines. Compared with Splat-SLAM, our method reduces the average ATE RMSE from 0.1296 m to 0.1015 m on UXO and from 0.5687 m to 0.4640 m on OPTI, corresponding to relative reductions of 21.7% and 18.4%, respectively. For rendering-based mapping, it increases the average PSNR from 28.05 dB to 29.61 dB on UXO and from 20.91 dB to 28.63 dB on OPTI while reducing the average LPIPS from 0.345 to 0.173 and from 0.450 to 0.303, respectively. Full article

Internal temperature distributions in black tea fermentation regulate enzymatic oxidation, heat accumulation and fermentation uniformity, but continuous three-dimensional measurements remain difficult in practical processing. We developed a CFD-prior-constrained physics-informed neural network (PINN + CFD) to reconstruct the three-dimensional thermal state of a 1.35 m × 0.96 m × 0.08 m fermentation bed under sparse sensing. Nine sensors at z = 0.04 m were used for training, and six held-out depth-wise sensors at z = 0.02 m and z = 0.06 m were reserved for depth-wise validation. The model integrated measured temperatures, transient heat-transfer physics, convective boundary conditions and a CFD-derived volumetric soft spatial prior, which guided spatial extrapolation rather than serving as ground-truth temperature data. Although the multilayer perceptron achieved the lowest fitting error at the instrumented z = 0.04 m plane, PINN + CFD showed better depth-wise extrapolation, with RMSEs of 0.128, 0.129 and 0.296 °C across the three stages. However, its advantage was stage- and validation-target-dependent: the baseline PINN was slightly better in part of the dynamic-stage validation, and standalone CFD had the lowest surface infrared error in the constant-temperature stage, indicating that PINN + CFD mainly improved spatial extrapolation rather than uniformly minimizing all error metrics. The inferred apparent process-level heat-source index Q_react(t) varied continuously, and its cumulative trajectory showed a descriptive association with cumulative polyphenol loss. These results indicate that PINN + CFD enables physically consistent thermal-state reconstruction within the tested sparsely instrumented black tea fermentation bed. Full article

Robot person following (RPF) in dense crowds requires a mobile robot to maintain an appropriate relative position with respect to a moving target while avoiding surrounding pedestrians and satisfying rear-following and social constraints. This paper proposes PPO-GAT-Follow, an interaction-aware reinforcement learning framework for dense-crowd RPF under geometric visibility loss with available target-relative pose estimates. The follower, target pedestrian, and surrounding pedestrians are represented as graph nodes, and a graph attention encoder models their local interactions. A task-oriented reward mechanism jointly accounts for target maintenance, visibility preservation, collision avoidance, proximity-aware social compliance, rear position maintenance, post-arrival stabilization, and action stability. Experiments are conducted in IR-SIM under fixed-route and random-route settings, with comparisons against MPC, DWA, SFM, and an adapted SARL baseline. In the fixed-route setting with 12 background pedestrians, PPO-GAT-Follow achieves a task success rate of $98.8\%$ and a collision rate of $1.1\%$, improving task success by 10.9 percentage points over MPC. In the random-route setting at the training density, it achieves $83.1\%$ task success and an SPL of $0.815$, outperforming MPC by 18.3 percentage points in task success; at this density, it also surpasses SARL in the main task-level metrics. Zero-shot evaluations across crowd densities, together with structural and reward ablations, reward weight sensitivity analysis, tolerance shift tests, multi-seed training, and stress testing under target pose noise and heterogeneous pedestrian dynamics, further demonstrate the effectiveness and reliability of the proposed framework. Gazebo-based validation also demonstrates system integration feasibility with localization, point cloud-based surrounding pedestrian perception, tracking, and UWB-like target-relative pose input. Nevertheless, visual target identification, re-identification, and perception-level occlusion recovery remain outside the scope of the present validation. Full article

Background: Hearing health in Kazakhstan is documented through parallel administrative systems whose indicators are often conflated. We examined three administrative layers of hearing health during 2020–2025: dispensary registration of ear and hearing-loss diagnoses, disability certification as a severe administrative endpoint, and cochlear implantation (CI) as the specialised treatment response, with reference to Global Burden of Disease (GBD) 2021 estimates. Methods: We conducted a national ecological, region-year, multi-source administrative registry study across all 20 regions of Kazakhstan from 2020 to 2025. The dataset included 114 region-year observations and integrated Dispensary Report Form No. 12, the State Register of Persons with Disabilities, and the Electronic Register of Inpatient Patients. Dispensary indicators were interpreted as registered healthcare-utilisation measures, not population prevalence or incidence. Trends were assessed using Mann–Kendall tests with Benjamini–Hochberg false-discovery-rate correction. Regional patterns were examined using Spearman correlation, inequality metrics, and exploratory clustering. Results: Among adults, registered H60–H95 annual morbidity increased from 231,209 to 278,897 cases (CAGR +3.8%; FDR p = 0.016), and first-time registrations rose from 126,513 to 167,674 (CAGR +5.8%; FDR p = 0.016). Active dispensary observation declined modestly from 21,840 to 19,881. H65–H66 chronic otitis first-time registration remained stable, while active observation declined sharply. Paediatric H90–H91 registration was stable. In 2025, 31,560 persons were registered with hearing impairment-related disability, including 26,522 adults. CI procedures increased from 285 to 523. Adult CI activity was intermittent between 2012 and 2018 (peak 140 in 2013), ceased in 2019–2020, and was re-established from 2021 (14 → 129 procedures in 2025), reaching 24.7% of all CI. Most procedures were concentrated in Almaty and Astana. Compared with the GBD-2021 estimate of 17,212 per 100,000, administrative indicators captured much narrower service-utilisation and severe-certification layers. Conclusions: Kazakhstan’s administrative data describe distinct hearing-health layers rather than competing prevalence estimates. Adult screening, regional audiology capacity, post-CI rehabilitation, and longitudinal outcome registries are priority areas. Full article

With the growing global emphasis on data privacy protection, particularly the enforcement of the “right to be forgotten” under the GDPR, effectively deleting private information from models has become an urgent challenge. Graph-structured data presents a particularly challenging unlearning scenario due to its non-Euclidean nature and strong relational dependencies, which are prevalent in real-world applications such as social and recommendation systems. To address this issue, graph unlearning has been introduced to eliminate the influence of deleted data on models while preserving their overall performance. The effectiveness of graph unlearning is typically evaluated by three key metrics: model performance, unlearning efficiency, and robustness against membership inference attacks, which together determine the overall quality of an unlearning method. Existing graph unlearning methods fall into exact and approximate regimes. Most studies focus on edge/node-level unlearning, and existing attempts at feature-level unlearning remain limited. Exact unlearning methods that adopt the SISA partition and retraining paradigm may inadvertently reintroduce the features intended to be unlearned during the aggregation phase, thereby leading to incomplete unlearning. Approximate methods, on the other hand, often incur excessive information loss in feature-level removal, which degrades predictive accuracy. Accordingly, we propose a graph unlearning framework specifically designed for feature-level unlearning, consisting of two main stages. In the first stage, we zero out the features of the unlearned nodes at each layer to block their propagation through the GNN, thereby reducing their influence on neighboring node representations. In the second stage, we induce misclassification of the unlearned nodes to progressively degrade model representations and learned knowledge associated with them, enabling more thorough feature-level unlearning. Experiments on multiple graph datasets and models demonstrate that our method achieves favorable overall unlearning performance in most settings, offering a balanced trade-off between accuracy, unlearning efficiency, and unlearning effectiveness. Full article

Evaluation of generative models for RF (Radio Frequency) signals remains largely ad hoc, with existing approaches relying on uncalibrated metrics imported from computer vision without systematic justification or baseline establishment. We present a multi-dimensional, distribution-level evaluation framework comprising ten metrics across six layers, calibrated against real-real baselines. A real-real baseline is computed by splitting authentic signals into two independent subsets and measuring the same metric between them; the resulting value sets the achievable ceiling for that metric, against which generated-vs-real scores are normalized. The framework is grounded in four design principles: distribution-level aggregation, real-real baseline normalization, signal-to-noise ratio (SNR)-modulation stratification, and multi-dimensional coverage. Application of the framework reveals two previously unreported phenomena. First, linear Short-Time Fourier Transform preprocessing creates a gradient imbalance between frequency-domain and time-domain loss components that causes catastrophic generation failure for analog amplitude modulation; logarithmic compression resolves this. Second, the widely adopted noise-prediction training objective exhibits systematic gradient suppression for amplitude-modulated signals due to SNR-dependent implicit loss weighting; switching to signal-prediction substantially improves temporal structure fidelity for amplitude-modulated signals, with only modest trade-offs for digital communication signals. The framework and calibration methodology establish reproducible standards for comparative assessment of RF generative models. Full article

Cloud removal is an indispensable preprocessing step in optical remote sensing. Reconstructing cloud-free imagery by combining multi-temporal optical observations with cloud-penetrating synthetic aperture radar (SAR) has become a mainstream approach. However, the existing studies mostly adopt simple composites, such as per-pixel least-cloudy selection or the temporal median, as baselines, and average accuracy metrics over entire scenes; together, these two practices may overstate the true gains of deep-learning methods. This paper proposes a temporal cross-modal cloud removal method (TCM-CR). In a multi-temporal sequence, the acquisition with the lowest cloud fraction retains true surface reflectance at its cloud-free pixels and is itself a high-accuracy baseline. TCM-CR exploits this baseline in two ways. First, on clear and light inputs, cloud-free pixels are taken unchanged from the reference image, so the true reflectance is preserved without loss, independent of training. Second, only cloud-covered pixels receive a bounded correction, in which SAR supplies the surface structure beneath clouds and multi-temporal observations are integrated along time while suppressing heavily clouded acquisitions. Experiments on the SEN12MS-CR-TS dataset show that TCM-CR maintains accuracy on par with the reference image on clear and light samples and improves the peak signal-to-noise ratio on heavy samples by 7.93 dB. In a cross-region experiment where one region is excluded from training entirely and used only for testing, heavy samples still improve by 7.27 dB. Full article

Timely identification of company financial risks is crucial for investors and regulators. However, existing studies overlook the class imbalance caused by the scarcity of high-risk samples, and the interpretability of deep models is insufficient, making it difficult to meet the practical needs. To address these problems, we propose a classification model named FA-LSTFormer, which models a company’s financial risk as low-, medium- and high-warning tasks. FA-LSTFormer employs LSTM and Transformer to decouple the short-term continuity and long-term dependency inherent in financial data. To further attend to indicator-level nuances, we incorporate a Risk-Sensitive Hierarchical Indicator Attention (RSHA) module. Moreover, given the pronounced class imbalance where high-risk events are substantially underrepresented, we further propose a Class-Imbalance-Aware Focal Loss (CIFL) function to prioritize these minor yet critical samples and suppress false negatives. On the dataset of Chinese A-share manufacturing listed companies, experimental results show that our FA-LSTFormer achieves superior performance in accuracy, precision, recall, F1-score and AUC, achieving 92.76%, 93.13%, 91.84%, 92.48%, and 95.27%, respectively. Compared to the suboptimal LTR-Net, it improves these metrics by 1.64–3.60%. Compared to the LSTM–Transformer baseline, FA-LSTFormer improves on it by 4.30–9.13%. In the risk-oriented decision evaluation, FA-LSTFormer achieves a warning ROC of 0.954 for the high-risk class and lowers the error rate to 9.82%. It maintains an accuracy rate of 84.46% even after three years of early warning and exhibits strong robustness across different warning thresholds and company sizes. These results verify the advantages of FA-LSTFormer in both algorithmic performance and practical early-warning applications. Full article

Chinese Spelling Correction (CSC) is a fundamental task in Natural Language Processing (NLP) aimed at identifying and correcting character errors in Chinese texts. It significantly enhances text readability and semantic accuracy. Most deep learning-based CSC methods focus on isometric correction, ensuring identical lengths for input and output sequences. However, they struggle with variable-length errors like splitting errors—where a single character is incorrectly divided into two (e.g., splitting “明” into “日” and “月”). These errors are challenging because they disrupt token alignment, preventing standard sequence-labeling models from mapping inputs to outputs effectively. To overcome this limitation, we propose KSEC (Knowledge-enhanced Splitting Error Corrector), a novel framework tailored for variable-length corrections. KSEC automatically constructs a splitting character knowledge base from public corpora to provide factual validation for correction outcomes. Furthermore, we design a variable-length architecture integrating an attention mechanism and introduce an alignment-aware loss function that optimizes sequence-to-sequence token mapping. Extensive experiments on standard CSC and CSEC benchmarks demonstrate that KSEC achieves state-of-the-art performance among lightweight models of similar size and outperforms existing methods across multiple evaluation metrics. Full article

Urban traffic noise poses a critical environmental health burden; however, its mitigation via urban greenery is constrained by the limitations of conventional two-dimensional (2D) metrics. To address this gap, the Landscape Forest Space Density (LFSD), as a novel volumetric index quantifying the ratio of acoustically effective biomass to total canopy volume via geometric crown approximation, is introduced in this study. Field investigations were conducted along a 30 m transect perpendicular to a six-lane expressway in Nanjing, China. A-weighted sound pressure levels were recorded at 18 nodes (3 forest types × 4 distances, plus controls) under controlled meteorological conditions, with seasonal replication during spring (full-leaf) and autumn (leaf-off). Results indicate that deciduous coniferous forests delivered superior noise attenuation, achieving a peak insertion loss of 27.16 dB at 20 m depth in spring—exceeding evergreen broadleaf configurations by over 10 dB. Noise reduction followed a logarithmic growth pattern relative to roadside greenbelt width, reaching saturation beyond 20 m. Regression analysis identified LFSD as the dominant predictor of insertion loss (R² = 0.86, N = 864), substantially outperforming traditional 2D metrics such as canopy coverage (R² = 0.15) and optical porosity (R² = 0.42). While a bivariate model incorporating both LFSD and porosity yielded an adjusted R² of 0.89–0.94, the marginal incremental gain confirms that volumetric biomass packing, rather than planar void fraction, governs acoustic attenuation. Notably, even during defoliation, high-LFSD conifers maintained robust sound scattering through their fractal branch architectures. On the whole, a quantitative design threshold of LFSD ≥ 0.4 is recommended to achieve >15 dB attenuation in spatially constrained urban environments. Full article

Background and Objectives: Persistent COVID-19 sequelae, often referred to as “long COVID”, still constitute a major global medical challenge. In particular, general fatigue is the most widely reported, serious symptom. In the framework of Japanese traditional (Kampo) medicine (JTM), post-infection fatigue is frequently considered a “Qi deficiency”, a significant loss of vital energy. In JTM, accurately diagnosing Qi deficiency is essential for determining the appropriate treatment. However, one challenge in JTM diagnosis is that mastering traditional diagnostic techniques takes time. We sought to validate the “Qi deficiency score”, a quantitative patient-reported outcome measure designed to assess Qi deficiency in patients suffering from long COVID. Materials and Methods: This was a methodological study. We conducted a cross-sectional study of 237 patients who sought treatment at the post-COVID-19 clinic (COVID-19 Aftercare Clinic) at Okayama University Hospital. The patient population was randomly split into two cohorts: one for Exploratory Factor Analysis (EFA, n = 122) and another for Confirmatory Factor Analysis (CFA, n = 115). The simple Qi-deficiency score comprises eight items derived from Terasawa’s diagnostic criteria, with a total range of 0–56. We evaluated internal consistency using Cronbach’s α/McDonald’s ω and examined structural validity. Convergent validity was established by examining correlations between the simple Qi deficiency score and the Fatigue Assessment Scale (FAS), the Self-rating Depression Scale (SDS), and Quality of Life (QOL) metric. Results: The score’s internal consistency (α = 0.64 and ω = 0.63) was considered modest for screening traditional clinical syndromes. EFA showed major factors representing Qi deficiency symptoms. CFA supported a one-factor model with modest fit indices (CFI = 0.87; RMSEA = 0.079; SRMR = 0.077; TLI = 0.82). Factor loadings revealed that “Body feels tired (now)” (0.98) and “Tires easily (tendency)” (0.73) were the most significant symptoms. Convergent validity showed a strong positive correlation with the FAS (r = 0.61, ρ = 0.60) and a moderate correlation with the SDS (r = 0.55, ρ = 0.55). A significant negative correlation was found with QOL scores (r = −0.42, ρ = −0.43). Conclusions: The simple Qi deficiency score may be useful for evaluating Qi deficiency in patients with post-COVID-19 condition. Full article