搜索结果：splitting

Human Activity Recognition (HAR) and exercise assessment models are increasingly used in healthcare to support clinical evaluation, rehabilitation, and remote monitoring. However, their real-world applicability critically depends on the ability to generalize across unseen subjects, whose movement patterns may differ substantially due to inter-individual variability. Despite this, many studies adopt random noncross-subject (NCS) data splits, where samples from the same individual appear in both training and test sets, potentially leading to overly optimistic and clinically misleading performance estimates. We investigate (i) how NCS and cross-subject (CS) splits affect performance estimation across machine learning and deep learning models under tasks of increasing complexity, (ii) how data splitting and differences between training and test sets contribute to predictive variance and stability. Experiments were performed using a large-scale HAR benchmark dataset (NTU RGB+D 120) and a rehabilitation-specific dataset (IntelliRehabDS). A total of 12 machine learning and deep learning models were trained across both tasks, and their performance was estimated and compared using a simulation-based approach. Predictive variance decomposition, via Generalized Linear Mixed-Effects models, was applied to link the split strategy and differences in training and test instances to model output stability. NCS splits consistently overestimated model performances, with discrepancies increasing alongside task and model complexity. DL architectures, in particular, showed markedly higher NCS performance compared to CS splits, generally with statistical significance. Variance decomposition revealed that greater subject difference between training and test sets often enhances predictive instability, while CS splitting reduces variance by promoting more generalizable representations. Improper dataset splits can mislead model evaluation, exaggerate generalization capabilities, and undermine clinical trust. Our study provides empirical evidence for computer vision-based rehabilitation models and offers methodological guidance for robust evaluation practices, supporting reproducible and trustworthy AI deployment in rehabilitation and broader healthcare applications.

One of the major causes of death in the general population is cardiovascular disease. Life-threatening cardiac disease is influenced by several factors, including age, gender, blood sugar, cholesterol, heart rate, and more. There are so many factors that it can be challenging for specialists to assess each one. The current approach utilizes electrocardiogram (ECG) data and magnetic resonance imaging (MRI) image features but suffers from poor performance and high error rates. To address this problem, we employ an ensemble-based fuzzy multilayer neural perceptron (EFMLNP) model to predict cardiac disease. Initially, an image from the University of California, Irvine (UCI) Machine Learning Repository was selected to analyze the prognosis of cardiovascular disease. To effectively replicate the raw data values in the dataset, a median box filter (MBF) is used to pre-process the MRI dataset, reducing irrelevant values. The second stage, segmentation, uses adaptive mean gray segmentation (AMGS) to initialize two clusters for regions of interest and non-interest. The dataset is then tested using a feature-selection method based on recursive spectral spider optimization (RSSO) to identify the most pertinent characteristics for diagnosing heart disease (optimal reduced-feature splitting). Lastly, we examine a machine learning feature-extraction model and perform test analysis on the reduced features. The proposed EFMLNP method is evaluated using metrics including precision, recall, and receiver operating characteristic (ROC). The experimental outcome demonstrates that the accuracy is 98.3%, the precision is 97.15%, the recall is 98.43%, the F1-score is 96.34%, and the ROC is 0.96.

Coupling excitons with quantized radiation has been shown to enable coherent ballistic transport at room temperature inside optical cavities. Previous theoretical works employ a simple description of the material, depicting it as a one-dimensional single-layer placed in the middle of an optical cavity, thereby ignoring the spatial variation of the radiation field. In contrast, in most experiments, the optical cavity is filled with organic molecules or multiple layers of two-dimensional materials. Here, we develop an efficient mixed-quantum-classical approach, introducing a bright layer description, that enables the simulation of exciton-polariton quantum dynamics in all three dimensions. Our simulations reveal that, for the same Rabi splitting, a multilayered material extends the quantum coherence lifetime and enhances transport compared to a single-layer material. We find that this enhanced coherence can be traced to a synchronization of phonon fluctuations over multiple layers, wherein the collective light-matter coupling in a multilayered material effectively suppresses the phonon-induced dynamical disorder.

Electrocatalytic reactions play crucial roles in energy conversion and green chemical synthesis; however, their practical applications are often constrained by sluggish reaction kinetics, high overpotentials, and insufficient operational stability. In recent years, defect engineering has emerged as an effective structural regulation strategy, providing new opportunities to enhance electrocatalytic performance through the deliberate introduction of non-ideal structural features, such as vacancies, heteroatom dopants and interfacial architectures. Accumulating studies have demonstrated that defects not only modulate the electronic structure of catalysts and reaction pathways, but that their dynamic evolution under specific reaction conditions also exerts profound influence on the real working-state structure and long-term catalytic performance. Owing to central importance in green hydrogen production, water-splitting reactions serve as representative model systems for elucidating defect-structure-activity relationships. In this review, we systematically summarize recent advances in defect engineering for electrocatalysis, with particular focus on water electrolysis. Emphasis is placed on defect types and structural characteristics, construction strategies, dynamic evolution behaviors, and their typical applications, while current challenges and future research directions are also discussed, aiming to provide valuable guidance for the rational design and practical implementation of high-performance defect-engineered electrocatalysts.

Current research has not yet drawn a definitive conclusion on the impact of glucagon-like peptide-1 receptor agonists (GLP-1RAs) on the risk of osteoarthritis (OA) in patients with type 2 diabetes mellitus (T2DM). To supplement the evidence base in this field, the present study sought to systematically investigate the association between GLP-1RA administration and the risk of OA in adult T2DM patients through a meta-analysis of relevant randomized controlled trials (RCTs). An exhaustive literature search was conducted across five major databases-PubMed, Embase, Web of Science, ClinicalTrials.gov, and the Cochrane Library-covering the period from each database's inception up to May 6, 2025. The main objective of this literature search was to identify RCTs that enrolled adult patients with T2DM receiving GLP-1RAs and reported OA incidence. The Cochrane Risk of Bias 2.0 tool was employed to evaluate the risk of bias for all included studies, while the GRADE system was applied to assess the overall certainty of the evidence. Additionally, the node-splitting method was utilized to perform inconsistency testing between direct and indirect evidence. The occurrence of OA events was defined as the primary outcome of the study. A series of analyses were performed to evaluate the impact of each intervention on OA risk, including traditional meta-analysis, frequentist network meta-analysis (NMA), and sensitivity analysis. Summary estimates of effect size were presented as relative risk (RR) along with 95% confidence intervals (CI). The present study ultimately included 74 RCTs with a combined sample size of 105,415 individuals. Conventional pairwise meta-analysis revealed no statistically significant difference in OA risk between GLP-1RAs and placebo, insulin, or other oral antidiabetic drugs (OADs). Stratified subgroup analyses based on GLP-1RA formulation, treatment duration, and control group type also found no statistically significant differences in OA risk across all subgroups. Furthermore, NMA results indicated no notable variations in OA risk between GLP-1RAs and control interventions, and no such differences were observed among the different GLP-1RA formulations either. Sensitivity analyses verified the stability and reliability of the study outcomes. Among adult patients with T2DM, there was no significant difference in OA risk between GLP-1RA treatment and placebo, insulin, or OADs; this absence of statistical significance was also observed across different GLP-1RA formulations, varying treatment durations, and distinct control groups. CRD420251034203.

This study investigated exciton-photon coupling in monolayer MoS2 integrated with a nanocorrugated SiN dielectric cavity, which supports tunable guided-mode resonances near a quasi-bound state in the continuum (q-BIC). The cavity exhibited a high quality-factor (Q-factor) of up to 6300 and near-field enhancement of approximately 2202. By engineering the SiN thickness and corrugation geometry, the cavity resonance was tuned across the MoS2 A-exciton, enabling a transition from weak-coupling regime to a pronounced polaritonic regime, as indicated by the emergence of two strong transmission dips. Full-wave finite-element simulations combined with Lorentz oscillator dispersion modeling revealed clear anti-crossing behavior and narrow spectral features with a high Q-factor of approximately 340. Depending on the corrugation amplitude, a Rabi splitting of approximately 27 meV was achieved in conjunction with high-Q polariton modes, confirming a strong coupling regime. Furthermore, the curvature-induced strain introduced an additional tuning mechanism by modulating the exciton energy and detuning, thereby enabling controllable polariton dispersion while maintaining robust coupling strength. Results revealed that a nanocorrugated dielectric cavity with a facile configuration can serve as a scalable platform for strong light-matter interactions in two-dimensional materials and for designing high-Q exciton-polaritonic quantum devices.

Buprenorphine low-dose initiation (LDI) protocols can reduce the risk of precipitated withdrawal by initiating buprenorphine at low doses without needing to be in opioid withdrawal when starting. However, they are challenging for patients to implement in outpatient settings, in part because they involve splitting doses and following a multi-day regimen with dynamic daily instructions. Blister packs offer a structured approach to completing LDI by providing pre-packaged, pre-split buprenorphine doses with simplified instructions; however there is limited data of buprenorphine blister pack implementation and experiences with their use. To explore attitudes of healthcare providers and patients with outpatient buprenorphine LDI blister packing to provide insight into their use in real-world settings and inform blister-packing implementation. A qualitative study of interviews with buprenorphine prescribers, pharmacists, and patients with opioid use disorder who had attempted LDI in the past 3 months to explore attitudes about using blister packs during LDI. Participants were recruited from two safety-net substance use clinics and Community Behavioral Health Services Pharmacy in San Francisco. Interviews were transcribed, coded, and analyzed using thematic analysis to identify major themes. A total of 33 participants were interviewed (9 buprenorphine prescribers, 5 pharmacists, and 19 patients). Major themes included: 1) blister packing facilitated ease of LDI and reduced medication errors; 2) close collaborations with a public health pharmacy, along with dedicated workflows and protocols, facilitated implementation; 3) challenges with blister pack preparation and packaging were identified as barriers for further expansion to other pharmacies. Blister packing was identified as a tool to allow patients to successfully complete outpatient buprenorphine LDI. Standardized protocols and collaboration with pharmacists at a public health pharmacy were key facilitators. Partnering with local pharmacies to inform implementation strategies and identify barriers, along with pharmacist education, may help expand use of buprenorphine LDI blister-packing.

N-terminal (Nt) methionine formylation, once thought restricted to bacteria and organelles, is now recognized as a stress-inducible initiator modification in the eukaryotic cytosol. Under metabolic or environmental stress, mitochondrial methionyl-transfer RNA (tRNA) formyltransferase mislocalizes to the cytosol, generating formylated initiator tRNA (fMet-tRNAi) that initiates translation with N-formylmethionine (fMet). Nascent chains bearing Nt-fMet activate an fMet-directed ribosome-associated quality control checkpoint early in elongation, recruiting ribosome-splitting and disaggregation factors. Stalled complexes are routed to stress granules, conserving mRNA, translation machinery, and energy, while limiting aggregation. During prolonged stress, newly synthesized fMet proteins undergo maturation or selective degradation via the fMet/N-degron pathway. In mammals, E3 ligase TRIM52 acts as an Nt-fMet recognin, modulating apoptosis. Proteolytic clearance of cytosolic fMet substrates releases formylated peptides and free fMet, which are elevated in critical illness and activate formyl peptide receptors - linking translation surveillance to innate immune and inflammatory signaling in sepsis and age-related disease. Advances in N-terminomics and anti-fMet reagents now allow direct detection and quantification of cytosolic fMet proteoforms. This Review integrates bacterial and organellar paradigms with emerging cytosolic mechanisms, examines regulatory gating of Nt-formylation, and highlights therapeutic strategies to restore proteostasis and counter fMet-associated pathology.

The study aims to evaluate and rank cholinesterase inhibitors, the NMDA antagonist memantine, anti-amyloid monoclonal antibodies, and non-drug modalities with respect to cognitive outcomes, functional status, neuropsychiatric symptoms, and tolerability. We registered a protocol in PROSPERO and searched PubMed/MEDLINE, Embase, CENTRAL, Web of Science, trial registries, and gray literature through June 2025. Eligible randomized phase II/III trials in adults with clinically diagnosed AD were screened in duplicate. Data on interventions, comparators, outcomes (e.g., MMSE, ADAS-Cog, CDR-SB), and adverse events were extracted. Risk of bias was assessed using Cochrane RoB 2. A Bayesian random-effects NMA synthesized 125 trials (n > 30,000), estimating standardized Mean Differences (SMDs) with 95% Credible Intervals (CrIs). Heterogeneity (I²) and inconsistency (design-by-treatment, node-splitting) were evaluated. The network was well connected, with low-to-moderate heterogeneity (global I² = 38.5%) and no significant inconsistency (p = 0.48). Cognitive training (SMD = 0.45; 95% CrI 0.30-0.60; SUCRA 92%), aerobic exercise (SMD = 0.55; 95% CrI 0.35-0.75; SUCRA 87%), and galantamine (SMD = 0.40; 95% CrI 0.22-0.58; SUCRA 84%) ranked highest versus placebo. Donepezil (SMD = 0.21; 95% CrI 0.11-0.30; SUCRA 78%) and memantine (SMD = 0.24; 95% CrI 0.13-0.35; SUCRA 72%) showed modest benefits. Risk-of-bias ratings were low in 37% of trials, some concerns in 48%, and high in 15%. Subgroup analyses confirmed greater cholinesterase inhibitor efficacy in mild AD and superior memantine effects in moderate-to-severe disease. Non-pharmacological interventions demonstrated short-term cognitive benefits primarily in mild Alzheimer's disease populations and should be interpreted as adjunctive symptomatic strategies rather than direct substitutes for pharmacological therapy.

The oxygen evolution reaction (OER) serves as a critical electrode process for electrochemical hydrogen production, yet it is hindered by its sluggish kinetics and poor stability of catalysts. Carbides anchored on highly electronegative oxide nanoparticles are attractive new catalytic materials for OER due to the unique carbide/oxide heterostructures. Herein, we report a strong carbide/oxide interfacial electronic coupling strategy to obtain high-performance OER catalysts of Ni3Mo3C anchored on MoO2 nanoparticles (Ni3Mo3C@MoO2-CN). Ni3Mo3C@MoO2-CN catalyst exhibits excellent OER performance with a low overpotential of 197 mV at 10 mA cm-2. Furthermore, a Pt/C || Ni3Mo3C@MoO2-CN electrolyzer achieves a low cell voltage of 1.92 V at 500 mA cm-2 with 2000 h of outstanding durability for overall water splitting. Our work reveals that carbide/oxide interfacial electronic coupling induces electron transfer from Ni3Mo3C to MoO2, which facilitates π* electron back-filling effect into the Ni─O*, resulting in the optimized adsorption of O*. Consequently, the key reaction step of OOH* formation in Ni3Mo3C@MoO2-CN becomes a spontaneous process, leading to a low energy barrier of the OER process and excellent catalytic performance. This study provides fundamental insights into modulating the catalytic performance by fabricating interfacial electronic coupling between carbides and highly electronegative oxide nanoparticles.

This study aimed to develop and compare two deep learning-based segmentation-radiomics pipelines - YOLOv8-Hybrid and nnU-Net v2 - for automated sex classification and age estimation from maxillary sinus morphometry on panoramic radiographs. A balanced dataset of 1,024 panoramic radiographs (512 males, 512 females; age 18-81 years) was collected from Near East University, North Cyprus. Ground truth sinus annotations were generated by an expert oral radiologist and validated through dual-annotator inter-observer reliability assessment (ICC (2,1) = 0.94-0.97). The YOLOv8-Hybrid pipeline employed YOLOv8n-seg coarse segmentation, U-Net boundary refinement, > 120 morphometric and radiomic features, and CatBoost/XGBoost classifiers. The nnU-Net v2 pipeline used auto-configured 2D U-Net segmentation with identical feature extraction and XGBoost prediction. Both pipelines underwent 5-fold cross-validation with patient-level splitting, transfer learning, Bayesian hyperparameter optimization, and SHAP interpretability analysis. nnU-Net v2 achieved statistically significant superiority in sex classification (AUC = 0.927 [95% CI: 0.881-0.964]) over YOLOv8-CatBoost (AUC = 0.893 [0.841-0.938]; DeLong p = 0.024, Cohen's d = 0.48). Both pipelines demonstrated comparable age estimation performance (MAE ≈ 7.2 years). YOLOv8 showed exceptional consistency (mAP@50 = 98.19%, CV = 0.77%). SHAP analysis identified bilateral area difference as the most determinant feature (sex: 0.42, age: 0.51). External validation on 50 independent images confirmed model generalizability. This study provides the first systematic comparison of YOLOv8 and nnU-Net v2 for forensic maxillary sinus analysis. nnU-Net v2 is recommended for precision-critical forensic reporting, while YOLOv8-Hybrid is suited for high-throughput screening. The > 120 radiomic/morphometric features establish a comprehensive framework for automated biological profiling.

Achieving stable operation at high current densities is a critical challenge for acidic water electrolysis, where intensified bubble evolution induces concentration polarization and mechanical stress that accelerate catalyst degradation. Here, we report Mo2C nanoclusters in situ anchored onto nitrogen-doped carbon nanotubes (NCNTs) via strong Mo─C and Mo─N covalent bonds. The covalent integration, achieved through electrostatically guided self-assembly followed by carbonization, yields a porous and conductive network that combines efficient charge transport with resistance to acidic corrosion. The resulting catalyst delivers overpotentials of 256 mV at 500 mA cm-2 and 396 mV at 1000 mA cm-2 in 0.5 m H2SO4, and operates stably at 865 mA cm-2 for over 240 h. In a proton exchange membrane water electrolyzer, it sustains overall water splitting at 1000 mA cm-2 with a cell voltage of 2.03 V for more than 150 h. This hierarchical covalent design enhances conductivity, durability, and bubble management, representing a scalable strategy for next-generation electrocatalysts capable of ampere-level operation in acidic media.

Grown in a transplant garden that provides field conditions but prevents predation by pocket gophers, plants of Erythronium grandiflorum (Liliaceae) have been exhumed annually (as dormant corms), photographed, and weighed over 33 years. From seed, plants grow to flowering size in about 5-6 years. They subsequently regulate their size by occasional vegetative splitting and by flowering and fruiting; producing one fruit costs a plant about 8% of the weight it would have gained if it had not flowered. Death is rare: a few plants have gradually lost weight and died in a way consistent with classical senescence, but others have died suddenly from fungal infection after previously growing robustly. Additional years of observation will be needed to clarify the issue of senescence. The data are archived and future collaborators are sought.

To address environmental pollution and sustainable energy challenges, lead-free Ag-Bi double perovskites Cs2AgBiX6 (X = Cl, Br) and their ammonium-substituted variants CsNH4AgBiX6 and (NH4)2AgBiX6 are investigated using first-principles FP-LAPW calculations within density functional theory. Ammonium incorporation slightly reduces lattice size while enhancing structural flexibility. Band-structure analysis (GGA, SOC, hybrid-PBE) shows decreasing band gaps with NH4 doping from 2.52 eV to 2.09 eV, with the CBM dominated by Bi states and the VBM by halide p states. Effective mass calculations indicate high carrier mobility due to the low effective mass of (NH4)2AgBiX6 (X = Cl, Br) compared to Cs-based double perovskites, which results in values that are between 0.524 and 0.939 eV, and values that are between 1.2 and 1.645 eV. The stability of these compounds is confirmed through mechanical (C ij ), formation of enthalpy ΔH f and Goldschmidt tolerance factor (τ G) analyses. The elastic constants confirm the mechanically stable and ductile nature of these materials. Furthermore, ab initio molecular dynamics simulations and phonon band-structure calculations have been performed and confirm the stability of the materials. Optical properties reveal stronger light absorption (∼45 × 104 cm-1 in the visible region) and an enhanced dielectric response after NH4 + substitution. Band-edge alignment analysis supports the potential for photocatalytic water splitting, while SLME analysis identifies (NH4)2AgBiBr6 (η max = 6.47%) as the most promising photovoltaic absorber. Overall, A-site ammonium engineering effectively tunes the structural, electronic, optical, and photocatalytic properties of Ag-Bi double perovskites for energy applications.

Staphylococcus aureus (S. aureus) is a common zoonotic bacteria, responsible for a wide range of infections and is well known for developing resistance to multiple antibiotics. In Ethiopia, information on methicillin-resistant S. aureus (MRSA), particularly from a One-Health perspective, is limited. This study aimed to detect S. aureus, identify MRSA strains, and assess their antibiogram patterns in swab samples collected from two municipal abattoirs in Northwest Ethiopia. A cross-sectional study was conducted between January 2021 and April 2022. A total of 150 swab samples were collected from beef carcasses, abattoir equipment, surfaces, workers' hands and clothes. Isolation and identification of S. aureus followed ISO 6888-2 standards. Antimicrobial susceptibility was tested against ten commonly used antibiotics using the disk diffusion method. Conventional PCR was used for detection of the mecA gene and Sanger method was used for sequencing. Overall, S. aureus was isolated from 25.3% (38/150) of the samples. The prevalence of S. aureus was 27.1% in beef carcasses, 26.9% in abattoir workers, and 23.1% in the abattoir environment. The prevalence was 22.7% in Bahir Dar and 28.0% in Debre Markos abattoirs. The highest detection rate (35.7%) was from workers' hands and hooks, while the lowest (11.1%) was from splitting axes. All isolates were susceptible to gentamicin but resistant to penicillin and methicillin. Multidrug resistance was observed in 60.5% of the isolates. Sequencing and phylogenetic analysis of the mecA gene showed that the current isolates were highly similar and clustered closely with mecA from Staphylococcus capitis and Staphylococcus fleurettii, but were distinct from other S. aureus strains. The detection of S. aureus and MRSA in beef carcasses, abattoir environments, and workers highlights potential risks to workers, consumers, and surrounding environments exposed to abattoir waste. Strengthening hygiene and sanitary practices in abattoirs is essential within a One Health framework.

Controlling collective behavior in the microscale is essential for advancing autonomous robotic systems in complex environments. While biohybrid microrobotic swarms offer considerable promise for targeted therapeutic and remediation applications, their programmable assembly and collective behavior remain challenging. Here, we describe an attractive light-triggered approach for enabling reconfigurable swarming of biohybrid microrobots based on the green microalga Chlamydomonas reinhardtii (CR). Such reversible swarming behavior is realized by combining the wavelength-dependent assembly ability of CR and its inherent phototactic properties with light exposures through a series of different mask openings that define the desired swarm geometry. Changes in the projected light enable dynamic modulation of the swarm shape and size, including real-time swarm splitting and merging behaviors. The concept was explored toward artificial intelligence-assisted wound targeting applications through the creation of microrobot swarms customized to exposed wound areas. Such powerful swarming capabilities offer considerable promise for the collective behavior of biohybrid microrobots toward important practical applications.