Browse Articles

Hemodynamic instability and retinal vein occlusion in glaucoma: Comparative analysis of heart rate variability and choroidal perfusion

Purpose To assess the hemodynamic and structural differences between glaucoma patients who developed retinal vein occlusion (RVO) and those who did not. Study design Retrospective, single-center, case-control study Methods This study included glaucoma patients who underwent a heart rate variability (HRV) test between January 2018 and July 2024. Patients were subdivided into RVO and non-RVO groups. Baseline mean deviation (MD) and pattern deviation (PSD) of the visual field and optical coherence tomography parameters were analyzed. Results Twenty-nine glaucoma patients with RVO and 34 glaucoma patients without RVO were included. Baseline MD and PSD had no difference (MD: −5.98 ± 9.05 vs. −3.70 ± 4.70, p  = 0.114; PSD: 3.88 ± 3.78 vs. 3.98 ± 3.96, p  = 0.883). HRV parameters, specifically Standard Deviation of NN interval (SDNN) and root-mean-square of successive differences (rMSSD), were significantly lower in the RVO group (SDNN: 22.12 ± 8.27 vs. 36.71 ± 24.74, p  = 0.002; rMSSD: 16.34 ± 9.55 vs. 29.87 ± 31.58, p  = 0.022). A significant difference in choroidal vascularity index was also observed between groups (64.62 ± 7.38 vs. 67.49 ± 5.90, p  = 0.045). In a parsimonious multivariable logistic regression model, no variable retained statistical significance. Conclusion This study suggests that reduced heart rate variability, reflecting autonomic dysfunction, is associated with the development of RVO in glaucoma patients. Although these associations did not remain statistically significant in multivariable analysis, the consistent univariate findings indicate that HRV may serve as a potential vulnerability marker rather than an independent predictor. Further prospective studies are warranted to clarify the clinical utility of HRV in risk stratification for RVO in glaucoma patients.

Machine learning prediction of weight gain after antiretroviral therapy initiation in people with HIV: Insights from a large french real-world cohort

Excessive weight gain after initiation of antiretroviral therapy (ART) has become a recognized concern among people living with HIV. Individual weight trajectories remain highly heterogeneous and challenging to predict using conventional methods. We leveraged the French Dat’AIDS national cohort to assess whether machine learning (ML) could enhance the prediction of individual body weight at 6, 12, and 24 months after ART initiation. Using 112 baseline variables encompassing demographic, clinical, laboratory, and treatment-related data, we trained XGBoost models and evaluated performance using root mean square error (RMSE), R², and mean prediction error. A simple benchmark model based on baseline weight was used for comparison. Among 24,014 eligible ART-naïve adults, the ML models achieved RMSEs of approximately 4.6 kg, 5.3 kg, and 6.4 kg at 6, 12, and 24 months respectively, with declining predictive power over time. Baseline weight (Weight_M0) consistently emerged as the strongest predictor, while other factors contributed minimally. Although ML marginally outperformed the benchmark (Weight_M0), accuracy remained insufficient for clinical decision-making. Sensitivity analyses excluding individuals with implausibly large monthly weight changes modestly improved RMSE (3.9–6.0 kg), underscoring the impact of data quality. Our results demonstrate that, despite large sample size and rich clinical variables, ML lacks the precision necessary for individual weight forecasting in this context. These findings highlight the limitations of applying artificial intelligence to heterogeneous real-world cohorts and underscore the need to incorporate behavioral and lifestyle factors to improve predictive modeling.

Power management and performance optimization of underwater wireless sensor networks based on MARL

In underwater wireless sensor network communication, communication performance degrades due to factors such as complex underwater channels and limited node resources. To reduce node redundancy energy consumption, improve transmission reliability, and extend the overall network lifetime, this study proposes an intelligent network performance optimization algorithm based on multi-agent reinforcement learning. By constructing an underwater wireless sensor network system model including fixed and mobile nodes, the network performance optimization problem is formalized as a partially observable Markov decision process. Then, multi-agent reinforcement learning is used to construct a comprehensive team reward function containing fair reuse rewards and survival time penalties, thereby establishing a distributed intelligent power management scheme. This solution enables each node to make transmission power decisions based on local observations, combined with the underlying media access control protocol, to collaboratively optimize higher-layer network performance indicators. The results show that in heterogeneous network scenarios, the proposed method achieves a network capacity of 245.68 kb and a fairness reuse index of 1.85. In imperfect networks with 5% node failures, the average communication latency is only 6.18 time slots, which is superior to the comparative algorithm. Under dynamic environments with a signal-to-interference-plus-noise ratio of 10–16 dB and a water flow velocity of 2.0 m/s, it can still maintain a network capacity of over 32,045 kb and an energy efficiency of 0.4 kb/J. These findings demonstrate that the proposed method significantly improves the robustness of underwater wireless sensor networks, providing communication support for ocean monitoring.

Qualitative analysis of genomic mutations and antibiotic susceptibility testing of Pseudomonas aeruginosa isolates from chronic lung infections

P. aeruginosa is intrinsically resistant to many antibiotics and may acquire resistance to others. The aim was to reconcile phenotypic resistance of isolates obtained from patients with chronic respiratory infection with the results of WGS. A total of 497 isolates were recovered from 4 countries between 2002 and 2016 from patients chronic pulmonary conditions, especially cystic fibrosis. Minimum inhibitory concentrations were determined previously by broth microdilution method. Sequencing was performed with Illumina technology and data were analyzed using ResFinder, PubMLST, and the CARD database. In this collection, resistance varied from 4.0% for colistin to 58.8% for ciprofloxacin. Acquired antibiotic resistance genes were found in 17.1% of the isolates, involving six different genes, but could not explain the majority of resistance. Single amino acid changes often did not lead to Minimal Inhibitory Concentrations (MICs) above the EUCAST susceptibility breakpoint. Determining the contribution of each amino acid change to clinical resistance was complicated by the large number of described changes found, the often low frequency of these changes and the high variability of the proteins involved. In particular, the diversity among OXA-ß-lactamases was large; despite over 900 OXA-types in the database, more than half of the variants in the isolate set were undescribed. Resistant isolates frequently had two or more amino acid changes. Four amino acid changes possibly related to β-lactam resistance were more common: two in AmpC (V239A and V356I) and two in PBP3 (R153S and R504C), three of which occurred more often in resistant isolates. Ciprofloxacin resistance could be linked to alterations in GyrA (in particular T82I and D87N) and to loss of or changes in MexZ. AmpD, NfxB, and PmrA, which are associated with resistance, were not detected in similar percentages of resistant and susceptible isolate. It can be concluded that frequently multiple mechanisms make a partial contribution to antibiotic resistance in this set of isolates.

Cost-effectiveness of sacituzumab tirumotecan in previously treated metastatic triple-negative breast cancer in China

Background The OptiTROP-Breast01 trial demonstrated the efficacy of sacituzumab tirumotecan for patients with metastatic triple-negative breast cancer (TNBC). The current analysis evaluated the cost-effectiveness of sacituzumab tirumotecan compared with chemotherapy for patients with metastatic TNBC from the Chinese health-care system perspective. Methods A partitioned survival model (PSM) was developed to simulate 3-week patients in 10-year time horizon to access the disease course and cost-effectiveness of sacituzumab tirumotecan compared with chemotherapy for metastatic TNBC patients, cost and utility values were gathered from the dataset and published studies, annual discount rate of 5% was used. Total cost, life-years (LYs), quality-adjusted life-years (QALYs), and incremental cost-effectiveness ratio (ICER) were the model outputs. Sensitivity analyses and subgroup analyses were conducted to estimate the robustness of the model outcomes. Results In base-case analysis, compared with chemotherapy, sacituzumab tirumotecan could bring additional 0.41 LYs and 0.35 QALYs, with marginal costs of $55,927.31, resulting in the ICER of $162,799.04/QALY, which high than the willingness-to-pay (WTP) threshold of $40,326 per additional QALY gained. One-way sensitivity analyses revealed that the utility value was the main driver of the model outputs. Probabilistic sensitivity analyses showed the cost-effective probability of sacituzumab tirumotecan was 0% at the WTP threshold of $40,326/QALY. Subgroup analyses suggested that sacituzumab tirumotecan could not be considered cost-effective for all subgroup patients. Conclusion Sacituzumab tirumotecan was unlikely to be the cost-effective option for patients with metastatic TNBC compared with chemotherapy from the Chinese health-care system perspective, reduced the price of sacituzumab tirumotecan could increase its cost-effective.

The need for patient-centric medicine design: investigating key physical characteristics of oral solid medications to improve acceptance of older patients in Addis Ababa, Ethiopia

Patient-centric medicine design emphasizes developing medication products tailored to patients’ specific needs and preferences, including size, shape, color, texture, packaging, and labeling. However, despite increasing recognition of patient-centered pharmaceutical design, evidence remains scarce on how medication and patient characteristics influence medication acceptance among older patients, particularly in low-resource settings such as Ethiopia. This study aimed to identify factors associated with the acceptance of oral solid medications among older adults. An institutional-based cross-sectional study was conducted from May to July 2024 in five public hospitals in Addis Ababa, Ethiopia. Data were collected through an interviewer-administered questionnaire capturing patient-related information (socio-demographics, clinical conditions, and medication beliefs) and medication-related characteristics (size, shape, score line, packaging, texture, dosage form, and labeling). Medication-level acceptance was measured using the MAQ-2019 tool. Descriptive statistics summarized the data, and bivariable and multivariable logistic regression analyses identified predictors of acceptance. Within-patient clustering was accounted for using cluster-robust standard errors. Among 408 older patients and 1,256 oral solid medications, 75% of medications were accepted. Medications sized 6–9 mm were more likely to be accepted (AOR = 6.50, 95% CI: 3.84–10.99, p < 0.001), while those ≥18 mm were less likely to be accepted (AOR = 0.15, 95% CI: 0.06–0.35, p < 0.001) compared with ≤6 mm. Medications with a score line were twice as likely to be accepted (AOR = 1.99, 95% CI: 1.32–3.00, p = 0.001). Older patients preferred small to medium-sized, coated, and scored tablets. Findings underscore the importance of age-appropriate formulation design and patient involvement in medication development.

The rehabilitation experiences, individual and combined effects of cognitive and physical rehabilitation on health and social outcomes in older athletes: A scoping review protocol

Background Evidence indicates that combined cognitive and physical rehabilitation can yield substantial improvements in health and social outcomes within the general aging population. However, the specific effects of such interventions on older athletes, who often exhibit enhanced resilience due to their competitive training backgrounds, remain inadequately explored. Objectives This study aims to describe the treatment regime (frequency, intensity, time/duration, and type of volume and progression of rehab) and components of cognitive and physical rehabilitation interventions, describe their combined effects on health and social outcomes and explore older athletes’ experiences with these interventions. Methods We will conduct a systematic scoping review following the five stages of the Arksey and O'Malley Framework. The search strategy will be developed in collaboration with a health and rehabilitation librarian, and searched in multiple data bases: MedLine, Web of Science, ProQuest, EBM Reviews, JBI EBP, Embase, APA PsychInfo, and Social Services Abstracts. Multiple reviewers will screen titles and abstracts, followed by full-text assessments in COVIDENCE, utilizing predefined inclusion and exclusion criteria. Articles will be included if participants were older athletes who actively engage in organized sports, characterized by regular competition and systematic training, and who have undergone cognitive and physical rehabilitation interventions. Data extraction will occur in pairs, focusing on relevant information to fulfill the study objectives. The pilot testing of all components—titles, abstracts, full texts, and data extraction protocols—will be conducted prior to the main review process. We will categories the physical and cognitive interventions using as frequency, intensity, time/duration, type of volume, progression of rehabilitation, and type of exercise. Additionally, data on intervention effects, including effect size and mean differences, will be summarized narratively. The experiences of the older athletes would be thematically analyzed. Discussion This scoping review will provide valuable information for clinicians, researchers, and policymakers to develop combined physical and cognitive rehabilitation programs tailored to the distinct physiological, cognitive and social needs of older athletes.

Correction: Disposal practices of cigarettes and electronic nicotine products among adults, findings from Wave 6 (2021) of the PATH Study

Configurational and chain-mediational path-ways linking natural environment perception to restorative environmental perception

While the link between natural environment perception (NEP) and restorative environmental perception (REP) is well-established, the specific psychological mechanisms—how environmental preference (EP) and place attachment (PA) configure this relationship—remain underexplored. This study aims to bridge this gap by examining the symmetric and asymmetric pathways translating nature perception into restorative outcomes. Analyzing survey data from 432 visitors to Zhangjiajie National Forest Park, China, we employed Partial Least Squares Structural Equation Modeling (PLS-SEM) to test serial mediation and Fuzzy-set Qualitative Comparative Analysis (fsQCA) to identify equifinal configurations. PLS-SEM results indicate that NEP enhances REP directly and indirectly through EP. Furthermore, a significant serial mediation path (NEP → EP → PA → REP) was established. Complementing these findings, fsQCA revealed EP as a “core condition” across all configurations for high REP, whereas PA serves as a “peripheral condition.” Theoretically, these results identify environmental preference as the critical gatekeeper for restoration. Practically, they suggest managers should prioritize aesthetics that trigger immediate preference to enhance restorative experiences.

Salmonella enterica persister cells exhibit distinct susceptibility profiles following exposure to human serum and macrophages

Salmonella enterica , particularly non-typhoidal serovars (NTS), is a leading cause of foodborne illness, with invasive infections posing high mortality risks in developing countries. Fluoroquinolones and third-generation cephalosporins, such as ceftazidime (CAZ), are used to treat severe infections, yet they are facing concerning rates of antimicrobial resistance. Furthermore, recalcitrant and/or persistent infections are often linked to persister cells, a phenotype that enables cells to survive in the presence of high concentrations of antibiotics. Although persisters are associated with chronic infections, their interactions with the human immune system, particularly serum resistance and opsonophagocytosis, are not well understood. Here, three NTS isolates from the food protein chain (S45, S48, and 4SA(2)) were used. Persister cells were selected by exposure to CAZ concentration 100 times higher than the minimum inhibitory concentration and then assessed for serum resistance, opsonophagocytosis, and intracellular survival in primary human macrophages. The isolates exhibited heterogeneous persister fractions (1.06%–39.55% survival after 72h of CAZ exposure). Persisters exhibited equal or greater serum resistance than regular cells. Isolate 4SA(2) proliferated in 100% human serum, with persister-derived cells showing higher growth rates. Following opsonization, serum-resistant persisters of all isolates were phagocytosed at significantly higher rates than serum-resistant regular cells. Intracellular survival varied: S45 persisters proliferated post-internalization; S48 persisters and regulars were eradicated; 4SA(2) showed no phenotype difference. Complement enhanced the intracellular survival of S45 but not S48 or 4SA(2). Despite having different intracellular outcomes, Salmonella persisters showed higher levels of opsonophagocytosis and serum resistance. These findings suggest that cell surface modifications may facilitate host cell uptake and contribute to antimicrobial treatment failure and long-term infection. The phenotypic diversity among isolates underscores the importance of considering persister heterogeneity and host-pathogen immune interactions in order to understand recalcitrant infection dynamics and design more effective therapeutic strategies.

Bovine brucellosis seropositivity in Mpumalanga Province, South Africa, 2021–2024: Temporal, and spatial trends

Introduction Bovine brucellosis, caused primarily by Brucella abortus, remains a major constraint to livestock productivity and a persistent zoonotic threat. Although brucellosis is a controlled disease in South Africa, detailed subnational epidemiological evidence is limited, particularly for Mpumalanga Province. Understanding temporal, seasonal, and spatial patterns is essential for improving risk-based surveillance and control. Materials and methods A retrospective cross-sectional analysis was conducted using routine diagnostic records from the Mpumalanga Provincial Veterinary Laboratory between January 2021 and December 2024. Rose Bengal Test (RBT) results from cattle originating from 17 Local Municipality Areas (LMAs) were analysed. Annual, seasonal, and spatial seroprevalence estimates were calculated, and independent predictors of RBT seropositivity were evaluated using multivariable logistic regression. Results A total of 67,974 cattle serum samples were tested, of which 6,182 were RBT-positive, yielding an overall seroprevalence of 9.1% (95% CI: 8.9–9.3). Annual seroprevalence increased from 6.9% in 2021 to 7.4% in 2022, peaked at 13.1% in 2023, and declined to 7.5% in 2024. Clear seasonal variation was observed, with higher seroprevalence in spring (10.6%) and summer (10.2%) compared with autumn (6.8%) and winter (6.9%). Pronounced spatial heterogeneity was evident, with Emalahleni (13.3%), Victor Khanye (13.0%), and Mbombela (12.0%) identified as high-burden municipalities, while Mkhondo (1.7%) and Albert Luthuli (2.7%) recorded the lowest prevalence. In adjusted analyses, testing in 2023 was associated with nearly double the odds of seropositivity compared with 2021 (AOR 1.95; 95% CI: 1.81–2.11), and spring and summer remained significant predictors. Conclusion Bovine brucellosis in Mpumalanga exhibits marked temporal variability, seasonal peaks, and spatial clustering. These findings support targeted, risk-based surveillance, strategically timed vaccination, and strengthened biosecurity, prioritising hotspot municipalities and high-risk seasons within a One Health framework.

Unbiased quantification of persistent postural and motor deficits following spinal cord injury in mice

Spinal cord injury (SCI) causes multifaceted postural and motor impairments that are challenging to quantify. Conventional behavioral tests, such as the Basso mouse scale (BMS), rely on qualitative observations, which do not detect subtle yet significant functional deficits. To perform unbiased and quantitative evaluation of motor function and posture after SCI, we employed high-speed video tracking with machine learning-based whole body pose estimation and weight-bearing analyses using the Blackbox system in freely moving mice. We found enduring alterations in posture induced by SCI not captured by conventional metrics. Postural deficits included reduced hindpaw spacing with concomitant increased forepaw spacing, narrowed femur width, and altered hindpaw angles, which remained evident beyond 42 days post-injury (dpi). Furthermore, sustained deficits in locomotor activity were identified as decreased distance traveled, decreased exploratory behavior, and disrupted fore-to-hindpaw speed ratio. Additionally, we analyzed behavioral motifs using Keypoint MoSeq software and found frequency changes in unique motor syllables correlating with forward acceleration and turning after SCI. Interestingly, while motor deficits persisted, sensory deficits, such as thermal and mechanical sensitivity, returned to baseline levels by 21 days after injury in C57BL/6J mice, with no subsequent hypersensitivity. Lastly, we developed a web-based application to assist in visualization and analyses of Blackbox-based kinematic data. Altogether, our study identifies distinct postural and motor deficits pre- and post-SCI using accurate, unbiased, and quantitative behavioral assessments. By tracking the unique features of motor recovery trajectories, researchers can more accurately assess the effectiveness of SCI therapeutic strategies.

“It becomes more difficult when people don’t empathize with us”: COVID-19-related stigmatization experienced by survivors in Nepal

The COVID-19 pandemic caused widespread social disruption, with stigma emerging as a significant challenge for individuals who survived infection. This qualitative study explored the forms, drivers, and impacts of COVID-19-related stigma among survivors in Eastern Nepal. In-depth interviews were conducted with 15 COVID-19 survivors who had reported stigma in a preceding cross-sectional survey. Due to pandemic-related restrictions, interviews were conducted over the phone. Data were analysed thematically following the process outlined by Braun and Clarke. COVID-19 stigma was multifaceted, including social rejection, internalized stigma, and discriminatory practices by community members. Key drivers of stigma included self-directed fear of infection and death, misinformation and limited awareness about COVID-19 transmission and prevention, and a fragile health system and policy responses. Although COVID-19-related stigma may have declined as the pandemic evolved, the findings illustrate how stigma can emerge rapidly during health emergencies that can have social consequences related to trust, disclosure, and help-seeking behavior in future crisis. The study highlights the importance of outbreak preparedness strategies that integrate clear communication, strengthened health system capacity, and social protection measures to mitigate stigma and its harms during future public health crises.

A full belly counsels well: More stabilisation with more responsibility

Design/methodology/approach This paper uses the data of Chinese A-share listed companies from 2009 to 2022 to study the impact of supply chain stability on corporate ESG performance. Purpose To study the impact and mechanism of supply chain stability on corporate ESG performance. Findings This paper found that the improvement of supply chain stability can improve the ESG performance of firms, and this conclusion still holds after the instrumental variables method, systematic GMM method, PSM method, omitted variables test and Double Machine Learning (DML) approach, and the improvement of supply chain stability can optimise the ESG performance of firms through the channels of reducing the corporate risk-taking, reducing the agency costs, and reducing the financing constraints, and this facilitating effect is more significant in the large firms, firms with higher-standard audit supervision, firms located in western regions, and non-technology-intensive firms. Practical implications The findings of this paper can provide a realistic framework for national and local governments to actively promote the stable development of supply chains in order to achieve sustainable economic development. Social implications This paper deepens the understanding of the external stakeholders of enterprise sustainable development, and provides an opportunity to actively play the role of external stakeholders in monitoring the development of enterprises, participate in corporate governance of enterprises, and achieve a win-win situation of supply chain stability and environmental sustainability. Originality/value This study contributes to the literature by shedding light on the relationship between supply chain stability and ESG in the context of external stakeholders.

Psychosocial stressors, accelerated biological aging, and multiple morbidities: Evidence from an age-diverse sample

Exposure to psychosocial stress is a well-established risk factor for poor health and premature mortality, yet most research has focused narrowly on single sources of stress without simultaneously modeling multiple stress exposures occurring across the life span. Using data from a state-representative sample of 2,267 adults ages 18–103, we examined associations between four psychosocial stressors – adverse childhood experiences (ACEs), stressful life events, chronic financial strains, and everyday discrimination – and DNA methylation-based biological aging clocks (GrimAge2 and DunedinPACE) alongside six indicators of physical and mental health outcomes. All stressors were associated with accelerated epigenetic aging and poorer health when examined individually. However, when considered simultaneously, financial strains and everyday discrimination emerged as more consistent predictors across all outcomes, relative to childhood adversity and stressful events in adulthood. Overall, stressor effects were more pronounced for mental health compared to physical health or biological aging. These findings highlight the importance of considering multiple sources of stress on varying indicators of aging, disease, and distress to fully account for the health significance of stress exposure.

Exploring episodic specificity induction on divergent thinking in children

Previous studies suggest that Episodic Specificity Induction (ESI) improves the recall of episodic details and facilitates transfer to other cognitive tasks requiring episodic thinking (i.e., divergent thinking). However, the only study examining an adapted future-oriented ESI in children has failed to show benefits in subsequent cognitive tasks. To investigate this, two experiments were conducted using the standard ESI protocol with children. Experiment 1 tested second graders, fifth graders, and young adults using children-adapted materials (i.e., TV cartoons), while Experiment 2 tested fifth graders using non-adapted materials. Both experiments confirmed that ESI improved the recall of episodic details compared to a control condition. Additionally, developmental differences in episodic recall in Experiment 1 disappeared after controlling for total verbal production, suggesting that children’s episodic memory benefits when recalling materials that are child-friendly. Conversely, unexpected findings regarding transfer effect to divergent thinking revealed no transfer effects in Experiment 2 (non-adapted materials) and a significant increase in idea fluency and flexibility following the control condition in Experiment 1 (children-adapted materials). This result may be explained by a positive mood induction, as general questions accompanied by child-friendly videos could enhance creative performance following the control condition. These findings highlight the importance of carefully selecting and adapting ESI materials to children population in future studies.

Moderating role of CEO expertise on the relationship between capital structure and financial reporting timeliness of Saudi-listed companies

In this study, we investigated the effect of capital structure on financial reporting timeliness with an interaction role of CEO financial expertise. Using the fixed effects technique, we analysed data from listed firms on the Saudi Stock Market between 2014 and 2023. Our results showed that capital structure choice through debt financing may significantly influence firms to reveal their audited accounts on a timely basis to signal their financial capabilities. Additionally, the results provide strong evidence that a CEO’s financial expertise may enhance the role of debt financing in reducing audit report delays, consistent with upper echelons and agency theories. The findings appear to be robust with the use of alternative measures, the COVID-19 effect and endogeneity control.

Defect detection method of printed circuit boards based on EDF-YOLOv10

To address the challenges of inadequate feature representation for small objects and slow model convergence in printed circuit board (PCB) defect detection, this paper proposes an improved YOLOv10 algorithm and develops a real-time detection system with a co-optimized hardware and software architecture. The efficient channel attention (ECA) mechanism is used to enhance the ability of the model to extract key channel features; the dynamic snake convolution (DSConv) in the backbone strengthens the model’s capacity to recognize the geometric structures of small targets through deformable kernels and multi-directional feature fusion; the Focaler-CIoU loss emphasizes samples with low intersection over union (IoU) values to boost hard sample learning and improve convergence efficiency. To simulate real-world industrial environments, multiple data augmentation strategies are utilized to expand the PKU-Market-PCB dataset, thereby enhancing the model’s generalization and robustness in complex scenarios. Experimental results demonstrate that the proposed EDF-YOLOv10 achieves mAP@0.50 of 90.6% and mAP@0.50:0.95 of 48.4% on the experimental dataset, representing improvements of 3.0 and 1.6 percentage points over the baseline, respectively. Furthermore, We also develope a real-time interactive detection system for identifying PCB defects. This system utilizes industrial cameras, a controllable light source, and a graphical user interface developed with the PyQt5 framework, employing the EDF-YOLOv10 model. Our approach serves as a methodological reference for detecting PCB defects in complex industrial environments.

Engaging stakeholders, shaping AI ethics: Targeted engagement in corporate AI ethics statements

Corporate AI ethics statements are being increasingly integrated into sustainability communication to showcase responsible AI performance as part of broader sustainable efforts. This study examines how leading AI companies (developers and adopters) engage with diverse stakeholders through their AI ethics statements. Integrating stakeholder theory with the engagement system, it analyses how Strategic, Responsive, and Operational engagement strategies address the needs of primary and secondary stakeholders. The analysis included 122 English-language statements (265,952 words) from those firms, covering their AI ethics guidelines and policies, press releases, and corporate AI ethics reports. The findings show statistically significant differences in usage frequencies in how primary and secondary stakeholders communicated with targeted engagement strategies. For primary stakeholders, particularly customers and third-party developers, companies frequently used three Operation-targeted strategies to detail data protection measures and AI ethics-related product updates. When addressing corporate management and local regulators, Responsive Acknowledge and Strategic Endorse were more prevalent. In contrast, communication with secondary stakeholders, particularly society, interest groups, and academia, frequently used Strategy-targeted engagement to outline broader ethical AI development plans. These findings inform the development of a stakeholder-oriented engagement model for corporate disclosures on their own trustworthy AI practices.

Social determinants of healthy aging: An investigation using the all of us cohort

Introduction The increasing aging population raises significant concerns about the ability of individuals to age healthily, avoiding chronic diseases and maintaining cognitive and physical functions. However, the pathways through which SDOH factors are associated with healthy aging remain unclear. Methods This retrospective cohort study uses the registered tier data from the All of Us Research Program (AoURP) registered tier dataset v7. Eligible study participants are those aged 50 and older who have responded to any of the SDOH survey questions with available EHR data. Three different algorithms were trained (logistic regression [LR], multi-layer perceptron [MLP], and extreme gradient boosting [XGBoost]). The outcome is healthy aging, which is measured by a composite score of the status for 1) comorbidities, 2) cognitive conditions, and 3) mobility function. We evaluate the model performance by area under the receiver operating characteristic curve (AUROC) and assess the fairness of best-performed model through predictive parity. Feature importance is analyzed using SHapley Additive exPlanations (SHAP) values. Results Our study included 99,935 participants aged 50 and above, and the mean (SD) age was 74 (9.3), with 55,294 (55.3%) females, 67,457 (67.5%) Whites, 11,109 (11.1%) Hispanic ethnicity, and 44,109 (44.1%) are classified as healthy aging. Most of the individuals lived in their own house (64%), were married (51%), obtained college or advanced degrees (74%), and had Medicare (56.2%). The best predictive model was XGBoost with random oversampler, with a performance of AUROC [95% CI]: 0.793 [0.788–0.796], F1 score: 0.697 [0.692–0.701], recall: 0.739 [0.732–0.748], precision: 0.659 [0.655–0.663], and accuracy: 0.716 [0.712–0.720], and the XGBoost model achieved predictive parity by similar positive and negative predictive values across race and sex groups (0.86–1.06). In feature importance analysis, health insurance type is ranked as the most predictive feature, followed by employment status, substance use, and health insurance coverage (yes/no). Conclusion In this cohort study, XGBoost model accurately predicted individuals achieving healthy aging, outperforming LR and MLP. Our findings underscore the significant role of health insurance in contributing to healthy aging.