Search Results (45,777)

Background: Artificial intelligence (AI) is becoming increasingly integrated into orthopaedic surgery for tasks such as implant positioning, dislocation risk prediction, and surgical decision-making. However, the current evidence varies widely across anatomical regions and applications. Methods: A structured narrative review was conducted using PubMed and Web of Science Core Collection to identify studies applying machine learning or deep learning in orthopaedic procedures, focusing on parameters such as the anatomical region addressed, data types used, primary AI tasks, evaluation designs, and validation strategies. Reviews and meta-analyses were excluded. Study selection was summarized using a PRISMA-style flow diagram, and included studies were narratively synthesized according to anatomical region, AI task, imaging modality, validation strategy, and clinical relevance. Results: We identified three main application areas: (1) AI in imaging-driven planning and implant positioning, often linked with navigation or robotic systems; (2) postoperative evaluation related to implants; and (3) prediction of clinically relevant outcomes such as dislocation risk. The strongest evidence is found in hip arthroplasty, where AI improves measurement accuracy and workflow efficiency, whereas applications in knee, shoulder, and spine surgery are less developed and often supported by smaller studies. Although existing risk prediction models demonstrate good performance, their generalizability is hindered by limited external validation and inconsistent reporting. Conclusions: Overall, while AI shows significant promise in enhancing various aspects of orthopaedic surgery, stronger links between technical advancements and patient outcomes are needed. Future research should prioritize extensive validations, workflow-aware evaluations, failure analysis, and adherence to AI-specific reporting guidelines to facilitate safe and effective clinical implementation. Full article

Object detection in UAV remote sensing imagery is confronted with three primary challenges: severe scale variation, densely clustered small targets, and constrained computational resources. This work introduces a family of lightweight detection models guided by the “Capacity-Aware Configuration Regularity” and incorporates a Feature-Refinement C2f module to enhance representational efficiency. A dynamic coupling mechanism is identified between detection head capacity and the representational quality of Backbone features, which is further validated through systematic ablation studies spanning three parameter magnitudes. Evaluated on the VisDrone2019 benchmark, the proposed model family exhibits a progressive parameter scaling from 1.67 M to 6.15 M. The nano variant achieves 31.7% mAP₅₀ using only 55% of the parameter budget of YOLOv8n, surpassing it by 0.7 percentage points. The small variant, with a parameter budget comparable to YOLOv8n, attains 36.7% mAP₅₀, exceeding it by 5.7 points. The medium variant reaches 43.1% mAP₅₀ with 58% of the parameters of YOLOv8s, outperforming it by 4.1 points. The improvements are pronounced under the stricter mAP_50–95 metric, where the small variant outperforms YOLOv8n by 3.3 points and the medium variant surpasses YOLOv8s by 2.8 points, demonstrating robust localization accuracy across a wide range of IoU thresholds. This consistent superiority in the accuracy–efficiency trade-off extends to the DIOR dataset, confirming the robust generalization of the proposed models across diverse remote sensing scenarios. Moreover, the uncovered capacity-matching regularity offers transferable methodological guidance for designing lightweight detection models tailored to resource-constrained platforms. Full article

Fine-grained insect pest classification presents a particularly demanding visual recognition challenge due to severe class imbalance, pronounced intra-class morphological variability across developmental stages, and high inter-class visual similarity among taxonomically related species. Existing deep learning approaches typically rely on a single feature representation extracted from a single network depth, overlooking complementary discriminative cues distributed across multiple abstraction levels. Furthermore, classical attention mechanisms perform spatial weighting deterministically, without explicitly modeling the underlying statistical structure of the feature space, which is inherently multimodal on long-tailed benchmarks such as IP102. This study proposes a Multi-Scale Gaussian Mixture Model-Gated Mixture of Experts (GMM-MoE) architecture that operates as a plug-in module insertable into any convolutional backbone, evaluated here on DenseNet-121 at three distinct feature depths. The proposed module computes analytic GMM posterior responsibilities in closed form, softly assigning each spatial location to dedicated convolutional expert sub-networks. At the same time, a conditional prior mechanism π(x) adapts the routing strategy to individual image content rather than relying on fixed priors. The architecture is evaluated on the IP102 benchmark (102 pest classes, ~75,000 images) under a two-stage training protocol. Ablation experiments confirm that increasing the number of experts consistently improves accuracy across all three routing depths, and that multi-scale fusion surpasses any single-scale configuration. The proposed model achieves a mean top-1 accuracy of 74.12% (±0.25%, 95% CI) across three independent runs on the IP102 test set. To the best of our knowledge, this is the first work to employ GMM posterior responsibilities as a spatial routing mechanism within a multi-scale CNN feature hierarchy for fine-grained insect pest classification, establishing a principled probabilistic alternative to deterministic attention weighting in visual recognition systems. Full article

Background: Accurate identification and positioning of thoracic devices on chest radiographs is critical for patient safety in intensive care. Multimodal large language models (LLMs) offer potentially generalizable automated evaluation, but their performance in this domain is underexplored. Methods: Three multimodal LLMs (GPT-4o, gpt-4o-2024-08-06; Gemini 3.1 Flash Lite Preview; Claude Sonnet 4.6) were evaluated on 4813 chest radiographs from the RANZCR CLiP dataset for device presence and positioning of ETT, NGT, CVC, and Swan–Ganz catheters. Performance was quantified with 95% Wilson confidence intervals, balanced accuracy, MCC, Cochran’s Q, Bonferroni-corrected McNemar, and Cohen’s/Fleiss’ kappa. Six additional analyses were performed: a blinded paired reader study (n = 377; two board-certified radiologists, blinded to ground truth and to all LLM outputs), external validation on PadChest (n = 200, device-presence detection only—PadChest lacks granular position labels), three-variant prompt-sensitivity analysis (n = 103), repeat-inference stability across three runs (n = 50), systematic error taxonomy, and a failure-case analysis. Results: Device-presence performance varied widely across models; abnormal-position sensitivity was uniformly poor (MCC ≤ 0.028; balanced accuracy 0.41–0.53). Inter-model agreement was poor to slight (Fleiss’ κ: 0.005–0.383 for presence; −0.280 to −0.025 for classification). Radiologists numerically outperformed all three LLMs in 42/42 paired comparisons; the superiority was statistically significant after Bonferroni correction in 33/42 (32/42 at p < 0.001). PadChest replicated the negative finding for device-presence detection (malposition not externally validated). Prompts and inference stochasticity introduced 2–3× sensitivity swings and run-to-run κ from 0.20 to 0.85. Case failures concentrated systematically in multi-device cases (p < 0.0001) but not in abnormal-position cases (p = 0.14). Conclusions: Current general-purpose multimodal LLMs are not yet reliable for autonomous thoracic-device assessment; their failure patterns are structurally characterizable across models, prompts, and case types and support, at most a circumscribed role, as adjunct device-presence screening tools. The findings do not generalize to purpose-built, regulator-approved clinical AI systems. Full article

Traditional visual SLAM pipelines are typically designed under the static-world assumption and often degrade severely in indoor environments with frequent human motion. To improve trajectory accuracy and front-end stability in such scenarios while maintaining real-time throughput, we present SY-SLAM, an RGB-D SLAM system for dynamic indoor environments with frequent human motion. (S stands for SuperPoint, which is used as a detector-only learned keypoint front-end, and Y stands for YOLO, which provides asynchronous person-aware keypoint suppression based on detected human bounding boxes.) We integrate a TensorRT-deployed detector-only SuperPoint module to improve keypoint repeatability and robustness while retaining ORB binary descriptors for efficient matching and place recognition within the ORB-SLAM3 framework. To avoid feature starvation while preserving keypoint quality, we further introduce an adaptive SuperPoint keypoint selection strategy that applies stricter filtering when keypoints are abundant and relaxes the selection constraints when they are scarce. In parallel, an asynchronous YOLOv8s TensorRT thread performs person detection with temporal bounding-box memory, and keypoints inside detected person regions are removed before ORB descriptor computation and matching to reduce dynamic-feature contamination in the front end. We evaluate SY-SLAM on five dynamic TUM RGB-D fr3 sequences using ATE and RPE metrics. Compared with ORB-SLAM3, SY-SLAM reduces ATE RMSE by 93.45% across four dynamic walking sequences. On the widely reported fr3/w/x sequence, SY-SLAM achieves competitive accuracy with recent dynamic SLAM methods while maintaining real-time performance. The system runs in real time at 46.8 Hz (21.36 ms per frame) on an Intel i9-13900H CPU with an NVIDIA RTX 4070 Laptop GPU. Full article

Introduction: Accurate detection and characterization of intracardiac masses remain a major challenge in cardiovascular imaging due to overlapping morphological features between tumors, thrombi, and vegetations, as well as the inherent limitations of echocardiography, including operator dependency and variable image quality. Although echocardiography is the first-line imaging modality for evaluating cardiac masses, diagnostic uncertainty frequently necessitates additional multimodality imaging. Artificial intelligence (AI), including machine learning and deep learning approaches, has emerged as a promising strategy to improve image interpretation, automate feature extraction, and enhance diagnostic consistency. Objective: This narrative review aims to examine current advances in AI-enhanced echocardiography for cardiac tumor detection, with a particular focus on detection, segmentation, classification, multimodal integration, and clinical translation. Methods: A narrative literature review was conducted using PubMed, Scopus, and Google Scholar databases. Relevant English-language studies published between 2016 and 2026 were identified using keywords including “artificial intelligence”, “machine learning”, “deep learning”, “echocardiography”, “cardiac tumors”, “intracardiac masses”, “multimodal imaging”, and “ultrasomics”. Original studies, reviews, and methodological papers related to AI-assisted cardiovascular imaging were evaluated. Discussion: Current evidence suggests that AI-driven techniques, including radiomics (ultrasomics), convolutional neural networks, and multimodal learning frameworks, can improve the detection, segmentation, and classification of intracardiac masses. Experimental studies have reported high diagnostic performance, with some deep learning models achieving diagnostic accuracies exceeding 95% under controlled conditions. AI-assisted systems may also reduce interobserver variability and improve workflow efficiency. Multimodal AI approaches integrating echocardiography with cardiac magnetic resonance imaging, computed tomography, electrocardiography, and clinical data appear particularly promising for improving diagnostic discrimination. However, current models remain limited by small and imbalanced datasets, insufficient external validation, data heterogeneity, and limited generalizability across institutions and imaging protocols. Additional barriers to clinical implementation include annotation variability, limited interpretability of deep learning models, and regulatory considerations. Conclusions: AI-enhanced echocardiography has substantial potential to improve the detection and characterization of intracardiac masses by augmenting diagnostic consistency and supporting clinical decision-making. Nevertheless, current evidence remains largely based on retrospective and experimental studies. Future progress will depend on large multicenter collaborations, standardized imaging datasets, explainable AI frameworks, and prospective clinical validation to enable safe and effective integration into routine cardiovascular practice. Full article

Background/Objectives: Capsular contracture (CC) is the most frequent complication of breast implant surgery, affecting up to 20% of augmentation and up to 40% of post-mastectomy reconstruction patients. Diagnosis relies on the Baker classification with poor interobserver reliability (κ = 0.55). This study evaluated shear-wave elastography (SWE) as an objective diagnostic tool for CC via quantitative measurement of periprosthetic capsule stiffness. Methods: A prospective single-center comparative study (Romania) enrolled 26 augmentation patients (50 breasts) with asymptomatic Baker I/II CC as controls, and 25 breasts with confirmed Baker III/IV CC in post-mastectomy reconstruction patients as the study group. Stiffness was measured using the SuperSonic MACH 30 platform (mean, median, min, max, SD in kPa). Analysis included Mann-Whitney U tests, ROC curves with bootstrapped 95% CIs, and Youden’s J index. Confounder analyses (Spearman correlations, multivariable logistic regression, partial correlations) assessed the independence of SWE findings from implant depth, periprosthetic tissue thickness, region-of-interest (ROI) diameter, and body mass index (BMI). Results: All four primary stiffness parameters differed significantly between groups (p < 10⁻¹¹, r > 0.97). Control median stiffness was 32.6 kPa versus 138.0 kPa in the study group. All four parameters achieved outstanding discriminative performance (AUC 0.988–0.994); SWE median yielded the highest AUC (0.994; 95% CI 0.980–1.000). A threshold of 82 kPa provided 100% sensitivity, 98% specificity, and 100% NPV. Baker Grades III (~92 kPa) and IV (~147 kPa) also differed significantly (p = 0.0001). No covariate (implant depth, periprosthetic tissue thickness, ROI diameter, BMI) significantly influenced SWE values within either group (all intra-group Spearman ρ p > 0.05), and SWE median stiffness remained the sole significant predictor in the fully adjusted multivariable model (adjusted OR = 1.18, 95% CI 1.08–1.31, p < 0.001). Conclusions: SWE objectively differentiates normal periprosthetic capsules from clinically significant CC with outstanding accuracy. An 82 kPa median stiffness threshold offers a reproducible, non-invasive complement to the Baker classification and provides a foundation for elastography-based CC staging. Full article

Aspect-Based Sentiment Analysis explores sentiment trends related to specific opinion aspects and holds significant commercial potential for monitoring brand reputation, understanding customer satisfaction, and personalizing recommendations. However, traditional methods rely exclusively on textual input and often struggle when the target aspect is not mentioned in the sentence. Multimodal Aspect-Based Sentiment Analysis addresses this limitation by incorporating both textual and visual modalities to enable more comprehensive sentiment understanding. Despite advancements in deep learning and transformer-based architectures, existing models often suffer from suboptimal modality fusion and weak aspect grounding, limiting their classification accuracy. To overcome these challenges, we propose an Attention-Driven Feature Fusion (ADFF) approach based on a three-stage hierarchical attention mechanism. First, it only fuses text and image embeddings. Second, it incorporates aspect-level features. Third, a multi-head attention layer further enhances cross-modal dependencies. The resulting representation is passed to a Long Short-Term Memory (LSTM) classifier for sentiment polarity prediction. We evaluate our model on three benchmark datasets, namely Twitter-2015, Twitter-2017, and MASAD. The experimental results demonstrate that the proposed model substantially outperforms state-of-the-art multimodal and unimodal baselines, improves both accuracy and F1-score, achieving 82.55% accuracy and 81.05% F1-score on Twitter-2015, 77.07% accuracy and 77.15% F1-score on Twitter-2017, and up to 99.67% accuracy and F1-score in the Plant domain of MASAD, where we observe consistent improvements across all seven domains. These results highlight the effectiveness and scalability of the hierarchical attention-based fusion strategy for real-world aspect-based sentiment analysis tasks. Full article

Background: Vestibular migraine (VM) is a common cause of episodic vertigo, yet its diagnosis remains primarily clinical and is often complicated by the absence of reliable objective biomarkers. Pupillary nystagmus, reflecting spontaneous oscillations of pupil diameter, has been proposed as a potential clinical sign of VM, but its quantitative characterization remains limited. Objective: The objective of this study is to evaluate the diagnostic value of pupillary nystagmus in VM and to provide a quantitative assessment using infrared pupillometry. Methods: In this case–control study, 137 patients with vestibular migraine and 102 healthy controls underwent comprehensive neuro-otological evaluation, including vestibular testing and pupillometric assessment. Pupillary activity was recorded using a dedicated infrared pupillometer, and oscillatory dynamics were quantified using the Pupillary Unrest Activity Level (PUAL), which was derived through spectral analysis (Larson–Neice algorithm). Statistical comparisons were performed using non-parametric methods. Results: PUAL values differed significantly between VM patients and controls (Mann–Whitney test p < 0.001), demonstrating a clear separation between groups. A cut-off value of 0.325 was identified as the upper limit of normality, suggesting that elevated PUAL values may indicate vestibular migraine. Conclusions: Pupillary nystagmus represents a clinically accessible sign that can be objectively quantified through infrared pupillometry. The PUAL index provides a measurable parameter reflecting altered vestibulo–autonomic dynamics in VM and may serve as a promising neuro-otological biomarker. The integration of pupillometric analysis with clinical evaluation may improve diagnostic accuracy and support the development of objective diagnostic tools in vestibular migraine. Full article

The abdominal pressure–volume (P–V) relationship during laparoscopic insufflation is curvilinear and subject to substantial inter-individual variability, yet clinical practice relies on universal pressure targets derived from population-level guidelines. The Smart Inline Compliance Module (SICM) is a novel inline retrofit device that acquires intra-abdominal pressure and insufflation gas flow through physically separated sensing circuits, reconstructs insufflated volume by numerical integration of the flow signal, and derives the abdominal P–V curve and its biomechanical parameters in real time. This study reports the first two-arm pilot technical evaluation of the SICM system. Arm A comprised an exploratory biomechanical phantom with three defined stiffness levels (Soft, Medium, Rigid) tested under Continuous and Stepwise insufflation protocols (30 curves). Arm B comprised three female feline cadavers assessed under the same dual-protocol design (18 curves). This study should be interpreted as an early-stage technical evaluation rather than as a definitive validation benchmark. Signal quality was consistently high across both arms (Curve Quality Index: 1.0000 in the phantom arm; 0.9974 ± 0.0009 in the cadaveric arm). Volume integration accuracy was confirmed against an independent offline reference (mean absolute percentage difference: 0.07%). The system extracted reproducible biomechanical parameters under the Continuous protocol: in the cadaveric arm, maximum compliance (Cmax) ranged from 116.8 to 191.4 mL/mmHg across subjects, with intra-session coefficients of variation below 16%; Knee Pressure (Pknee), defined as a working operational index of the compliance transition, was 3.33–4.17 mmHg with CV below 8%. The Rigid phantom and cadaveric datasets showed partial numerical overlap in selected shape-derived parameters, which was interpreted only as an internal consistency check and not as evidence of biomechanical equivalence. The Stepwise protocol exposed the current methodological limits of the parameter-extraction workflow and identified specific targets for the next development iteration. These results are interpreted exclusively within the scope of technical feasibility and preliminary biomechanical characterisation; clinical applicability and optimal pressure guidance require adequately powered in vivo studies. Full article

Chunmee tea quality is critical to its grading, and accurate identification is essential for quality evaluation and market valuation. However, traditional machine learning relies on manual feature extraction and causes spectral information loss, while conventional one-dimensional convolutional neural networks (1D-CNNs) are restricted by fixed kernels and narrow receptive fields, making multi-scale feature capture difficult. In this study, an improved DeepSpectra model integrated with the Inception module and residual connections was proposed for end-to-end automatic grading of Chunmee tea. A total of 360 samples across six grades (60 samples per grade) were collected using an Antaris II near-infrared spectrometer and preprocessed by multiplicative scatter correction (MSC). The proposed model was compared with other models. Results showed that under a 7:1:2 train–validation–test split, the proposed DeepSpectra achieved an average test accuracy of 96.39 ± 1.63% across ten random sample divisions, significantly outperforming the other models (p < 0.05). The model also exhibited excellent stability in five-fold cross-validation and superior generalization in small-sample scenarios, and a lightweight structure with low inference latency of 2.2 ms, which is suitable for real-time industrial applications. This work provides a reliable, efficient, and end-to-end method for grading Chunmee tea and offers a promising strategy for intelligent and rapid quality control of green tea. Full article

The aim of this study is to address the limitations of fixed parameters and poor adaptability in traditional damage models for damage assessment of reinforced concrete (RC) columns under variable axial load. An improved damage model considering the influence of variable axial load was proposed herein. Based on quasi−static tests of RC columns under variable axial load, a fiber finite element model was established, and its reliability was verified using experimental data. The limitations of classical damage models were systematically analyzed, and the quantitative relationship between core parameters and axial load ratio was derived via numerical simulation of multi−level axial load ratio working conditions, on the basis of which the traditional model was modified. The applicability of the improved model was evaluated through full factorial combination working conditions, and the quantitative correlation among damage indices, stiffness degradation, and load−bearing capacity degradation was established. The results indicate that the improved model addresses the limitation of fixed parameters of traditional models, maintains stable calculation accuracy for circular RC columns under the investigated ranges of axial load ratio, shear−span ratio, and reinforcement ratio, and enables quantitative prediction of mechanical properties based on the damage index. Full article

The rapid advancement of unmanned aerial vehicles (UAVs) has greatly increased the application of various swarm intelligence algorithms in UAV path planning. To address the potential issues with the dung beetle optimizer (DBO) in UAV trajectory planning, such as low convergence accuracy, tendency to get trapped in local optima, and imbalance between global search and local exploration, a hybrid algorithm termed DBO-PSO is proposed by integrating DBO with particle swarm optimization (PSO) to solve the UAV path planning model. The Kent chaotic map is introduced to enhance population diversity and distribution uniformity, and the velocity–position update mechanism of PSO is incorporated into DBO to strengthen its global search capability. Comparative experiments are conducted on CEC2022 benchmark functions, and multiple classical swarm intelligence algorithms are selected for comparison using six evaluation metrics, along with Wilcoxon rank-sum and Friedman statistical tests. An ablation study is also performed to evaluate the contribution of each improvement component. The path planning experimental results demonstrate that compared to DBO, PSO, IDBO, and ECFDBO under the population size of 50, DBO-PSO reduces the total path cost by 44.2%, 17.3%, 8.9%, and 45.1%, respectively. The ablation study verifies that both improvement components contribute positively, which demonstrates its competitive performance and practical applicability in UAV three-dimensional path planning. The source codes to support the presented results are publicly available on GitHub. Full article

Visual explainable artificial intelligence (XAI) is an important mechanism for connecting analytically complex decision models with practitioners who must interpret and act upon their outputs in industrial supply chains. In facility localization problems, wafer foundries and other capital-intensive manufacturers must evaluate geographically dispersed candidate sites against multiple uncertain criteria. The ability to communicate fuzzy group decision-making (FGDM) outcomes in a transparent, interpretable form has direct operational relevance. The literature has introduced hanging gradient bar charts, gradient bidirectional scatterplots, and traceable aggregation charts as visual XAI instruments for semiconductor supply chain localization that show substantial reductions in interpretation error versus conventional plots. However, the quantitative assessment of explanation quality itself remains underdeveloped. To address such a gap, this research proposes a quantitative explainability quality index (XQI) that formalizes visual explanation quality in FGDM as a composite measurable construct. XQI integrates two complementary layers: (1) An objective explainability layer (OEI), consisting of normalized fuzzy interpretation deviation, response time, ranking fidelity, and interpretation accuracy, and (2) a subjective explainability layer (SEI), consisting of perceived understanding, perceived transparency, decision confidence, and cognitive load. Trust, acceptance, and decision quality are downstream outcome constructs rather than components of the index. A weighted linear combination of OEI and SEI produces a single index for systematic, reproducible comparison across competing visualization designs. A structural equation model is specified as a planned validation mechanism for examining how explanation quality may relate to trust, acceptance, and downstream decision quality. The proposed validation framework includes a semiconductor facility localization scenario, three visualization conditions, and a planned participant pool of 150–240 supply chain managers, engineers, and graduate students. The XQI framework transforms visual XAI from a descriptive communication aid into a testable decision-support construct, thereby addressing a key evaluation gap in the FGDM visualization literature. Full article

Transfer accuracy is widely used to evaluate orthodontic indirect bonding workflows, particularly in the context of digital CAD/CAM planning and three-dimensional bracket positioning. However, substantial heterogeneity in its definition, measurement, and reporting may limit comparability and clinical interpretability across systematic reviews. This methodological umbrella review examined how transfer accuracy is operationalized as an outcome construct, with specific focus on conceptual definitions, dimensional frameworks, reference systems, measurement pipelines, and interpretative strategies rather than pooled quantitative deviation estimates. A systematic search of major biomedical databases was conducted to identify systematic reviews evaluating transfer accuracy in orthodontic indirect bonding. Data extraction was performed independently by two reviewers using a predefined methodological mapping framework, and methodological quality was assessed with AMSTAR-2. Four systematic reviews met the inclusion criteria. Across reviews, transfer accuracy was operationalized through heterogeneous linear and angular geometric deviation metrics derived from planned–achieved bracket position comparisons, without use of a standardized composite accuracy indicator. Nevertheless, substantial heterogeneity was found in outcome definitions, dimensional architectures, reference system selection, and analytical workflows, resulting in structurally non-equivalent representations of transfer accuracy and limiting cross-review comparability. Within the included systematic reviews, transfer accuracy functioned primarily as a workflow-dependent geometric measurement construct rather than as an outcome systematically operationalized within clinically validated frameworks. We recommend standardized construct definitions, mandatory reporting of reference systems and registration algorithms, routine uncertainty quantification, and harmonized dimensional frameworks as essential steps toward valid evidence synthesis, reproducible digital orthodontic workflows, and clinically interpretable transfer accuracy measurement. Full article