Medical image registration is an important task in medicine for providing an accurate anatomical correspondence between multimodal images, which is essential for diagnosis, treatment planning, and longitudinal follow-up. Nevertheless, the natural variability of biological shapes as well as the absence of annotated data in unsupervised settings make conventional registration-related algorithms extremely challenging to apply. In this work, we introduce a new Recursive Deformable Pyramid Network (RDPN) for the task of unsupervised medical image registration that aims to model both global and local deformation fields by hierarchically incorporating multi-scale feature representations. The adopted network consists of a deformable convolutional backbone that is used recursively on the pyramid level, where such instantiation allows us to estimate the adaptive spatial transformation without inheriting ground truth correspondences. Our method is tested on a synthetic brain MRI dataset which was meticulously generated to simulate inter-patient and intra-patient anatomical variability and a real abdominal CT dataset. Experiments show that RDPN achieves a robust performance gain in comparison with the state-of-the-art methods in the aspects of Dice similarity coefficient, target registration error and deformation smoothness. The pyramid mechanism in recursive level is especially effective for the purpose of aligning fine-grained anatomical structures with global structure consistency. Moreover, a thorough ablation study demonstrates the importance of recursive feature fusion and deformable modeling in learning a robust unsupervised registration solution. This work contributes by presenting a pragmatic approach to scalable registration for difficult registration problems in medical imaging, which could elevate the quality of the clinical workflow downstream.
Chronic psychological suffering may persist despite adequate diagnosis, evidence-based treatment and preserved cognitive insight. In clinical practice, this persistence is often interpreted as treatment resistance, symptom severity or insufficient adherence. However, such explanations may fail to capture the structural organization through which suffering becomes stabilized over time. This theoretical article proposes an alternative conceptualization of psychopathological chronicity as a structural mode of organization sustained by recursive affective-symbolic loops, rather than by the mere persistence of symptoms. It aims to explain the paradox of insight without recovery and to clarify how chronic suffering may acquire a homeostatic function. The article integrates concepts from clinical psychopathology, phenomenology, affective neuroscience and predictive processing. It also draws on converging evidence from clinical domains in which psychological status, perceived health and self-regulatory appraisals modify relationships among symptom reports, objective findings, functional impairment and somatic outcomes. Clinical vignettes are used to illustrate the proposed model. The proposed framework suggests that cognition may become functionally recruited in the stabilization of chronic suffering. Within this organization, chronicity can operate as a form of "pathological health", protecting the subject from greater psychic fragmentation while simultaneously maintaining distress. The model also highlights how therapeutic interactions may inadvertently reinforce recursive loops when interventions remain confined to cognitive insight, reassurance or premature disruption. Psychopathological chronicity may be better understood as an affective-symbolic architecture that organizes symptoms, self-appraisal, embodiment and relational patterns. Structural change may therefore require embodied and transferential interventions capable of disrupting the recursive core, together with operational markers for distinguishing timely therapeutic destabilization from premature disruption. This model is primarily applicable to adult and older adolescent populations and may contribute to a more nuanced evaluation of chronicity in clinical practice.
Digital fluoroscopy offers high temporal resolution but is limited by substantial noise. Recursive filtration improves the signal-to-noise ratio (SNR) by combining information across frames, though potentially at the cost of motion blur and contrast loss. This study evaluates recursive filtering performance in a commercial angiographic system, specifically where filter parameters (e.g. K-factor) are not explicitly disclosed. Fluoroscopic images were acquired using a rotating phantom with high- and low-contrast objects. Image quality was assessed using noise power spectrum (NPS), contrast-to-noise ratio (CNR), and lag analysis. Manual K-factor settings were compared with automatic configurations, and a model was developed to estimate the effective K-factor based on NPS characteristics. Higher K-factors yielded substantial noise reduction but increased lag and decreased contrast. Automatic filtering modes provided a more balanced outcome, achieving noise suppression comparable to high K-factor values while preserving contrast and minimising motion artefacts. Estimated K-factors from automatic modes varied with temporal frequency, reflecting adaptive filter behaviour. Understanding the effects of undisclosed filter settings is critical for protocol optimisation. These findings support adaptive filtering strategies tailored to clinical tasks, enhancing imaging performance and patient safety.
An adaptive fractional-order model predictive control (FO-MPC) framework of DC-DC boost converters, which incorporates the Exponential Recursive Least Squares (ERLS) identification, the use of the fractional-order dynamics, and the application of the Grey Wolf Optimization (GWO) is presented in this paper. An important discovery is that the combination creates synergistic effects: ERLS convergence is improved by 47% compared to standalone implementations, since the damping of adaptation transients is of fractional-order-damping-like-density, which previous combined methods (such as Wang et al. (2020)) or (Ghamari et al. (2025)) did not provide. The ERLS algorithm allows adaptation model free and convergence in 15 samples even without the use of exact mathematical models. An optimized α = 0.85, fractional-order operator in the noise rejection case and better stability margins is observed, which is 15dB more than the traditional MPC implementations. GWO, executed offline during commissioning, achieves 25 to 30 times faster convergence than conventional metaheuristics (GA, PSO) when tuning the controller parameters. Arduino DUE (84 MHz ARM Cortex-M3) hardware validation has shown that settling time is significantly decreased to 0.42s (83% lower than the baseline), that overshoot is kept to less than 1% (95% lower than the baseline), and that steady-state error is only 20mV (87% smaller than the baseline). The controller is stable in the 30% variations in parameter and 10 times changes in load with an execution time of 85µs, which is compatible with 10 kHz control frequency. Monte Carlo simulations (n = 1000) confirm a success rate of 98.2% in combined disturbances, and statistical significance is validated using the Wilcoxon signed-rank tests (p < 0.001, Cohen's d > 2.0). The industrial use has been tested and supported with 168 h continuous operation and IEC 61000-4-3 EMI compliance test.
Internalized racism, although widely studied, remains conceptually fragmented and theoretically underdeveloped. I advance the internalized racism process model (IRPM), a process framework that conceptualizes internalized racism as a recursive, motive-driven system of self-regulation through which structurally organized ethnoracial hierarchy becomes embedded in psychological functioning. The IRPM specifies how structural conditions, sociocultural socialization, and identity processes converge to organize internalized racism. The model identifies a functional-motivational architecture in which needs for self-integrity and belonging give rise to three regulatory functions-threat minimization, ideological acceptance, and identity translation-that structure diverse forms of internalized racism and their behavioral enactments. Through recursive feedback loops, these processes stabilize into enduring patterns of self-regulation. The IRPM integrates previously fragmented literatures by distinguishing forms from manifestations, linking motives to regulatory functions, and specifying cross-level dynamics between structural contexts and individual regulation. It explains variability in who internalizes (via vulnerability and resistance resources), when internalization is likely (under conditions of threat and belonging pressure), and why consequences differ (as a function of form and configuration). Finally, the model specifies how strategies that provide short-term regulatory relief can consolidate into entrenched patterns that undermine psychological functioning. Although designed to generalize across hierarchically organized systems, the IRPM centers on internalized racism among subordinated groups, where empirical foundations are most developed, while also providing a framework for examining internalized racial dominance among advantaged groups. The IRPM offers a unifying account that advances explanation, generates testable predictions, and establishes an agenda for future research on the psychological organization of inequality. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
Addressing society's most pressing problems requires sustained coordination among stakeholders across organizational boundaries. A central dilemma of such grand challenges is how to sustain engagement among actors who hold potentially incommensurate objectives, priorities, and values while pursuing a shared vision of social impact. Yet existing scholarship reveals little about how novel organizing forms are built and maintained throughout grand challenge initiatives and even less about the iterative cycles of interactions among their stakeholders. We examine this dilemma through a 3-year action research study of a multistakeholder initiative focused on eliminating mental health stigma. We draw on participant observation, interviews, and archival data to develop a grounded model of community weaving-an organizing approach that enables sustained engagement among heterogeneous actors under conditions of complexity, uncertainty, and evaluative heterogeneity. Community weaving combines a central coordinating architecture with semiautonomous substructures that support partial alignment around a multivocal shared purpose while enabling distributed, locally meaningful action. Rather than resolving tensions in stakeholder goals, values, and priorities, the initiative sustains participation by recursively reworking persistent tensions as it grows. Our model and theorizing advance research on organizing for social impact by showing how collaboration can be maintained without full convergence or centralized control. In doing so, we shift attention from consensus and governance toward partial alignment, multivocality, and recursive coordination as mechanisms for sustaining collective action in complex social systems. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
Open neuroscience repositories support reuse by making datasets accessible, but event level reuse also depends on whether task and stimulus annotations can be interpreted by software. I audited public latest OpenNeuro BIDS snapshots using public GraphQL metadata, recursive file trees, small events.json sidecars, and bounded events.tsv header ranges. Raw neural data were left untouched. Among 1,713 public latest snapshots, 1,483 had task metadata or observed event TSV files and formed the primary event relevant denominator. Event TSV files were present in 1,175/1,483 snapshots (79.2%; Wilson 95% CI 77.1%-81.2%). Candidate event JSON sidecars were present in 604/1,483 snapshots (40.7%), but an inheritance aware path and entity check found applicable JSON sidecars for 590/1,483 (39.8%), descriptive applicable sidecars for 550/1,483 (37.1%), and experiment specific applicable sidecars for 491/1,483 (33.1%). Across the full recursive file tree, 162,034/301,681 event TSV files (53.7%) had an applicable event JSON sidecar, and 147,412/301,681 (48.9%) had an applicable experiment specific sidecar. HED was detected in event JSON for 45/1,483 snapshots (3.0%) and in sampled TSV headers for 2/1,483 (0.13%). EEG had higher sidecar coverage and HED detection than fMRI, but 21.2% of EEG candidate JSON files were bookkeeping only. These results identify a repository visible metadata gap: event timing is common, but software interpretable event meaning remains uneven.
Worldwide, anemia in children under-five is a major public health issue, particularly in sub-Saharan Africa. Sub-Saharan Africa also has the highest burden of malaria. This study aimed to develop an ensemble machine learning model to estimate anemia burden and potential predictors in under-five children in malaria-endemic sub-Saharan African countries. A cross-sectional study was conducted using Demographic and Health Survey data from sub-Saharan African countries. Samples were selected through a two-stage stratified cluster sampling method. Data analysis was performed using Python 3.8, with a total weighted sample of 21,249. The dataset was split into 80% for training and 20% for testing and validation purposes. To address class imbalance, a hybrid data balancing approach combining SMOTE (Synthetic Minority Over-sampling Technique) and Tomek Links was applied. Four machine learning algorithms were developed and evaluated using standard performance metrics. Recursive Feature Elimination with a Random Forest classifier was used to identify potential predictors of anemia among children under five living in malaria-endemic SSA countries. In this study, XGBoost showed the best performance, achieving an accuracy of 83.69%, a precision of 85.81%, and an F1 score of 83.19%. Additionally, XGBoost attained the highest ROC AUC of 90.1 and Precision Recall AUC of 90.0. According to Recursive Feature Elimination with a Random Forest classifier, region, birth order, child age, wealth index, and number of mosquito nets were identified as the associated factors of anemia among under-five children in malaria-endemic SSA countries. To reduce anemia among under-five children in malaria-endemic regions of sub-Saharan Africa, interventions should prioritize implementing geographically targeted programs, focus on younger children and those with high birth orders by integrating anemia screening into routine check-ups. In addition, enhancing economic support for low-income families and distributing and educating families on the proper use of mosquito nets are essential.
Brain-computer interfaces (BCIs) based on selective auditory attention aim to restore communication by decoding selective attention from auditory evoked potentials. Clinical translation of such BCIs requires maintaining sufficient decoding performance despite brain-state non-stationarity.

Approach: We compared classifiers across four evaluation settings: offline baseline classifiers using shuffled 5-fold cross-validation; a causal classifier using a chronological 20%/80% calibration/test split; simulated real-time deployment with a static classifier calibrated on 20 trials; and simulated real-time deployment with an adaptive recursive least squares (RLS) classifier, evaluated within-subject and in a leave-one-subject-out (LOSO) setting. The analysis used 62-channel electroencephalography recorded from 25 healthy adults (18 retained after artifact rejection).

Main results: The best offline baseline classifier, logistic regression with point-to-point features, achieved a mean ROC AUC of 0.75 and an estimated information transfer rate (ITR) of 2.46 bits/min, derived from ROC AUC via a conservative heuristic. Under causal application, performance decreased to ROC AUC = 0.63 and ITR = 0.68 bits/min. In simulated real-time deployment, static classification dropped further to ROC AUC = 0.51, whereas adaptive RLS improved ROC AUC to 0.68 and ITR from 0.14 bits/min to 1.42 bits/min (p < 0.001, Cohen's d > 1.49). In the LOSO setting, RLS achieved ROC AUC = 0.57 and ITR = 0.86 bits/min. The LOSO result further suggests that zero-calibration deployment is feasible, with personalization occurring trial-by-trial.

Significance: Brain-state non-stationarity is a major driver of performance decline in auditory BCIs. Lightweight adaptive recalibration substantially restores real-time performance and supports the translational potential of ERP-based communication paradigms.
Non-adherence to medication represents an important global challenge that compromises patient outcomes and increases healthcare costs, particularly in Spain due to the high prevalence of chronic conditions. Therefore, identifying the key factors influencing adherence is a valuable approach for developing targeted interventions. This study analysed two large real-world primary care databases from Madrid and Catalonia using four feature selection methods and three machine learning classifiers, together with threshold optimisation, calibration analysis, bootstrap confidence intervals, and importance analyses. Recursive Feature Elimination with Cross-Validation (RFECV), a cross-validated procedure that iteratively removes less informative variables) combined with Extreme Gradient Boosting (XGBoost), a tree-based algorithm that combines multiple decision trees, achieved the best performance in both cohorts. Overall, 48 structural factors were identified, 19 in Madrid and 29 in Catalonia, with consistent validation and test performance (AUROC 0.6953/0.6952 and 0.7775/0.7788, respectively). In Madrid, the number of medications and chronic disease burden were the most relevant factors, whereas in Catalonia smoking-related factors, rural or urban context, and prescription-timing factors were important. These findings support the value of using artificial intelligence to identify patterns and develop patient-centred adherence strategies in clinical practice. Although such data-driven AI models can reveal useful patterns, their interpretation should remain grounded in comprehensive adherence frameworks and and the results should be interpreted considering the heterogeneity between the databases, indirect adherence measures, and the absence of richer social determinants or external validation. Future research should therefore address these limitations to strengthen the generalisability of the findings.
There is a difficulty reconciling the anti-humanistic approach of systems theory (as developed by Niklas Luhmann) with methodologies centering the individual. Society is viewed as a constellation of systems - politics, law, economy, media - and not the agglomeration of billions of individuals. This fore fronting of systems, and the unknowability of the cognitive processes of individuals, makes empirical operationalization of systems theory difficult. Can an interview method ever be accepted as a robust empirical operationalization of a theory which views society as a collection of intertwined and self-reinforcing systems, recursively communicating according to their own internalized contingencies? This paper attempts a theoretical justification for a methodology where individuals (researcher and interviewees) observe a system from which, theoretically, the human agent is excised and the focus is only on the construction of systemic possibilities. Reconciling theory and method is an important step in recognizing the possibilities of systems theory as a way of examining society.
This study aimed to identify and validate aging-related genes (ARGs) implicated in multiple myeloma (MM), thereby advancing the understanding of the molecular mechanisms underlying the disease. mRNA expression data were retrieved from the Gene Expression Omnibus, and used as a training set to identify ARGs. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes enrichment analyses were conducted to explore the functional roles of the ARGs in MM. Candidate genes identified through least absolute shrinkage and selection operator (LASSO) logistic regression and support vector machine recursive feature elimination (SVM-RFE) were compared using a Venn diagram, which revealed the overlap between the genes identified by the two algorithms. A receiver operating characteristic curve was generated based on the screening results. The candidate genes were further validated using the GSE39754 and GSE5900 datasets. Real-time quantitative polymerase chain reaction (RT-qPCR) was employed to validate the mRNA expression of key genes. Compared to normal individuals, patients with MM exhibited differential expression of 19 genes, of which 5 were upregulated and 14 downregulated. A total of 6 candidate genes (TXN, JUN, FOS, HIF1A, CAT and KCNA3) were identified through LASSO regression and SVM-RFE screening. Among them, TXN exhibited the most significant differential expression, suggesting its potential as a diagnostic biomarker for MM. In addition, in vitro RT-qPCR analysis confirmed that the mRNA levels of TXN in MM cells aligned with the bioinformatics findings, showing higher expression compared to normal B-lymphocyte cell lines. In conclusion, this study identified age-associated molecular patterns in MM and highlighted the diagnostic potential of TXN, offering novel insights for clinical applications in MM.
Host metabolic programs are increasingly recognized as key modulators of tuberculosis (TB) immunopathology, yet manganese metabolism-related transcriptional signatures with diagnostic and immunological relevance remain insufficiently characterized. We aimed to identify manganese metabolism-linked biomarkers for TB diagnosis and to delineate their immune and regulatory context. Three Gene Expression Omnibus microarray datasets were analyzed. A curated Mn metabolism gene set was retrieved from GeneCards. Candidate diagnostic genes were prioritized by integrating Least Absolute Shrinkage and Selection Operator regression (glmnet), Support Vector Machine-Recursive Feature Elimination (caret), and Boruta feature selection. Diagnostic performance was assessed by Receiver Operating Characteristic analysis (pROC), and an eight-gene diagnostic model was visualized by a nomogram with calibration and Decision Curve Analysis. Immune landscapes were profiled using CIBERSORT (IOBR) and single-sample Gene Set Enrichment Analysis (Gene Set Variation Analysis, GSVA). We performed single-gene Gene Set Enrichment Analysis (GSEA) using MSigDB Gene Ontology and Kyoto Encyclopedia of Genes and Genomes gene sets to infer functional programs associated with each candidate gene. Non-negative Matrix Factorization clustering defined TB molecular subtypes, and a competing endogenous RNA (ceRNA) network was constructed using miRTarBase miRNA-mRNA interactions combined with curated lncRNA-miRNA links. From 241 DEGs in GSE83456, 30 Mn-DEGs were obtained and were enriched in immune and infection-related pathways. Machine-learning integration converged on eight diagnostic genes. In training, individual AUCs ranged from 0.924 to 0.975 and the combined model reached 0.995; external validation achieved AUC = 1.000 in both cohorts. Immune analyses revealed TB-associated shifts toward innate/inflammatory signatures and gene-immune correlations, while single-gene GSEA highlighted host-defense, lysosomal, and innate signaling pathways. NMF identified two TB subtypes with distinct immune compositions and immune functional scores. The ceRNA network comprised 8 mRNAs, 28 miRNAs, and 65 lncRNAs. An eight-gene manganese metabolism-associated signature enables accurate TB discrimination and captures immune heterogeneity, providing a framework for biomarker-guided stratification and mechanistic hypothesis generation.
Metabolic dysfunction-associated fatty liver disease (MAFLD) affects 38.9% of the global adult population and is associated with increased mortality when it co-occurs with depression. Women exhibit a 1.5- to 3-fold higher prevalence of depression, with postmenopausal hormonal imbalances amplifying susceptibility. This underscores the urgent need for sex-specific predictive models. The aim of this study was to develop a lightweight, high-accuracy model to identify key predictors of depression risk in female MAFLD patients using a five-year longitudinal UK Biobank cohort. We analyzed routine blood biomarkers (hematology and metabolites), lifestyle factors, and reproductive history from female participants with MAFLD identified within the UK Biobank cohort. Logistic regression was adjusted for socioeconomic, lifestyle, multimorbidity, and medication use, and was employed to evaluate sex-specific factors. Feature selection employed a two-stage approach to ensure balanced covariate distributions between cases and controls: Random Forest-based stratified bootstrap resampling (1000 iterations) instead of traditional random sampling, followed by recursive feature elimination. Five models-Light Gradient Boosting Machine (LightGBM), Extreme Gradient Boosting (XGBoost), Random Forest, Feature Tokenizer Transformer (FT-Transformer), and Gated Adaptive Network for Deep Automated Learning of Features (GANDALF)-were assessed via 10-fold cross-validation. SHapley Additive exPlanations (SHAP) and restricted cubic splines (RCS) elucidated feature importance and nonlinear effects. A total of 39,430 female MAFLD patients were included in the final analysis, among whom 611 (1.55%) developed incident depression over the five-year follow-up. Adjusted logistic regression identified younger age at first live birth (<20 years) and early menopause (<35 years) as significant risk factors for depression. Among the evaluated models, GANDALF demonstrated superior performance (area under the receiver operating characteristic curve [ROC-AUC] = 0.96 ± 0.03, Matthews' correlation coefficient [MCC] = 0.830 ± 0.068), significantly outperforming conventional machine learning approaches (MCC range: 0.720 to 0.760) and exhibiting better calibration (Brier score: 0.066 vs. 0.093-0.115). SHAP analysis identified red blood cell count, Townsend deprivation index, and neutrophil count as the most influential predictors among the 17-feature panel. RCS analyses revealed nonlinear protective effects of moderate physical activity and higher red blood cell counts, contrasted by adverse effects from elevated white blood cell counts, overweight (body mass index [BMI] >25 kg/m2) and obesity (body mass index [BMI] >30 kg/m2), and early reproductive milestones. Additionally, our feature set showed enhanced predictive validity compared to established biomarker panels from prior studies (ROC-AUC 0.96 vs. 0.886-0.940). This study introduces a highly accurate, lightweight predictive model tailored for female MAFLD patients, leveraging 17 key features to improve the prediction of depression risk. By enabling personalized risk assessment and targeted interventions, our model offers a transformative approach to improve mental health outcomes and care quality in this vulnerable population.
Cumulative mental fatigue poses a significant threat to safety, productivity, and health in the workplace. In this study, we aimed to establish a robust machine learning framework using optimized resting-state electroencephalography (rs-EEG) features to detect such fatigue and to validate a 4-day high-stress cognitive competition paradigm for its induction. EEG signals were recorded from participants under eyes-closed (EC) and eyes-open (EO) conditions during fatigue and recovery phases. We extracted 544 features spanning power spectral density, entropy, and nonlinear complexity. Support Vector Machine Recursive Feature Elimination (SVM-RFE) was used for feature selection. The derived model index (Mean Model Result, MMR) was correlated with a subjective sleepiness index (the Stanford Sleepiness Scale, SSS) and sleep duration. Analysis of participant data identified a discriminative subset of 65 features from the EC EEG. The model achieved an accuracy of 90.37% in classifying deeply fatigued versus fully recovered states, significantly outperforming the EO-based model (86.54%). The MMR demonstrated a significant negative correlation with SSS scores (rs = -0.358, p = 0.020) and a positive correlation with sleep duration (rs = 0.494, p < 0.001). The results of this study demonstrate the superior efficacy of EC rs-EEG for monitoring cumulative fatigue, establishing a quantifiable EEG-sleep relationship and supporting the practical feasibility of this framework for occupational fatigue risk assessment.
Glycolytic markers such as carbonic anhydrase 9 (CA9) have been implicated in hepatocellular carcinoma (HCC) prognosis; however, sex-based differences and the underlying pathological mechanisms remain unclear. Herein, we investigated sex-stratified prognostic significance of CA9 and developed a pathomics-based predictive model. Data from the TCGA-LIHC cohort were analyzed, to determine the optimal CA9 expression cutoff using the "survminer" package. Kaplan_Meier survival analysis and Cox regression were performed to assess sex-specific prognostic associations. Tissue segmentation was conducted using the Otsu method, and features extracted with PyRadiomics were optimized through a combined minimum redundancy maximum relevance_recursive feature elimination_Akaike Information Criterion pipeline to construct a logistic classifier. Ferroptosis-related pathways were interrogated using FerrDb V2 genes with gene set variation analysis (GSVA) and weighted gene co-expression network analysis (WGCNA). Model performance was evaluated using ROC-AUC and calibration curves. Patients with high CA9 expression experienced significantly shorter overall survival than those with low expression (42.37 vs. 84.4 months, P < 0.001). CA9 emerged as an independent risk factor (HR = 2.131, P < 0.001) with a more pronounced prognostic impact in males than in females (HR = 2.836 vs. 1.261, P = 0.038). The pathomics model achieved AUCs of 0.746 in the training cohort and 0.706 in the validation cohort; the predictive score (PS) strongly correlated with CA9 expression (P < 0.001). Patients in the high-PS group had reduced survival (56.17 months vs. 82.87 months; P < 0.001), with no significant sex-based difference observed (P = 0.17). Tumors with high PS exhibited TGF-β/mTOR signaling, whereas low-PS tumors demonstrated upregulated ferroptosis-related CGAS expression and enrichment of NADPH oxidase- and EGFR resistance-related pathways (P < 0.01). CA9 represents a sex-dimorphic prognostic biomarker in HCC. When combined with a pathomics model, it offers a precise framework for risk stratification and provides insight into the contribution of the glycolysis-ferroptosis axis to HCC progression.
Metabolic dysfunction-associated steatotic liver disease (MASLD) represents a growing global health burden, yet early detection remains difficult, especially in low-resource settings. This study aimed to develop an interpretable machine learning model using both conventional and non-invasive, low-cost predictors to assess the probability of MASLD in the general population. Data from 3120 adults were split into training (n = 2,184) and testing (n = 936) sets. Six ML algorithms, including Random Forest, Gradient Boosting Machine, kernel-based Support Vector Machine, k-Nearest Neighbors, Neural Network, and Recursive Partitioning, were evaluated using standard performance metrics. Model interpretability was examined through SHapley Additive exPlanations. Analyses were also stratified by sex to explore heterogeneity in predictor influence. In the whole-variable model, Random Forest and Gradient Boosting Machine achieved the strongest performance (AUC 0.858 and 0.855, respectively). SHAP consistently identified composite hepatic indices, particularly the Fatty Liver Index and Hepatic Steatosis Index, along with waist circumference and adiposity-related measures, as leading contributors to MASLD probability. Sex-specific patterns were evident, with metabolic traits such as fasting glucose and diastolic blood pressure contributing more strongly in women, while hepatic markers and lipid-related variables were more influential in men. In a reduced model restricted to non-invasive, low-cost predictors, waist circumference remained the dominant determinant across all subgroups, followed by diastolic blood pressure, age, and a healthy lifestyle score. These findings highlight the association of visceral adiposity and metabolic dysfunction with MASLD and demonstrate that models based on widely accessible indicators can distinguish individuals with and without MASLD within this cohort, suggesting the potential feasibility of scalable screening approaches in resource-limited settings.
Primary angle-closure glaucoma (PACG) is increasingly recognized as involving brain alterations beyond the visual pathway, but the dynamic organization of large-scale brain activity and its biological context remain unclear. Co-activation pattern (CAP) analysis can characterize transient brain states and may provide insight into state-specific functional reorganization in PACG. Resting-state fMRI data were collected from 44 PACG patients and 57 healthy controls. CAP analysis was performed across multiple frequency bands, and six CAP states were identified. Group differences in CAP temporal dynamics and transition profiles were examined. Spatial associations between CAP-related alterations and normative transcriptomic, cell-type, and neurotransmitter receptor maps were assessed using Allen Human Brain Atlas data, enrichment analyses, cell-type-specific profiling, and receptor/transporter density maps. CAP-derived features were further evaluated using support vector machine-recursive feature elimination and multiple machine-learning classifiers. PACG patients showed state-specific alterations in CAP dynamics, with increased occurrence, dwell time, fractional occupancy, and self-transition probability of selected CAP states, alongside reduced engagement of complementary states. These altered states were organized into limbic-centered, spatially antithetical configurations involving attention, sensorimotor, and control networks. Imaging-transcriptomic analysis identified a dominant normative transcriptional gradient spatially associated with CAP alterations, involving genes enriched for neuronal excitability and synaptic regulation. Cell-type analyses indicated preferential enrichment in excitatory neurons, inhibitory neurons, and endothelial cells. CAP-related alterations also showed spatial associations with cholinergic, serotonergic, glutamatergic, and synaptic vesicle-related receptor systems. Machine-learning analyses demonstrated modest discriminative performance of CAP-derived features, with relatively high specificity but limited sensitivity. PACG is associated with state-specific alterations in intrinsic brain dynamics that spatially align with normative molecular, cellular, and neuromodulatory architectures. These findings provide a multiscale, hypothesis-generating framework for understanding brain functional alterations in PACG.
In spacecraft, precision equipment is severely affected by vibration excitation generated by rotor imbalance in the Control Moment Gyroscope (CMG), which is the core actuator for low-frequency attitude adjustments. Active vibration isolation (AVI) has been extensively studied by many scholars. However, traditional actuators suffer from contact friction and response delay; feedback and feedforward control alone have performance bottlenecks. As a result, the low-frequency tracking and mid-frequency isolation performance of CMG AVI are severely limited. Therefore, this paper proposes a magnetic levitation vibration isolation system (MLVIS) with composite control to optimize the tracking and AVI performance. In terms of structure, a magnetic levitation actuator is used to eliminate slow response and friction hysteresis. For the control algorithm, a composite strategy integrating integral force feedback and filtered-x recursive least squares is adopted. Through mutual compensation between the two control methods, this composite strategy mitigates the inherent time delay of feedback control and the instability of feedforward control. Experimental results show that the tracking error in the low-frequency band is 2.71% and 11.67% under single- and dual-frequency excitation, respectively, while 91.86% attenuation of the natural-frequency vibration amplitude is achieved in the mid-frequency band. These results verify the tracking and AVI performance of the MLVIS.
For 6G wireless networks, efficient resource allocation is a significant problem, especially with the growing need for ultra-low latency, high-speed communication, and efficient energy consumption. The traditional approach is found to be inadequate to meet the dynamic changes and service-allocation requirements. The application of AI and DL is seen as an efficient approach to making intelligent, timely decisions in complex scenarios. This paper proposes an integrated AI- and DL-based approach for efficient, intelligent resource allocation in 6G wireless communication. The difficulties encountered in dynamic spectrum allocation, energy depletion, and attenuation are addressed through an integrated approach that combines optimal path selection with efficient allocation mechanisms. The input parameters considered are residual battery indicator (RBI), channel matrix (H), normalized spectrum availability (v), SINR values, node pairs (s, d), service levels, and historical statistics. To ensure data quality, a Recursive Hampel Filter-Based Estimation Model (ReHF-EM) has been employed. Furthermore, for fundamental decision-making, a Dual-Stage Multi-Time-Scale Temporal Attention-Based LSTM network (D-MTSTA-LSTM) has been architected, which effectively learns short- and long-term relationships in network trends, thereby precisely predicting optimal communication routes and associated power and spectrum allocation. Additionally, the parameters of the proposed model have been fine-tuned using the Pied Kingfisher Optimizer (PKfO) for better efficiency, thus reducing complexities associated with the model. The proposed model has been implemented using Python, and various performance parameters such as Spectrum Efficiency (SE), Energy Efficiency (EE), SINR margin, Bit Error Rate (BER), Computational Time (CT), and Accuracy have been considered to evaluate the proposed model. The results show a 23.6% increase in Energy Efficiency and a 19.2% reduction in Bit Error Rate.