Search Results (685)

Fairness-aware classification with respect to sensitive attributes, such as gender and race, is one of the most important topics in machine learning. Although numerous studies have made outstanding progress through various approaches, one key limitation is that they necessarily require additional labels of sensitive attributes for training. This poses a significant challenge since sensitive attributes typically correspond to personal information. To this end, we propose a novel reweighting method that dynamically gives more weights to underrepresented groups across potential sensitive attributes. Without auxiliary networks or strong assumptions about sensitive attributes, the proposed method significantly improves fairness under various scenarios on benchmark datasets, outperforming the existing state-of-the-art methods. Full article

Recent advances in deep learning (DL) have enabled the integration of diverse biomedical data for disease prediction and risk stratification. Building on this progress, the overall objective of this study was to develop and evaluate a multimodal DL framework for robust multi-label classification (MLC) of major comorbidities in patients with obstructive sleep apnea (OSA) using physiological time series signals and clinical data. This study proposes a robust framework for multi-label classification (MLC) of comorbidities in patients with OSA using diverse physiological and clinical data sources. We conducted a retrospective observational study including a convenience sample of 144 patients referred for overnight polysomnography at the Sleep Medicine Center (SleepLab Split), University Hospital Centre Split (KBC Split), Split, Croatia. Patients were selected based on predefined inclusion criteria and data availability. A one-dimensional Convolutional Neural Network (1D-CNN) was developed to process and fuse time series signals, oxygen saturation ($Sp{O}_2$), derived $Sp{O}_2$ features, and nasal airflow (FP0), with demographic and physiological parameters, enabling the identification of key comorbidities such as arterial hypertension, diabetes mellitus, and asthma/COPD. The instruments included polysomnography-derived signals (SpO₂ and FP0 airflow) and structured demographic/physiological parameters. Signals were preprocessed and used as inputs to the proposed fusion model. The proposed model was trained and fine-tuned using the Optuna hyperparameter optimization framework, addressing class imbalance through weighted loss adjustments. Its performance was comprehensively assessed using multi-label evaluation metrics, including macro/micro F1-score, AUC-ROC, AUC-PR, subset and partial accuracy, Hamming loss, and multi-label confusion matrix (MLCM). The study protocol was approved by the Ethics Committee of the School of Medicine, University of Split (Approval No. 003-08/23-03/0015, Date: 17 October 2023). The 1D-CNN achieved superior predictive performance compared to traditional machine learning (ML) classifiers with macro AUC-ROC = 0.731 and AUC-PR = 0.750. The model demonstrated consistent behavior across age, gender, and BMI groups, indicating strong generalization and minimal demographic bias. In conclusion, the results confirm that $Sp{O}_2$ and airflow signals inherently encode comorbidity-specific physiological patterns, enabling efficient and scalable screening of OSA-related comorbidities without the need for full polysomnography. Although the study is limited by data set size, it provides a methodological basis for the application of multi-label DL models in clinical decision support systems. Future research should focus on the expansion of multi-center datasets, thereby improving model interpretability and potential clinical adoption. Full article

Background: Preterm birth is a significant risk factor for neurodevelopmental delays, but the appropriate use and timing of age correction for developmental assessment remain debated. Objective: This study investigated psychomotor development in preterm children at two years of age, with the aim of clarifying whether age correction remains necessary at this stage, particularly across different gestational age groups. Methods: A total of 161 preterm infants were assessed at a mean chronological age of 25.4 months (mean corrected age: 23.3 months) and compared with two control groups of typically developing children matched for gender and either corrected age (Control–Corr, N = 88) or chronological age (Control–Chron, N = 87). The preterm group was further stratified by gestational age: extremely preterm (<28 weeks), very preterm (28–31 weeks), and moderate-to-late preterm (32–36 weeks). Cognitive, Language (Receptive, Expressive), and Motor (fine, gross) scales of Bayley-III were analysed using t-tests and MANOVAs. Results: Using corrected age, preterm children showed a selective profile, with deficits in Receptive Language, borderline mean score in Gross Motor, and preserved performance in Cognitive, Expressive Communication, and Fine Motor. When compared with controls of the same age, significant differences emerged in the Cognitive, Language, and Gross Motor, but not Fine Motor, domains. In contrast, scoring by chronological age produced a generalised delay, with preterm children performing significantly worse than chronological-age controls across all domains. Subgroup analyses further showed that extremely preterm children already displayed marked Language vulnerabilities at corrected age, which became more severe with chronological scoring and extended to other domains. Very preterm children also fell into the deficit range in Cognitive, Language, and Gross Motor scales/subscales when chronological age was applied, whereas moderate-to-late preterm children performed comparatively better. Conclusions: Developmental assessment using corrected age remains essential at least until 24 months, especially for extremely and very preterm children, to avoid substantial overestimation of developmental difficulties. Chronological scoring, while helpful to highlight persistent vulnerabilities, may inflate delay classification if used too early. Tailoring correction strategies by gestational age and developmental domain could provide a more accurate and clinically meaningful representation of preterm children’s developmental trajectories. Full article

Background/Objectives: Most patients with alcohol use disorder (AUD) suffer from mild cognitive decline, which does not meet the diagnostic criteria of the severe form of alcohol-related cognitive impairment (ARCI). ARCI is associated with executive abnormalities in addictive behaviors and therefore influences relapse and daily functioning. Abnormalities in speech production reflect cognitive disturbances. The aim of this study was to examine the temporal speech parameters (TSPs) in ARCI. Methods: The TSPs were measured with the S-GAP Test^® on 34 AUD patients with intact cognitive functions and 31 age- and gender-matched control participants. Results: Ten out of fifteen parameters of TSPs were significantly different between the AUD and healthy groups. Speech tempo and the total pause duration rate have significant classification potential. Conclusions: Our exploratory study revealed that filled pause-related temporal parameters appear to be particularly altered in ARCI and indicated that marked TSP alterations could serve as early indicators of cognitive deficits. Full article

This study examined Croatian consumer attitudes towards fresh pork from the Black Slavonian pig, focusing on the following sustainability dimensions: environmental, social, economic sustainability, and animal welfare. A survey of 410 consumers was conducted in June 2021, using an online questionnaire assessing consumption habits, breed knowledge, and socio-demographic characteristics. Factor analysis identified four key dimensions: attention to animal welfare, support for local production and biodiversity, origin and information, and price and intrinsic quality. Cluster analysis revealed three distinct consumer segments: conscious consumers (32.4%), value-oriented consumers (37.3%), and uninvolved meat consumers (30.2%). Multinomial logistic regression showed that age, region, family economic status, and place of purchase significantly predicted cluster membership (Nagelkerke R² = 0.251, classification accuracy = 52.9%), while gender, education level, and household composition did not. Conscious consumers were characterised by older age, higher income, and a preference for direct purchasing channels, while value-oriented consumers favoured supermarkets and mid-range pricing. These findings highlight the need for improved consumer education, transparent labelling, targeted marketing strategies, and enhanced policy support to promote sustainable indigenous pig breed production and conservation. Full article

The COVID-19 pandemic exacerbated structural, social, economic, and racial inequalities affecting immigrant, refugee, and asylum-seeking men—vulnerable populations often overlooked in men’s health research. This study investigated the health conditions of immigrant, refugee, and asylum-seeking men during the COVID-19 pandemic. A scoping review was conducted following Joanna Briggs Institute guidance, and a qualitative lexical analysis (text-mining of standardized study syntheses) was performed in IRaMuTeQ using similarity analysis, descending hierarchical classification, and factorial correspondence analysis. We identified 93 studies published between 2020 and 2023 across 35 countries. The evidence highlighted vaccine hesitancy, high epidemiological risks (infection, hospitalization, and mortality), barriers to accessing services and information, socioeconomic vulnerabilities, psychological distress (e.g., anxiety and depression), and structural inequalities. Findings were synthesized into four integrated thematic categories emphasizing the role of gender constructs in help-seeking and gaps in governmental responses. Most studies focused on immigrants, with limited evidence on refugees and especially asylum seekers; therefore, conclusions should be interpreted cautiously for these groups. Overall, the review underscores the urgency of multisectoral interventions, universal access to healthcare regardless of migration status, culturally and linguistically appropriate outreach, and gender-sensitive primary care strategies to support inclusive and resilient health systems. Full article

Background and Objectives: Type 2 diabetes mellitus (T2DM) is increasing in prevalence worldwide, placing a substantial burden on healthcare systems, particularly in resource-limited settings. In Yemen, limited screening and diagnostic capacity contribute to delayed detection and management. Prediabetes, a reversible state of dysglycemia, carries significant cardiovascular risk and frequently progresses to diabetes. Early identification of both conditions is vital for prevention and public health planning. Materials and Methods: This cross-sectional study, conducted from July 2024 to May 2025 in three medical centers in Ibb Governorate, Yemen, assessed 1045 adults aged 18–60 years without known diabetes or prediabetes. Glycaemic status was classified according to the 2025 American Diabetes Association criteria. Undiagnosed diabetes was defined using three diagnostic combinations: FBS + OGTT, FBS + HbA1c, and OGTT + HbA1c. Results: The prevalence of undiagnosed diabetes was 8.4% (FBS + OGTT) and 9.76% (FBS + HbA1c or OGTT + HbA1c). Prediabetes prevalence was 23.4%, 14.7% and 26.4% based on FBS, OGTT, and HbA1c, respectively. Females represented a higher proportion of undiagnosed diabetes and prediabetes cases. Age was significantly associated with glycemic status across all tests, while gender showed significant associations with FBS and HbA1c. Family history of chronic disease was significantly associated with HbA1c-based classification. Approximately 8–10% of adults in Ibb had undiagnosed diabetes, while up to one-quarter had prediabetes. Age and family history were key predictors of dysglycaemia. Conclusions: These findings highlight the need for targeted, multi-marker screening and early intervention strategies, particularly in relatively stable regions of conflict-affected settings, to prevent progression to diabetes and reduce long-term complications and healthcare burden. Full article

Many states around the world create an unprotected class of migrants by legally categorizing them as “unauthorized”. Yet, we have a limited understanding of the state of knowledge that has resulted from this empirical phenomenon, particularly outside the U.S. and over time. In this article, we map the state of knowledge on unauthorized migration by analyzing the last 30 years of papers published in a leading migration journal. Articles were identified through a comprehensive keyword-based search strategy and analyzed using a computational pipeline that combines natural language processing and large language model-assisted classification. Our findings reveal a persistent empirical emphasis on Mexico–U.S. migration, with economic drivers and disparities, as well as immigration laws and policies, dominating articles’ content. Our analysis also identifies underexplored or peripheral topics, including studies on gender or the environment, highlighting the need for more diversified, cross-national research. Ultimately, by providing a detailed account of our computational mapping approach, we hope this study will serve as a blueprint for future scholars to track how migration research evolves into the future. Full article

Background: Malnutrition is a common yet often underestimated risk factor for adverse outcomes in hospitalized patients, including heart transplant recipients. Assessing nutritional status in this population is challenging due to comorbidities, pharmacotherapy, and the urgent nature of surgery. This study aimed to evaluate the reliability of routinely measured laboratory and anthropometric parameters in diagnosing malnutrition in heart transplant patients. Methods: This retrospective study included 53 adult patients who underwent orthotopic heart transplantation between 2021 and 2024 at the Silesian Center for Heart Diseases in Zabrze, Poland. Anthropometric data (gender, age, BMI) and laboratory parameters—albumin, total protein, hemoglobin, cholesterol, C-reactive protein (CRP), and neutrophil-to-lymphocyte ratio (NLR)—were analyzed. Malnutrition was defined as BMI < 22 kg/m². Cut-off points were based on literature data. Correlations between laboratory parameters and nutritional status were assessed, and concordance in malnutrition classification was evaluated. Results: Malnutrition criteria were met by 15% of patients. Only CRP predicted malnutrition risk, though its values did not differ significantly between BMI groups (p = 0.106). Negative correlations were found between CRP and total protein (−0.342; p = 0.012), albumin (−0.666; p < 0.0001), cholesterol (−0.287; p = 0.037), and hemoglobin (−0.383; p = 0.0046). A positive correlation was observed between CRP and NLR (0.333; p = 0.014). Conclusions: Malnutrition assessment in heart transplant recipients should not rely solely on individual laboratory parameters. A multifactorial approach integrating biochemical, anthropometric, and clinical data is necessary. Further research is needed to identify novel biomarkers to improve malnutrition risk evaluation and guide nutritional interventions in this population. Full article

Guided by Sustainable Development Goal (SDG) 5 on achieving gender equality and empowering all women and girls, our study examines the age, period, and cohort effects of gender income inequality across China’s public and private sector employment by utilizing hierarchical age–period–cohort cross-classification random-effects models (HAPC-CCREMs) and repeated cross-sectional data from the Chinese General Social Survey from 2003 to 2021 (N = 29,367). The results demonstrate the following: (1) Age effects of gender income inequality diverge between public and private sector employment. In public sector employment, inequality undergoes a progressive decline over individuals’ career spans, as age is institutionalized as a sector-specific capital and compresses inequality through seniority-based accumulation. In private sector employment, inequality follows an inverted U-shaped trend as age is marketized as a proxy for labor productivity, producing steeper inequality in individuals’ early careers and sharp declines thereafter. (2) Period effects of gender income inequality manifest significant developing differences across public and private sector employment between 2003 and 2021. In public sector employment, the state redistributive mechanism maintains inequality at a consistently low and stable level. In private sector employment, inequality fluctuates with China’s post-transition economic restructuring, expanding during rapid market growth (2003–2008), contracting amid structural upgrading (2010–2013), and rising again under deeper market integration (2015–2021). (3) Cohort effects are negligible, reflecting that mechanisms sustaining gender income inequality exhibit intergenerational continuity. These results demonstrate that institutional segmentation structures gendered income dynamics throughout the life course via distinct resource allocation mechanisms. Our study extends life course approaches to social inequality, emphasizing the role of gender-equality-oriented governance, lifecycle-spanning support mechanisms, and cross-sectoral coordination in mitigating gender disparities. Full article

Community Question-Answering (cQA) sites have an urgent need to be increasingly efficient at (a) offering contextualized/personalized content and (b) linking open questions to people willing to answer. Most recent ideas with respect to attaining this goal combine demographic factors (i.e., gender) with deep neural networks. In essence, recent studies have shown that high gender classification rates are perfectly viable by independently modeling profile images or textual interactions. This paper advances this body of knowledge by leveraging bimodal transformers that fuse gender signals from text and images. Qualitative results suggest that (a) profile avatars reinforce one of the genders manifested across textual inputs, (b) their positive contribution grows in tandem with the number of community fellows that provide this picture, and (c) their use might be detrimental if the goal is distinguishing throwaway/fake profiles. From a quantitative standpoint, ViLT proved to be a better alternative when coping with sparse datasets such as Stack Exchange, whereas CLIP and FLAVA excel with a large-scale collection—namely, Yahoo! answers and Reddit. Full article

Historically, singlehood is a growing demographic trend shaped by economic, social, and personal factors. This study examines the key influences associated with this phenomenon across diverse global contexts based on empirical evidence provided by WVS (World Values Survey), which covers over 100 countries and spans four decades. A multi-technique analytical approach is applied to identify the most robust predictors of singlehood. This approach involves feature selection, cross-validation, robustness checks, and statistical modeling (parsimonious models with near-excellent or excellent classification accuracy as AUCROC > 0.9). The results indicate that age and parental status are negatively associated with singlehood, while precarious employment status is positively linked. Co-residence with parents also appears closely related to singlehood. Other factors, including education level, social class, and settlement size, also correlate with singlehood patterns, as resulting from supplemental analyses. Moreover, gender and regional analyses reveal some variations in these associations, highlighting the interplay between personal, cultural, and economic contexts. These findings also align with social and economic theories of marriage, emphasizing the impact of life course factors, financial stability, and cultural norms. They contribute to a deeper understanding of demographic shifts. They also provide meaningful and well-founded insights as well as strategic guidance for policy in areas such as youth employment, social welfare, urban planning, and demographic adaptation. Full article

Background: Preservation of posterior occlusal support is critical for maintaining residual dentition in partially edentulous patients. Although implant-supported restorations are widely used, long-term evidence on their role in preventing tooth loss remains limited. Methods: A retrospective cohort study of 225 patients (80 males, 145 females; mean age 53 ± 9 years) who received posterior implant-supported restorations, with a mean follow-up of 21 years. Post-treatment occlusal support was classified using the Eichner classification. Tooth and implant loss were recorded. Associations between residual occlusal support and tooth loss were analyzed using Pearson’s chi-square test, with thresholds of ≥1, ≥2, or ≥3 teeth lost. Multivariate logistic regression analyses adjusted for age and gender were performed to evaluate the independent association of the Eichner classification. Results: After prosthesis delivery, 159 patients were classified as Eichner A, 48 as B1, 16 as B2, and 2 as B3. The mean number of lost teeth was 2.4, and 35 implants failed in 22 patients. Tooth loss ≥3 occurred more frequently in the Eichner B groups than in A. (48.5% vs. 34.0%). In multivariate analyses, Eichner classification significantly predicted loss of ≥3 teeth (adjusted OR = 1.83; 95% CI: 1.00–3.33; p = 0.048), but not ≥1 or ≥2 teeth. Conclusions: Posterior implant therapy provides durable occlusal support and contributes to long-term preservation of residual dentition. Patients with limited occlusal support remain at a higher risk of substantial tooth loss, even after adjustment for age and gender. The Eichner classification represents a practical clinical tool for evaluating occlusal support and predicting long-term prognosis in partially edentulous patients. Full article

Lymphovascular invasion (LVI) is a potent yet underutilized prognostic marker in bladder cancer, particularly in non–muscle-invasive disease (NMIBC). We aimed to develop and internally validate a predictive nomogram to estimate the probability of LVI at initial transurethral resection of bladder tumors (TURBT), utilizing preoperative clinical parameters. In this retrospective cohort study, 413 patients with histologically confirmed urothelial carcinoma who underwent initial TURBT were included. LVI was identified histologically in 9.2% of cases. Univariate and multivariate logistic regression, in conjunction with the least absolute shrinkage and selection operator modeling, revealed eight significant predictors: papillary architecture, Box–Cox–transformed tumor size, urinary cytology classification, age ≥ 75 years, pedunculated morphology, gender, hydronephrosis, and tumor multiplicity. The resulting nomogram demonstrated excellent discriminative performance, with an AUC of 0.888 in the training cohort and 0.827 in the validation cohort, and exhibited good calibration based on weighted plots. This model facilitates individualized prediction of LVI using routinely available clinical data. Early detection of LVI may inform risk-adapted management strategies, including repeat resection, or intensified surveillance in patients with bladder cancer. The model complements existing predictive frameworks and can contribute to more personalized and effective bladder cancer care. Full article

Background/Objectives: The ascending pharyngeal artery (APA) exhibits considerable variability in origin. Understanding its anatomy is essential for head and neck surgery, endovascular procedures, and skull base approaches. This study aimed to (1) systematically characterize APA origin sites, (2) evaluate bilateral patterns, and (3) establish a comprehensive sequential classification system for external carotid artery (ECA) branching. Methods: Bilateral computed tomography angiography assessment was performed in 85 patients (170 carotid axes; 54 men, 31 women; mean age 69 ± 10 years). APA origins were classified into six types: Type 0 (absent), Type I (ECA medial wall), Type II (ECA posterior wall), Type III (occipitopharyngeal trunk), Type IV (internal carotid artery), and Type V (other origins). A novel sequential classification system (S-types) documented the complete ECA branching order. Results: APA was absent in 14.71% of cases; APA’s absence or internal carotid origin was noted in 19.41% of cases. Type I occurred in 26.47%, Type II in 35.88%, Type III in 17.06%, Type IV in 4.71%, and Type V in 1.18%. Forty distinct S-types were identified, representing the most comprehensive documentation of ECA branching diversity. No statistically significant side-related (χ² = 42.12, p = 0.379) or gender-related (χ² = 49.81, p = 0.138) differences were found. Twenty-three types occurred in fewer than five cases each. Conclusions: This first comprehensive sequential classification system reveals extraordinary anatomical diversity in ECA branching patterns. The absence of predictable side or gender patterns necessitates bilateral preoperative imaging for surgical planning. Full article