High-dimensional feature spaces and severe class imbalance remain fundamental challenges for Machine Learning-based Network Intrusion Detection Systems (ML-NIDS), where minority attack categories are frequently overlooked during feature selection. Existing feature selection approaches commonly rely on global feature relevance measures and manually specified feature counts, which favor majority traffic classes and reduce sensitivity toward rare but critical attack categories. To address these limitations, this study proposes the Adaptive Class-Aware Feature Selection (ACAFS) framework for multi-class intrusion detection. Unlike conventional approaches, ACAFS introduces a data-driven adaptive feature count mechanism based on permutation null hypothesis testing, a Class-Aware Composite Mutual Information scoring strategy that explicitly preserves minority-class discriminative information, and a coordinated two-stage feature selection framework that combines statistical filtering with XGBoost-based model refinement. The framework was evaluated independently on the CSE-CIC-IDS2018 benchmark dataset and a Simulated University Network Environment (SUNE) dataset representing Tanzanian higher learning institution networks. Experimental results demonstrate that ACAFS substantially reduces feature dimensionality while improving balanced intrusion detection performance. On CSE-CIC-IDS2018, ACAFS reduced the feature space from 74 to 22 features, representing a 70.3% dimensionality reduction, while the Two-Stage CNN achieved 99.39% accuracy, 99.40% F1-score, and a false positive rate of 0.09%. The framework further achieved 98.59% recall for Web_Attacks despite severe class imbalance, demonstrating effective preservation of minority-class discriminative features. On the SUNE dataset, ACAFS independently selected 18 features and maintained stable detection performance without dataset-specific manual tuning, confirming its adaptability across heterogeneous network environments. These results confirm that adaptive and class-aware feature selection can simultaneously reduce feature redundancy, improve minority attack detection, and maintain robust intrusion detection performance across diverse network traffic environments.
As machine-learning models for vocal pathology advance, their performance is increasingly constrained not by modelling techniques but by the taxonomic structures used to define the classification task itself. Conventional clinical frameworks, while grounded in diagnostic practice, often reflect conceptual groupings that do not map cleanly onto the acoustic patterns learned by modern Voice AI systems-contributing to the persistent performance gap between multi-class and binary detection tasks. Motivated by this mismatch, we introduce an alternative strategy: deriving a taxonomy from data-driven acoustic relationships rather than prescriptive clinical categories, with the goal of establishing a more model-aligned and generalisable foundation for voice disorder classification. We developed CarLab 2025, a novel data-driven classification framework derived from model confusion patterns. We conducted comprehensive experiments comparing its performance against existing clinical taxonomies, including the hierarchical USVAC 2025 framework, as well as Compton 2022, da Silva Moura 2024, and Za'im 2023, across multiple vocal tasks, features, and model architectures. We evaluated both in-domain performance and cross-database generalisation, including experiments with multi-task learning and targeted data injection. CarLab 2025 achieved superior in-domain classification accuracy compared to established clinical taxonomies, with balanced accuracy reaching 67.20% compared to 61.03% for the best-performing clinical framework. For out-of-domain generalisation, models trained with structured taxonomies consistently outperformed those trained with narrow, single-disorder labels, and training on a diverse set of vocal tasks proved more effective for cross-database performance than relying on a single task. Multi-task learning offered no advantage over single-task training, and while injecting a small amount of data from target domains significantly boosted binary detection accuracy, this improvement did not consistently translate to multi-class recall. Our experiments established a baseline performance exceeding that obtained with existing clinical classification frameworks by aligning more closely with acoustic manifestations of disorders. We further show that exposure to varied recording conditions is crucial for binary generalisation, while robust multi-class generalisation will require substantially more diverse multi-source training data. The results provide a clear, evidence-based path toward developing more robust and generalisable models for vocal pathology detection.
Bronchopulmonary dysplasia (BPD) is a heterogeneous syndrome encompassing multiple biological and structural phenotypes. This study aimed to (1) identify latent phenotypes of in-hospital respiratory trajectories among preterm infants born at ≤ 32 weeks, and (2) compare these data-driven subgroups with traditional BPD severity classifications in predicting in-hospital mortality. We conducted a retrospective cohort of preterm infants born at ≤ 32 weeks' gestation and admitted to a tertiary neonatal unit in Medellín, Colombia (2021-2023). Infants with major malformations or genetic syndromes were excluded. Clinical and respiratory variables were analyzed using latent class analysis (LCA) to identify subgroups based on markers of prematurity, oxygen dependency, and respiratory support. Among 236 infants, a three-class model best described the data (AIC = 12,002; BIC = 12,440; entropy = 0.83). Class 1 (64%) comprised more mature infants with mild respiratory disease and 5% mortality. Class 2 (27%) showed prolonged oxygen therapy and intermediate outcomes. Class 3 (9%), composed mainly of extremely preterm infants, exhibited severe respiratory failure and 88% mortality. The classes reflected a progressive gradient of respiratory severity and mortality. LCA revealed three distinct BPD phenotypes with increasing respiratory severity and mortality, paralleling biological models of parenchymal, interstitial, and vascular injury. This approach supports data-driven phenotyping as a foundation for precision neonatal care.
Dermoscopic skin lesion classification is challenged by class imbalance and the underutilization of clinical metadata, limiting the diagnostic reliability of existing deep learning systems. Building upon a Vision Transformer baseline that fuses dermoscopic images with patient metadata via cross-modal attention, we introduce three complementary improvements: class-balanced focal loss with square-root effective-number sampling, a patch-level cross-attention metadata-guided attention module, and supervised contrastive regularization. The extended model is evaluated on the ISIC 2019 and BCN 20000 datasets. Ablation studies demonstrate consistent improvements. On the ISIC 2019 validation set, the full model achieves a Macro AUC of 0.9613 and a Macro balanced accuracy of 0.8426, with strong performance on rare categories such as dermatofibroma (AUC = 0.991) and vascular lesions (AUC = 0.999). Principled loss design, spatially aware fusion, and representation-level regularization jointly improve the robustness of skin lesion classification under extreme class imbalance.
CT is the primary modality for kidney pathology, but deep-learning evaluation is undermined by slice-level leakage and poor probability calibration. A patient-disjoint, group-stratified hold-out benchmark is established for four-class kidney CT classification (Normal, Cyst, Tumor, Stone) on a 12,446-image multicenter cohort, and NephroNet - a compact (1.46 M-parameter) depthwise-separable CNN with squeeze-and-excitation and a light SpatialGate -b is proposed. A standardized pipeline (320 × 320 preprocessing; annealed MixUp/CutMix; class-weighted AdamW with warmup-cosine; EMA-only evaluation; TTA; post-hoc temperature scaling, T* = 1.42) is reported across accuracy, per-class and micro/macro ROC-AUC, Brier score, and ECE with bootstrap CIs. On the hold-out (N = 2,490), NephroNet attains accuracy 0.9997 (95% CI 0.9984-1.0000), macro-AUC 0.9969 (0.9953-0.9983), Brier 0.0007, ECE 0.0021, surpassing budget-matched CNN and transformer baselines. Transparent splits, explicit capacity control, and calibration-aware reporting support reproducible comparison; all data are single-region, so external and prospective validation is required.
Ductal carcinoma in situ (DCIS) spans a biologic continuum from atypical ductal hyperplasia (ADH) to high-grade lesions with variable risk of progression to invasive ductal carcinoma (IDC), yet morphologic assessment by hematoxylin and eosin (H&E) remains diagnostically limited, particularly at the benign versus ADH/low-grade DCIS boundary. TRPV4, a mechanosensitive ion channel with pathology-dependent subcellular localization in DCIS, offers a biologically motivated immunohistochemical (IHC) marker that may refine classification beyond routine H&E assessment. We tested whether deep learning models trained on TRPV4 IHC outperform H&E-based models across the DCIS progression spectrum. We assembled a multi-institutional cohort of H&E and TRPV4 IHC whole-slide images from 108 patients, comprising an internal development cohort (n = 69), an external test cohort (n = 39), yielding 24,248 annotated tiles. Histopathological tiles from annotated regions were grouped into four ordered classes: normal/benign, ADH/low-grade DCIS, high-grade DCIS, and IDC. Xception and EfficientNet-B0 convolutional neural networks were trained with patient-level three-fold cross-validation on the development cohort and evaluated as ensembles on the external test cohort. On external patient-level testing, H&E ensembles achieved macro-F1 values of 0.43-0.44 and macro-AUC values of 0.73-0.80, whereas TRPV4 IHC ensembles improved performance to macro-F1 values of 0.68-0.72 and macro-AUC values of 0.91-0.92, corresponding to a 54.5-67.4% relative improvement in patient-level macro-F1. Patient-level per-class analyses showed the largest AUC gains with TRPV4 IHC versus H&E for ADH/low-grade DCIS (0.94-0.95 versus 0.61-0.70) and IDC (0.77-0.85 versus 0.61-0.69). Per-class analyses showed the largest gains with TRPV4 IHC versus H&E for ADH/low-grade DCIS (AUC, 0.83-0.84 versus 0.70-0.81) and IDC (AUC, 0.74-0.79 versus 0.65-0.66). These findings support TRPV4 IHC as a mechanistically grounded complement to H&E that improves patient-level discrimination across the DCIS progression spectrum, with the strongest gains for ADH/low-grade DCIS and IDC, in a pilot multi-institutional setting.
Speech disorders (SD) present significant diagnostic challenges due to the complex and indistinct acoustic characteristics embedded within speech signals. The standalone convolutional neural networks (CNNs) and vision transformers (ViT) struggle to capture SD temporal and spectral dynamics. To address these limitations, this study proposes an end-to-end binary pathological SD detection framework that processes raw audio waveforms to differentiate disordered speech from healthy speech while improving feature representation, interpretability, and generalization. A hybrid feature extraction integrating one-dimensional CNNs and ViT's self-attention mechanism is developed to extract crucial SD features. An adaptive fusion and a Class-attention transformer (CaiT)-based feature refinement is introduced to transform the extracted features into a classification-optimized global representation. Through the integration of gradient-weighted class activation mapping (Grad-CAM) and attention-based visualization with the model architecture, the temporal-localization of disorder-relevant acoustic patterns is enabled. Two benchmark datasets are utilized for model's performance evaluation, benchmarking against state-of-the-art CNNs and ViT architectures. The model is trained and internally validated on the SVD dataset using subject-level splitting, while the PD and VOICED datasets are used exclusively for external validation to assess cross-dataset generalization across neurological and heterogeneous pathological voice conditions. The proposed model achieves 97.50% accuracy on both SVD and PD datasets and 95.80% accuracy on VOICED, while maintaining a lightweight design with 5.2 million parameters. Statistical analysis further confirms the results' reliability and significance. The integration of adaptive fusion and transformer-based refinement significantly enhances SD detection performance while ensuring interpretability. The suggested framework offers a sound and proof-of-concept for diagnosing SD, allowing clinicians and speech-language pathologists to make reliable predictions and pay attention to diagnostically significant time intervals.
How to support students' mathematics achievement while also maintaining their mathematics interest has become a central concern. Grounded in Carroll's model of school learning, the present study focuses on two central learning activities of classroom instruction and out-of-class homework to investigate how the factors of time and quality within these activities are associated with students' mathematics achievement and interest. Participants were 2808 eighth-grade students in China (45.5% female) from 76 schools across the nine counties. Students completed a questionnaire that included background information and scales measuring mathematics interest, instructional time, instructional clarity, homework time, and homework quality, as well as a mathematics achievement test. A series of multilevel regression models were conducted to examine the effects of instructional time, instructional clarity, homework time, and homework quality on mathematics achievement and mathematics interest. The study revealed that at the within-school level, mathematics achievement was positively associated with instructional clarity and non-linearly related to instructional time. Higher homework quality reduced the achievement gains of spending 60-90 min on homework compared with spending less than 30 min. Between-school effects were either non-significant or lacked substantial practical meaning. For mathematics interest, instructional clarity and homework quality played a more prominent role than instructional time and homework time. Instructional clarity strengthened the positive association between instructional time and mathematics interest. Moreover, mathematics interest was positively related to homework quality while negatively related to homework time. These findings provide insights into how time and quality in instructional and homework contexts are related to students' mathematics achievement and interest. They suggested that the quality-related factors (instructional clarity and homework quality) are more practically relevant to students' development of mathematics achievement and mathematics interest.
Despite effective interventions, mother-to-child transmission (MTCT) of hepatitis B virus (HBV) remains a concern. While health education is crucial, knowledge gaps and misconceptions persist among HBsAg-positive pregnant women. Traditional variable-centered approaches often treat this population as homogeneous, potentially overlooking distinct cognitive subgroups that require tailored interventions. This study aimed to apply person-centered latent class analysis (LCA) to identify and characterize such latent cognitive profiles regarding HBV transmission knowledge. We conducted a secondary LCA on cross-sectional survey data from 203 HBsAg-positive pregnant women in Taiyuan (2020-2022). Model fit was assessed using BIC and entropy. Derived classes were characterized and compared using Knowledge and Misconception scores (Wilcoxon test) and characterized demographically and clinically. LCA identified two profiles: a "Low-Misconception & High-Knowledge" profile (90.2%) and "High-Misconception & High-Knowledge" profile (9.8%) -a group demonstrating accurate knowledge of true transmission routes alongside persistent endorsement of casual transmission myths. The latter group endorsed numerous casual transmission myths (e.g., via food) despite knowing true routes, with a significantly higher Misconception Score. This group was younger, less reliant on internet information, and higher education was a protective factor against belonging to it. Using LCA, we identified a significant minority profile characterized by the co-occurrence of accurate knowledge and a high burden of casual transmission myths. This demonstrates the critical need for a precision public health approach to move beyond one-size-fits-all education. Screening for this cognitively conflicted subgroup and delivering myth-debunking counseling through trusted channels can optimize MTCT prevention efforts. Our findings provide an empirical basis for developing brief clinical screening tools and implementing tailored patient education during routine antenatal visits.
We characterized the genetic diversity of exon 2 of one major histocompatibility complex (MHC) Class II b chain gene in five non-native populations of the American bullfrog (Aquarana (Rana) catesbeiana)-a highly invasive amphibian. We discovered higher levels of genetic diversity and more signatures of positive selection in the MHC exon relative to a putatively neutral protein coding gene (Cytochrome b). Populations (even those in close geographic proximity) harboured highly varying levels of diversity at both loci, including alleles that have been previously shown to be associated with disease resistance in this species.
Transthoracic echocardiography is widely used to assess the volume status of patients with coronary heart disease combined with heart failure (CHD-HF) because of its noninvasive and convenient nature. Inferior vena cava (IVC) measurement is a commonly used alternative method for evaluating central venous pressure. This study analyzed the specific clinical applications of IVC diameter (IVCD) and collapse index (IVCCI) in assessing the cardiac dysfunction severity and prognosis of CHD-HF patients. A retrospective analysis was conducted on 340 CHD patients treated at our hospital between January 2021 and March 2023. Patients were categorized into CHD (n = 123) and CHD-HF (n = 217) groups. Ultrasound measurements of IVCD and IVCCI were performed, and 2-year follow-up data were compiled to record cardiovascular re-hospitalization and all-cause death. Spearman or Pearson was utilized for correlation analysis. ROC analysis was used for performance analysis, and Kaplan-Meier and Cox regression were harnessed for survival analysis. CHD-HF patients showed increased IVCD and decreased IVCCI. CHD-HF patients experiencing cardiovascular re-hospitalization or all-cause death exhibited elevated IVCD and reduced IVCCI. IVCD (AUC = 0.715, 95%CI = 0.638-0.792, cut-off = 14.15 mm; AUC = 0.795, 95%CI = 0.732-0.858, cut-off = 13.04 mm) and IVCCI (AUC = 0.790, 95%CI = 0.726-0.853, cut-off = 51.47%; AUC = 0.731, 95%CI = 0.661-0.801, cut-off = 50.00%) demonstrated high predictive values for cardiovascular re-hospitalization and all-cause death in CHD-HF patients. IVCD (HR = 1.124, 95%CI = 1.056-1.196; HR = 1.139, 95%CI = 1.078-1.203) and IVCCI (HR = 0.959, 95%CI = 0.921-0.999; HR = 0.956, 95%CI = 0.921-0.992) were both independently associated with cardiovascular re-hospitalization and all-cause death in CHD-HF patients. IVCD and IVCCI effectively predict cardiovascular re-hospitalization and all-cause death in CHD-HF patients.
Plant GATA transcription factors (TFs) constitute a conserved family of zinc finger proteins characterized by a type IV zinc finger motif (CX2CX18-20CX2C) that binds to the WGATAR (W = A/T, R = A/G). Given their essential roles in plant growth, development, and environmental adaptation, GATA TFs have been extensively investigated through genome-wide analyses across diverse plant species. However, despite its economic importance as a major root crop, the carrot (Daucus carota subsp. sativus) has not yet undergone a comprehensive genome-wide analysis of its GATA TF family. To address this gap, this study conducted a comprehensive in-silico characterization of the GATA TF family in the carrot genome. A total of 33 DcGATAs were classified into four classes, with Classes A and B containing relatively more members than Classes C and D. Each class displayed distinct conserved gene, domain, and motif structures. Chromosome distribution and collinearity analysis revealed that the 33 DcGATAs were unevenly distributed across the nine carrot chromosomes and formed 16 intergroup syntenic gene pairs, indicating that segmental duplication may have contributed to the expansion of the gene family. Secondary and tertiary structure analyses indicated that DcGATA genes were mainly enriched in random coils with lower proportions of α-helices, extended strands, and β-turns. Protein-protein interaction analysis suggested a broad interaction network among DcGATA genes at low confidence, with only a limited number of interactions detected at medium confidence. Cis-acting element analysis revealed 1,106 regulatory elements in DcGATA promoters, predominantly associated with light, stress, and hormone responses, with class-specific variation in element abundance. RNA-seq expression profiling showed that only a subset of DcGATA genes exhibited significant variation across developmental stages, root colors, and their interactions, whereas most genes maintained stable expression in 40- and 80-DAP yellow and orange roots. MiRNA target prediction revealed 45 miRNAs targeting 15 DcGATA genes, mainly involving MIR159, MIR165, and MIR396 families, with class-specific cleavage- and translation-mediated regulatory patterns. These results provide comprehensive insights into the evolution and functional diversification of DcGATAs by establishing an in-silico integrative regulatory framework; however, further experimental validation is required to elucidate their precise roles in development and stress adaptation.
Major depressive disorder (MDD) includes heterogeneous clinical dimensions, including depressive symptom severity, treatment refractoriness, and suicidality, which are commonly assessed using subjective rating scales and retrospective clinical histories. Dysfunction of frontal and anterior cingulate cortex (ACC) networks has been implicated in MDD, suggesting that electroencephalography (EEG)-based approaches combined with machine learning (ML) may help with objective characterization of clinical heterogeneity. Resting-state and cognition-modulated EEG data were analyzed from 209 patients with MDD. A rostral ACC-engaging cognitive task (RECT) was used to probe frontal-ACC circuitry. Linear and non-linear EEG features extracted from frontal electrodes across multiple frequency bands were integrated with several ML classifiers to perform exploratory classification of suicidality, depressive symptom severity, and treatment refractoriness. Class imbalance in suicidality was addressed using synthetic oversampling applied to the training data only. ML models, particularly Random Forest (RF), outperformed support vector machines across all outcomes. RF achieved classification accuracies of around 83% (area under curve (AUC) = 0.83) for depression severity and 87% (AUC = 0.87) for treatment refractoriness. Suicidality categorization performance improved following data balancing. Feature importance studies found consistent patterns across outcomes, with useful predictors primarily obtained from frontal electrodes and nonlinear EEG complexity measures. The combination of cognition-engaging frontal modulation and ensemble-based ML applied to EEG data suggests the feasibility of an exploratory, unified EEG-based approach for identifying important characteristics of MDD. These findings highlight the importance of frontal network dysfunction across the severity spectrum of MDD and the need for further validation in larger and longitudinal cohorts.
Patients with phenotypically mild hypertrophic cardiomyopathy (HCM) do not require symptom management, but may be at an earlier stage in the disease course, with potential to benefit from disease-modifying therapies. However, little is known about the natural history and predictors of major adverse cardiovascular events (MACE). Using the Sarcomeric Human Cardiomyopathy Registry, we identified predictors of incident MACE and characterized disease progression in phenotypically mild HCM. Phenotypically mild HCM was defined as: having shorter disease duration (<10 years since diagnosis or age ≤30 years), no previous MACE, being NYHA functional class I, and having a left ventricular (LV) maximal wall thickness (MWT) <25 mm. These individuals were followed prospectively for the development of symptoms or MACE: atrial fibrillation (AF), malignant ventricular arrhythmia (MVA) (sudden cardiac death, resuscitated arrest, or appropriate defibrillator therapy), heart failure (HF) (cardiac transplantation, LV assist device implantation, LV ejection fraction <35%, or NYHA functional class III or IV symptoms), stroke, or all-cause mortality. Cox regression identified MACE predictors. Linear and latent class mixed models characterized LV remodeling trajectories and risk clusters. Of 2,500 participants with phenotypically mild HCM (mean age 43 years, 31% women) followed for a mean duration of 7 ± 6 years, 534 (21%) developed MACE, including 289 with AF, 69 with MVA, and 193 with HF. Individuals who progressed from NYHA functional class I to ≥ II symptoms during follow-up (n = 585, 23%) were 2.79 times (95% CI: 2.30-3.39 times) more likely to experience MACE. Age at baseline (HR: 1.24; 95% CI: 1.17-1.32 per 10-year increase), body mass index (HR: 1.10; 95% CI: 1.01-1.21 per 5-kg/m2 increase), left atrial (LA) diameter (HR: 1.16; 95% CI: 1.09-1.25 per 5-mm increase), LV MWT (HR: 1.27; 95% CI: 1.10-1.46 per 5-mm increase), and LV outflow tract (LVOT) gradient (HR: 1.08; 95% CI: 1.05-1.12 per 15-mm Hg increase) associated with higher MACE rates. LV late gadolinium enhancement presence was associated with 36% (95% CI: 5%-76%) higher hazard of MACE. Remodeling trajectories during follow-up predicted risk with each 0.5 mm/year steeper increase in LA diameter associating with doubled AF (HR: 2.24; 95% CI: 1.69-2.97) and HF rates (HR: 2.22; 95% CI: 1.62-3.04) and each 0.5 mm/year steeper LV MWT increase associating with doubled MVA rates (HR: 1.92; 95% CI: 1.38-2.69). Higher sustained values and/or steeper increases in LA diameter, LV MWT, or LVOT gradient associated with the highest MACE rates. Approximately 21% of patients with phenotypically mild HCM developed MACE over medium-term follow-up. Older age, symptoms development, and increasing LA diameter, LV hypertrophy, or LVOT gradient associated with MACE, particularly in instances of steeper rate of change. These findings can guide management strategies and inform future studies of disease-modifying therapies.
Adolescent endometriosis (EMs) imposes substantial physical and psychological burdens on young female patients, and adaptive coping strategies are vital for their health management. This study aimed to explore the latent classes of parental rearing patterns and coping styles among adolescent females with endometriosis, and examine the interrelationships between parental rearing, psychological resilience and coping styles. The findings intend to provide theoretical evidence and practical references for mental health intervention and public health services targeting this vulnerable group. This cross-sectional study consecutively recruited participants from 2024 to 2026 at a tertiary Class A hospital in Shenyang. A total of 168 adolescents aged 12-20 years were enrolled via convenience sampling. Parental Rearing Patterns Scale, the Resilience Scale, and the Coping Styles Scale were used to collect relevant data. Latent class analysis (LCA) was adopted to classify parental rearing patterns and coping styles. Independent samples t-test, one-way ANOVA and bias-corrected Bootstrap mediation analysis (5,000 resamples) were further performed for statistical testing. LCA identified two parenting subtypes (31.5% positive-empowering, 68.5% high-pressure restrictive) and three coping subtypes (58.9% active problem-solving, 28.0% passive-avoidant, 13.1% social support-seeking). Psychological resilience differed significantly across parenting patterns and coping styles (P<0.05). Mediation analysis revealed that high-pressure restrictive parenting was negatively associated with psychological resilience, and resilience was correlated with coping styles (P<0.05). The indirect effect was 0.295 (P<0.05), and the direct effect was non-significant. As this was a cross-sectional study, the findings only indicate statistical indirect associations rather than causal mediation. Different parenting styles and coping patterns lead to varied psychological and clinical outcomes. Positive-empowering parenting and active coping facilitate patients' psychological adjustment and disease management. By contrast, high-pressure restrictive parenting, avoidant coping and low willingness to seek help tend to trigger negative emotions, reduce treatment adherence and delay intervention. Parenting styles affect coping behaviors partly through psychological resilience. Targeted improvement of psychological resilience can effectively enhance patients' ability to cope with the disease.
Maternal-fetal cardiac coupling has emerged as a promising non-invasive marker of fetal autonomic regulation and neurocardiac development. However, existing approaches are limited by unidirectional modeling, handcrafted coupling metrics, and a lack of interpretable and clinically validated abnormality assessment frameworks. We propose a transformer-based deep learning framework for bidirectional maternal-fetal cardiac coupling analysis using heart rate variability (HRV) features. The model learns fetal-to-maternal (FTM) and maternal-to-fetal (MTF) interactions using two independently trained UNETR-based models with identical architecture and no shared weights. The framework is evaluated on a real clinical dataset, incorporating time-delay, gestational, and classification analyses. Time-delay analysis revealed directional asymmetry, with MTF coupling peaking at 3 seconds and FTM at 5 seconds. Gestational analysis showed relatively stable MTF coupling, while FTM coupling was consistently reduced in abnormal cases. Classification using support vector machines demonstrated strong discrimination, with the full-feature model (MTF + FTM + gestational week) achieving ROC-AUC = 0.91 ± 0.08. Under class imbalance, the coupling-only model (MTF + FTM) achieved the highest PR-AUC, indicating robust minority-class sensitivity. These findings suggest that bidirectional coupling analysis may provide useful insight into maternal-fetal physiological interactions. However, the results should be interpreted as proof-of-concept evidence due to the limited and heterogeneous abnormal cohort. Further validation on larger and more homogeneous datasets is required to assess the potential of this approach for non-invasive prenatal monitoring.
Fry, AC, Schechter, EGK, Johnson, QR, and Cabarkapa, D. Tons of resistance exercise research, but does it apply to highly trained athletes? J Strength Cond Res XX(X): 000-000, 2026-The volume of scientific research reviews on resistance exercise has increased markedly in recent years. The purpose of this report was to determine what subject populations have been studied to determine whether the findings are relevant to highly trained and elite athletes. A representative sample of 26 highly cited systematic reviews and meta-analyses on resistance exercise topics from the most recent 15 years (2010-2024) were analyzed to determine the subject characteristics of each study cited in the reviews. All reviews included dependent variables of strength or hypertrophy, and dependent variables of exercise choice, order, intensity [load], volume, interset rest, frequency, or lifting tempo. Subjects were classified with a 6-tiered scale ranging from sedentary (tier 0) to world class athlete (tier 5). Of 18,574 subjects reported in the selected reviews, only 2.3% were classified as highly trained or elite athletes. Similar distributions were observed when reviews were analyzed separately for each resistance exercise topic. Cumulatively, the reviews included have been cited by scientific and lay literature more than 7,000 times to date, indicating how effectively their findings are being disseminated. Although each of the reviews cited may provide helpful information for the general population for health and fitness purposes, the findings are not primarily based on the study of highly trained elite athletes, and does not consider training performed outside the weight room, training histories, competition schedules, or the specific demands of their sports. Several practical suggestions are presented that may facilitate research access to highly trained or elite athletes.
Trauma exposure and structural inequities contribute to mental health disparities among Black men and create barriers to care. Black men face elevated risk from opioid and stimulant co-use, including overdose from unintentional opioid exposure, yet heterogeneity in trauma exposure among men who co-use substances is understudied. Using latent class analysis, this study identified unobserved trauma subgroups and examined associations with cocaine use in a sample of Black men co-using opioids and stimulants; most were formerly incarcerated (88%). Three profiles emerged: Low Trauma (36%), High Non-Sexual Trauma (42%), and High Trauma Across All Categories (21%). In regression models, men in the High Trauma Across All Categories class had higher odds of lifetime powder cocaine use (OR=4.13, p<.01) and crack cocaine use (OR=2.08, p<.05) versus the Low Trauma class. Older age was associated with stimulant use, while incarceration history predicted powder cocaine use only. Findings underscore within-group heterogeneity and support trauma-informed interventions to reduce overdose risk among Black men who co-use opioids and stimulants. Targeted screening and tailored services may strengthen engagement in care and harm reduction.
Olea europaea L. (olive tree) stands as an enduring emblem of Mediterranean culture and a treasure trove of structurally diverse phytochemicals with profound biological potential. This review presents a structured, data-driven overview of over 300 metabolites previously reported across various olive sources, including leaves, fruits, olive oil, pomace, flowers, etc. These compounds are systematically classified into 15 chemical classes, each described through its basic skeleton and position numbering, highlighting structural features that allow clear differentiation between closely related compounds. Secoiridoids emerge as the dominant class, with oleuropein and ligustroside serving as key representatives, and biosynthetic intermediates such as tyrosol, hydroxytyrosol, and elenolic acid playing central roles. The distribution of metabolites across organs revealed that leaves represent the most abundant source, followed by fruits and olive oil, while pomace also contained a considerable amount, emphasizing its potential value as a by-product for future exploitation. For each metabolite, comprehensive chemical identifiers (PubChem ID, SMILES, InChI, formula, and exact mass) and, for the first time, an analytical confidence level (validated via a five-level scoring system) are provided in the SI. This analysis reveals a predominance of moderately validated compounds (Level 3), highlighting a critical need for further structural elucidation. Collectively, this robust and cheminformatics-ready dataset serves as an accessible resource, poised to accelerate future studies in virtual screening, molecular docking, and network pharmacology. It also critically guides efforts to expand structural validation for low-confidence compounds and encourages exploration of underrepresented organs, thereby significantly enriching the landscape of therapeutic mapping. Additionally, this review integrates recent insights into isolation yields, impurity profiles, and toxicological aspects of Olea europaea metabolites, thereby providing a holistic framework for their therapeutic evaluation and safe exploitation.
Dilated cardiomyopathy (DCM) is a refractory cardiac disease with significant morbidity and mortality. Although immune adsorption combined with immunoglobulin G replacement therapy (IA/IG) has shown potential in treating DCM, only small-scale clinical trials have been reported. Its efficacy and safety characteristics still need to be further systematically evaluated. We conducted a systematic review and meta-analysis in accordance with PRISMA 2020 guidelines. A comprehensive search of PubMed, Embase, Web of Science, and the Cochrane Library was performed up to July 1, 2025. Clinical studies on IA/IG treatment for DCM were included. The primary outcome was the left ventricular ejection fraction (LVEF) change. Secondary outcomes included left ventricular end-diastolic dimension (LVEDD), NYHA functional class, N-terminal pro-brain natriuretic peptide (NT-proBNP), and VO2 peak. Study quality and risk of bias were assessed using the Cochrane ROB 2.0 and ROBINS-I tools. Sensitivity analysis was taken into consideration to determine the stability of the results. This review was registered with PROSPERO (CRD420251104796). Eighteen studies (2 RCTs and 16 non-RCTs) involving 809 participants were included. IA/IG therapy significantly improved LVEF [mean difference (MD) = 7.71%, 95% CI: 6.18-9.24, p < 0.00001] and reduced LVEDD (MD = -3.22 mm, 95% CI: -4.16 to -2.28, p < 0.00001) from baseline. Significant improvements were also observed in NYHA functional class (MD = -0.76, 95% CI: -0.91 to -0.60, p < 0.00001) and peak VO₂ (MD = 2.66 mL/kg/min, 95% CI: 1.26-4.06, p = 0.0002). Compared to controls, the IA/IG group demonstrated greater improvement in LVEF (MD = 8.31%, 95% CI: 6.45-10.18, p < 0.00001) and NYHA class (MD = -0.62, 95% CI: -1.00 to -0.25, p = 0.001). A sensitivity analysis of the results suggested that they were stable. Systematic reviews and meta-analyses suggest that IA/IG therapy may improve cardiac function and quality of life in patients with DCM. However, the number of RCTs included in the study is limited, so these results should be interpreted with caution. Further high-quality, large-scale trials are warranted to establish standardized treatment protocols and confirm the long-term benefits of IA/IG therapy. https://www.crd.york.ac.uk/PROSPERO/recorddashboard, PROSPERO CRD420251104796.