Browse Articles
Discover research articles across all indexed journals
The association between atherogenic index of plasma (AIP) and ultrasound attenuation parameter (UAP) in Chinese adults: a cross-sectional study
A CEL simulation approach for penetration-leveling of the multi-bucket foundation incorporating the temperature analogy method
Dynamic performance enhancement of adjustable blade pitch angle for wind generation system applications based on artificial neural network control techniques
Abstract The increasing reliance on the renewable energy, particularly wind power, introduces significant challenges for modern power systems and can compromise system stability. This study proposes an improved pitch-angle control strategy for a 1.5 MW large-scale Wind Energy Conversion System (WECS) based on a Doubly-Fed Induction Generator (DFIG). To address the limitations of conventional controllers, which struggle with system nonlinearity and the requirement for highly accurate mathematical models, this study examined Proportional-Integral-Derivative (PID) and Fractional PID (FPID) strategies. These were integrated with Neural Network (NN) architectures, specifically Multilayer Feedforward (MLFFNN), Cascade Forward (CFNN), and Elman NN, to improve control performance. The results, using MATLAB/Simulink, show that the MLFFNN architecture provides superior performance. With a minimum Mean Square Error of 0.0027024 and a power performance efficiency reaching a 98.9% under step, ramp, and random wind speed variations, the proposed NN controller consistently outperforms both PID and FPID systems, offering a robust solution for large-scale wind energy applications.
Machine learning-assisted validation of a high-isolation THz MIMO antenna for 6G communication and IoT application
TADM-CGAN: a resting-state to task activation map prediction framework using temporal attention-driven diffusion models and conditional generative adversarial networks
Abstract Predicting task-induced brain activation from resting-state fMRI (rs-fMRI) remains a significant challenge in computational neuroimaging, primarily due to the difficulty in simultaneously modeling the detailed temporal evolution and high spatial resolution of intrinsic neural activity. Most existing literature relies on parcel-based modeling using spatial functional connectivity features, neglecting the long-range temporal interactions and nonlinear fluctuations in rs-fMRI signals. To overcome these limitations, we introduce TADM–CGAN, a two-stage, grayordinate-level cascaded architecture that infers task activation maps directly from rs-fMRI time series, fully utilizing both temporal and spatial characteristics. In the first stage, a multi-head temporal attention–driven diffusion model (TADM) is employed to generate compact temporal embeddings for each grayordinate, capturing dependencies across the entire rs-fMRI time series. These unique temporal features then serve as inputs to 59,412 time series regression models, enabling highly localized, grayordinate-specific prediction of preliminary activation values. In the second stage, a principal component analysis (PCA)-conditioned conditional generative adversarial network (PCA-CGAN) is introduced, where PCA constrains adversarial refinement to a low-rank, biologically meaningful subspace, while the generator with the proposed activation fidelity loss (AFL) reduces noise and sharpens spatial details for the improved predictions. This proposed cascaded framework consistently outperforms prior task activation map prediction methods across a diverse set of task contrasts and datasets, underscoring its robustness and generalizability in practical applications.
Enhancing healthcare information security through VAE-driven anomaly detection in EHR access patterns
Abstract HIPAA breaches and unauthorized access to Electronic Health Records (EHRs) have been growing more likely due to the sudden digitalization of the healthcare sector. High endurance, privacy-based security practices have never been more in demand as hospitals and other medical facilities of this type have clung to the electronic system. The given study considers this problem by suggesting an anomaly detection model that determines the presence of abnormal or suspicious access patterns in EHR systems using Variational Autoencoders (VAEs). One of the main weaknesses of creating an efficient security model in the healthcare sector is that access to real-world information is limited and is constrained by privacy policies. To bridge this challenge, a synthetically enriched EHR access log dataset was generated and realistic features, including departmental affiliations, user roles, frequency of access, and timestamps, were ensured; the artificially generated dataset is, therefore, a close simulation of real-world hospital activities. By observing the access patterns of healthcare professionals that are generally typical in a latent space, the proposed VAE model can signal deviations that can indicate a possible security breach or policy violation without revealing or even relying on actual patient data, thus identifying both large and small-scale aberrations by modelling these latent representations. Evidence of the superiority of VAE-based detection over traditional machine learning algorithms, including Isolation Forest and One-Class Support Vector machines, using some key measures, like accuracy (F1-score: 0.93), lower false-positive rates, and greater sensitivity to noisy data, confirms this assertion. As a result, unsupervised deep generative modelling plus synthetic data generation provides a new, privacy-conserving approach to improving the cybersecurity of medical information systems. Based on the results, the VAE-based anomaly detection can become a trusted means of protecting sensitive healthcare infrastructure against the changing cyber risks.
Characterization of cast Ti30Cr20Mo15Zr10Ta5Nb20-xFex compositionally complex alloys
Abstract This study systematically investigates the relationship between the microstructure and performance of two cast Ti 30 Cr 20 Mo 15 Zr 10 Ta 5 Nb 20-x Fe x compositionally complex alloys (CCAs), prepared by vacuum arc melting. In the first alloy with (x = 0.0), 20 at.% Nb was added, resulting in the composition of Ti 30 Cr 20 Mo 15 Zr 10 Ta 5 Nb 20 CCA (20Nb), while in the second version (x = 10), a more cost-effective variant was developed by partially substituting Nb with 10 at.% Fe, yielding the composition of Ti 30 Cr 20 Mo 15 Zr 10 Ta 5 Nb 10 Fe 10 CCA (10Fe10Nb). Microstructural analysis showed that both alloys have a dendritic structure, with BCC1 as the main phase and a minor BCC2 phase. Some intermetallic phases, such as ZrCr 2 , MoNb, and MoTa, were also observed in the 20Nb alloy. In the Fe-containing CCA, more intermetallic compounds were formed with Zr, Cr, Ta, and Ti. The partial replacement of Nb with Fe in the 10Fe10Nb alloy reduced the intensity of the solid solution phases and promoted the formation of additional intermetallic compounds. The microstructure in both alloys was dendritic, with segregation of high-melting-point elements to the dendritic regions. In terms of mechanical properties, the 20Nb alloy exhibited a lower hardness (584 HV) than 10Fe10Nb (667 HV). The 10Fe10Nb alloy demonstrated a higher Young’s modulus of (102.47 GPa), while the 20Nb alloy measured (85.92 GPa). Regarding corrosion resistance in saline solution, the 20Nb alloy provided better corrosion protection than 10F10Nb without hydroxyapatite (HA) addition. However, both alloys showed excellent corrosion resistance in the presence of 3 g of HA inhibitor. The corrosion rate of 20Nb decreased from 39.09 μm/y without HA to 1.94 μm/y with 3 g HA, and that of 10Fe10Nb reduced from 61.84 μm/y without HA to 0.38 μm/y with 3 g HA. This emphasizes the effective interaction between Fe–Nb oxides and the deposited HA. Moreover, the incorporation of Nb promoted the formation and stabilization of a passive layer composed of Nb 2 O 5 and NbO 2 , while the addition of HA further enhanced the film’s thickness and compactness. Concluding, the properties of Ti 30 Cr 20 Mo 15 Zr 10 Ta 5 Nb 20-x Fe x CCAs can be tailored for a specific application through balancing Nb and Fe contents, and lower-cost versions can be produced.
Combating multidrug-resistant bacteria and associated virulence factors using Cichorium intybus extract: integrated microbiological characterization, phytochemical profiling, cytotoxicity assessment, and mechanistic insights
Abstract The global emergence of multidrug-resistant (MDR) Gram-negative pathogens necessitates novel therapeutic agents targeting bacterial growth and virulence. This study investigated the antibacterial, antibiofilm, antioxidant, cytotoxic, and apoptosis-inducing activities of Cichorium intybus leaf extract against MDR clinical isolates, alongside phytochemical profiling. Seventy-five clinical specimens were analyzed, predominantly blood (24%), sputum, pus, and urine (20% each), with stool samples underrepresented (4%; χ 2 (4) = 12.12, p = 0.033). A total of 75 bacterial isolates were identified using morphology, biochemical testing, and VITEK ® 2 (96–98% accuracy), including Klebsiella pneumoniae (56%), Escherichia coli (24%), and Acinetobacter baumannii (20%) (χ 2 (2) = 17.52, p < 0.001). The isolates exhibited high resistance to β-lactams, moderate resistance to aminoglycosides and tetracyclines, and greater susceptibility to imipenem and amikacin (χ 2 (3) = 92.4, p < 0.001). The C. intybus extract exhibited notable antibacterial activity, with the largest inhibition zone observed against K. pneumoniae (19.95 ± 2.16 mm). The minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) values ranged from 31.25 to 187.5 µg/mL and 62.5 to 375 µg/mL, respectively. Biofilm formation by A. baumannii , E. coli , and K. pneumoniae was significantly inhibited in a dose-dependent manner (0.78–50 µg/mL), with maximal suppression at 50 µg/m. Antioxidant assays demonstrated strong activity, with 92.2 ± 1.5% DPPH and 90.5 ± 2.0% ABTS inhibition at 1000 µg/mL, and IC 50 values of ~ 110 and ~ 115 µg/mL, respectively, showing strong correlation with ascorbic acid ( r = 0.997 and 0.995; p < 0.05). Cytotoxicity assays revealed selective, dose-dependent activity against PC3 and HepG2 cells compared with normal HFB4 cells (IC 50 = 24.6, 21.9, and 59.7 µg/mL, respectively), with apoptosis as the primary mechanism and minimal necrosis at lower doses. GC–MS and HPLC analyses identified bioactive compounds including chlorogenic, cichoric, linoleic, hexadecanoic, and octadecanoic acids, supporting synergistic antimicrobial, antioxidant, and anticancer effects. Overall, C. intybus demonstrates promising multifunctional bioactivity against MDR pathogens, supporting its potential as a natural therapeutic candidate.
The role of sand in an impact-based bedload monitoring system
Abstract Monitoring bedload transport is essential for understanding the morphology of rivers, yet traditional direct measurement techniques offer limited spatial and temporal resolution. Impact plates are a valuable addition for monitoring the transport of gravel-sized particles, although they require extensive field calibration. Recent research has therefore focused on developing globally applicable calibration approaches derived from controlled laboratory experiments. These efforts mainly aim to reproduce field-like flow velocities and roughness conditions. One aspect that remains largely unexamined is the influence of coincident sand transport. To address this gap, we conducted laboratory flume experiments that included additional sand transport, thereby replicating natural conditions during flood events more closely. Three sand feeding rates and seven gravel-sized diameter classes were investigated under high flow velocities, as they occur in mountainous rivers and creeks. The results were then compared with clear-water experiments. Signal and video analyses revealed that frequency-based signal properties, particularly centroid frequency, are strongly affected by the transported sand, whereas amplitude-based properties are more robust. Moreover, sand introduces biases in transport rate predictions derived from impulse counts. This study highlights the need to account for sand transport when calibrating impact-based monitoring systems, ultimately supporting more effective riverine sediment management and future revitalization measures.
Multimodality Imaging to Determine Underlying Causes of Myocardial Infarction With Nonobstructive Coronary Arteries in Women and Men
BACKGROUND: Myocardial infarction with nonobstructive coronary arteries (MINOCA) has several underlying causes, including mimicking conditions in some cases. Imaging is recommended to identify MINOCA etiologies, but it remains unclear which patients are most likely to have abnormal findings. We characterized MINOCA mechanisms, analyzed predictors of imaging abnormalities, and explored sex differences. METHODS: We enrolled patients with clinical diagnosis of myocardial infarction in an international, prospective, diagnostic study at 28 sites in the United States, Canada and United Kingdom. After a women-only phase, we included both sexes. Individuals with ≥50% diameter stenosis or coronary dissection on angiography, or alternate causes for the clinical presentation, were excluded. Participants had multivessel coronary optical coherence tomography (OCT) during index coronary angiography and cardiac magnetic resonance imaging (CMR) within 1 week. Independent core laboratories interpreted imaging, blinded to other results. RESULTS: Among 754 patients enrolled, 389 had MINOCA, and 336 with MINOCA underwent OCT (270 women and 66 men); CMR was completed in 284 (85%). An OCT-defined culprit lesion was identified in 45% (116 of 270 women [43%] and 35 of 66 men [53%], P =0.18). CMR demonstrated an ischemic pattern in 114 of 284 (40%), similar by sex (96 of 225 women [43%] versus 18 of 59 men [31%], P =0.12). A nonischemic pattern was observed in 23% (23% of women, 25% of men, P =0.78). We identified a cause of the clinical presentation in 79% of patients with both tests completed; 59% had an ischemic cause of MINOCA, and 20% had a non-ischemic mimicking condition. OCT alone found a MINOCA etiology in 151 of 336 (45%) and CMR alone in 180 of 284 (63%). Predictors of an OCT culprit lesion included age, abnormal angiogram, and number of vessels imaged, but 27% of normal angiograms harbored a culprit lesion. Predictors of abnormal CMR were peak troponin, shorter time to CMR, and non-Asian race, but CMR was abnormal in 40% when troponin was <4-fold above the upper reference limit. CONCLUSIONS: The combination of multivessel coronary OCT and CMR in patients with a clinical diagnosis of MINOCA confirmed myocardial infarction in 59% and identified an alternate cause (MINOCA mimic) in 20%. Clinical factors had limited usefulness to predict imaging abnormalities. No sex differences in imaging results were detected. REGISTRATION: URL: https://www.clinicaltrials.gov ; Unique identifier: NCT02905357.
OCT-based optic neuropathy diagnosis using explainable and privacy-preserving machine learning
Abstract Glaucoma shares similarities with neurodegenerative conditions like dementia, Parkinson’s disease, and ischaemic optic neuropathy, which affect ocular health. However, current studies exclude neurodegenerative cases in neuropathy diagnosis and rely on ‘black box’ models. To address this, we applied explainable machine learning to optical coherence tomography (OCT) data, integrating a privacy-preserving mechanism to create a reliable neuropathy diagnostic tool. OCT data from 268 glaucomatous, 334 normal, 56 dementia, 60 Parkinson’s, and 93 ION eyes were analysed from a Sydney-based eye clinic. Spatial and frequency domain features were extracted, followed by feature selection and hierarchical classification. Model interpretability was enhanced using SHapley Additive exPlanations and partial dependency analysis, and privacy was preserved incorporating a differential privacy mechanism. A team of three clinicians, with 12, 11, and 6 years of experience, evaluated their performance on the same dataset for a human versus machine comparison, with the machine achieving an area under the curve of 0.90 for classifying neuropathy. Overall, machine outperformed clinicians, with 26.3% higher accuracy for neuropathy and 24.8% higher accuracy for glaucoma diagnosis. In conclusion, both non-explainable and explainable methods show promise in enhancing diagnostic support for clinical decision-making, with the privacy-preserving approach safeguarding data privacy.
Recovery From Heart Failure: Microvascular Mechanisms
BACKGROUND: Heart failure (HF) is a significant global health problem. Left ventricular assist device (LVAD) implantation serves as a bridge for patients awaiting heart transplantation. Intriguingly, LVAD support often improves cardiac histology and function, sometimes enough to avoid transplantation after LVAD removal. However, the cellular programs underlying this recovery remain unclear. METHODS: Myocardial tissues were obtained from patients with HF at the time of LVAD implantation (pre LVAD) and explantation (post LVAD) for histological analysis and single-nucleus RNA sequencing. A murine model of HF recovery, combined with lineage tracing studies, was employed to define cellular sources of vascular repair. Cardiac function, fibrosis, and vascular density were assessed using echocardiography, histology, and fluorescent microsphere perfusion. A patient-derived cardiac nonmyocyte culture system was established to interrogate mechanisms of cell fate regulation. RESULTS: Post-LVAD myocardial tissues exhibited reduced fibrosis and increased capillary density compared with pre-LVAD samples. Across samples, fibroblast abundance was inversely correlated with endothelial cell abundance, consistent with enhanced angiogenesis during recovery. Single-nucleus RNA sequencing identified a fibroblast subset predisposed to undergo mesenchymal-to-endothelial transition, acquiring an endothelial cell identity. Additionally, nonmyocytes from pre-LVAD hearts proliferated poorly and failed to form vascular structures, whereas nonmyocytes from post-LVAD hearts displayed greater proliferation and angiogenesis capacity, forming vessel-like structures, reinforcing the association of HF recovery with angiogenic reprogramming. Mechanistically, knockdown of c-Myc (cellular myelocytomatosis oncogene) by small interfering RNA shifted post-LVAD nonmyocytes to a pre-LVAD–like state, while c-Myc overexpression by mRNA in pre-LVAD cells induced a post-LVAD–like phenotype, implicating c-Myc as 1 contributor to this fate switch. A model of HF recovery in mice mimicked the histological and functional changes in patients, with physiological evidence of increased microvascular perfusion, associated with a fibroblast-to-endothelial transition, documented by lineage tracing. CONCLUSIONS: HF recovery involves reduced fibrosis and enhanced microvascularization, partly driven by fibroblast-to-endothelial cell fate transition. c-Myc functions as 1 regulator of this transition, offering a mechanistic entry point to develop regenerative therapies in HF.
A vision-based framework for quantifying fish feeding behavior in industrial recirculating aquaculture systems
Abstract Accurate quantification of fish feeding intensity is critical for optimizing feeding strategies and reducing feed waste in industrial recirculating aquaculture systems (RAS). However, real-world aquaculture environments present significant challenges, including high-density fish populations, water surface disturbances, and dynamic behavioral variations. To address these issues, this study proposes a hybrid vision-based framework (HVIT) for robust feeding intensity analysis. The proposed method integrates a Convolutional Neural Network (CNN) for local feature extraction and a Vision Transformer (ViT) for global context modeling within a parallel architecture, enabling effective representation of complex group behaviors. Furthermore, a Long Short-Term Memory (LSTM) module is incorporated to capture temporal dynamics of feeding activity, allowing continuous characterization of feeding intensity over time. A dedicated dataset of largemouth bass (Micropterus salmoides) under industrial RAS conditions was constructed, with data augmentation strategies applied to improve robustness against environmental noise and visual disturbances. Experimental results demonstrate that the proposed framework achieves over 98% accuracy across four feeding intensity levels and outperforms conventional CNN-based approaches. More importantly, the proposed method enables quantitative evaluation of feeding activity, providing a practical basis for real-time feeding decision support and intelligent feeding system development in large-scale aquaculture.
Correction to: Breast Cancer Reveals Latent <i>BMPR2</i> -Related Susceptibility to Pulmonary Hypertension
Methylphenidate exposure alters brain gene expression and induces transgenerational DNA methylation changes in Poecilia reticulata guppies
Response by Croon and Khera to Letter Regarding Article, “Phenotypic Selectivity of Artificial Intelligence–Enhanced Electrocardiography in Cardiovascular Diagnosis and Risk Prediction”
Network analysis of serotonin CNVs shows biological convergence from genetic heterogeneity and discriminates between autism and developmental delay
Abstract Differentiating autism spectrum disorder (ASD) from developmental delay (DD) is critical for guiding early intervention, but overlapping features and shared biological mechanisms pose challenges. This study investigates whether copy number variations (CNVs) affecting serotonergic genes carry sufficient information to distinguish between these neurodevelopmental disorders (NDDs). Using network mapping and machine learning, we applied gene ontology (GO) terms related to serotonergic systems to filter CNVs and construct networks modeling genetic and biological variation in ASD and DD. We identified hub nodes and subnetworks reflecting distinct patterns in gene and GO term interactions. ASD networks analysis yielded six genetic clusters, five of which remarkably contained genes specifically linked to serotonergic receptor mechanisms. In contrast, DD networks exhibited greater genetic homogeneity, with just two clusters sharing serotonergic mechanisms. Random Forest classifiers using serotonergic gene features achieved an average prediction accuracy of 85.6%, increasing to 88.6% when combined with dopaminergic dosage features, consistent with the two systems capturing partially non-overlapping biological signal. GO-based features yielded comparable accuracy with fewer inputs, emphasizing their efficiency. These findings demonstrate that different genetic alterations may be associated with disruption of shared biological pathways, each leaving a distinct signature tied to clinical diagnoses. Importantly, although genetic heterogeneity is observed in ASD, we found a homogeneity of serotonergic biological terms, suggesting convergence of mechanisms, which was distinct for each condition. Together, these results suggest that serotonergic CNVs carry discriminatory information for ASD vs. DD classification, and that combining serotonergic and dopaminergic features captures partially non-overlapping biological signal.
Automated Echocardiographic Detection of Congenital Heart Disease Using Artificial Intelligence
BACKGROUND: Delayed or missed diagnosis of congenital heart disease (CHD) contributes to excess pediatric mortality worldwide. Echocardiography (echo) is central to diagnosing and triaging CHD, yet expert interpretation remains a scarce and maldistributed global resource. Artificial intelligence offers the potential to democratize diagnostics and to extend expert-level interpretation beyond large academic centers, but its application in CHD remains underexplored. METHODS: We developed EchoFocus-CHD, an artificial intelligence–enabled model for automated detection of 12 critical and 8 noncritical CHD lesions, individually and as composites. The composite critical CHD outcome was the primary end point. The model expands on a multitask, view-agnostic architecture (PanEcho) with a transformer encoder to improve focus on relevant echo views. The model was internally trained (80%) and tested (20%) on the first echo per patient from Boston Children’s Hospital, with further evaluation on a referral cohort of echo studies performed at external US and international centers. RESULTS: The internal and referral cohorts included 3.4 million videos from 54 727 echos (median age at echo, 7.1 years [interquartile range, 0.2–15.0 years]; 5.8% critical CHD, 23.6% noncritical CHD) and 167 484 videos from 3356 echos (median age at echo, 2.5 years [interquartile range, 0.3–9.4 years]; 29.4% critical CHD, 45.6% noncritical CHD), respectively. EchoFocus-CHD showed excellent internal ability to detect the composite critical CHD outcome (area under the receiver-operating curve [AUROC], 0.94; positive likelihood ratio, 7.50; negative likelihood ratio, 0.14) and individual critical lesions (AUROC, 0.83–1.00), as well as composite noncritical CHD (AUROC, 0.90; positive likelihood ratio, 5.00; negative likelihood ratio, 0.23) and individual noncritical lesions (AUROC, 0.70–0.96). Performance declined during evaluation on the referral cohort to detect critical CHD (AUROC, 0.77), coinciding with greater expert disagreement on referral cases (κ=0.72 versus 0.82 for internal cases). Explainability analyses demonstrated that the model prioritized the same clinically relevant views (parasternal long axis, parasternal short axis, subxiphoid long axis, apical) across internal and referral cohorts, whereas uniform manifold approximation and projection analysis revealed a domain shift between cohorts. Retraining on all available US patients attenuated domain shift effects, improving international critical CHD detection (AUROC, 0.87) and calibration. CONCLUSIONS: EchoFocus-CHD shows promise for automated CHD detection to advance equitable global cardiovascular care and highlights the need to address domain shift and to establish external validation before real-world deployment.