Therapeutic resistance to chemotherapy or radiotherapy is a significant issue in several cancers, including head and neck squamous cell carcinoma (HNSCC). One pathway associated with therapeutic resistance is the NFκB pathway, which promotes survival in response to the cytokine TNFα, a key mediator of chemotherapy and radiotherapy-induced cytotoxicity. However, direct targeting of the NFκB pathway is associated with significant toxicity and thus targeting the regulation of this pathway is a promising therapeutic target. We recently demonstrated that the USP14/UCHL5 inhibitor b-AP15 inhibits NFκB activity, inhibiting proliferation and inducing apoptosis in HNSCC cells. Furthermore, b-AP15 treatment sensitised HNSCC cells to the cytotoxic effects of TNFα, as well as TNF-inducing radiation treatment. Here, we investigated if b-AP15 sensitised HNSCC cells to tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), a cancer selective member of the TNF family. b-AP15 treatment sensitised HNSCC cells to TRAIL treatment. Mechanistically, we show that b-AP15 induced expression of the TRAIL receptor Death Receptor 5 (DR5)/TRAIL Receptor 2 (TRAILR2), which was required for b-AP15-mediated TRAIL sensitisation. b-AP15 induced reactive oxygen species (ROS) and activated the JNK signalling pathway and both ROS and JNK signalling were required for the induction of DR5 expression and TRAIL sensitisation. We further show that b-AP15-mediated reduction of the NFκB-dependent gene XIAP induced DR5 expression and TRAIL sensitisation and that combination between b-AP15 and IAP antagonists was synergistic in HNSCC cells in vitro. Our data further define the mechanism of b-AP15-mediated cytotoxicity and highlight potential combination treatments that warrant further exploration in pre-clinical studies in HNSCC.
The gut microbiome supports digestion, immunity, and metabolism; its imbalance (dysbiosis) drives inflammation and metabolic dysfunction, contributing to chronic diseases such as diabetes, cardiovascular disease, inflammatory bowel disease, and autoimmune disorders. Medicinal plants provide a wide range of phytochemicals (such as polyphenols, flavonoids, alkaloids, saponins), which reach the colon and undergo two-sided interactions with microbes in the gut, acting as potential microbiome modulators and substrates of biotransformation into bioactive metabolites. This structured narrative review synthesises evidence from peer-reviewed studies indexed in PubMed, Scopus, and Web of Science over the last 10 years on the role of medicinal plants in microbiome-mediated chronic disease modulation. This literature is organised into three mechanistic axes: (i) perturbations, defined here as measurable shifts in microbial diversity or taxonomic composition relative to a baseline or healthy reference state, together with beneficial taxa enrichment; (ii) alterations in microbial metabolite output, especially short-chain fatty acids (SCFAs) and other immunometabolic mediators; and (iii) downstream host metabolic and immune signalling. Rather than broad descriptive summaries, the literature is organised using an axis-based mechanistic framework, highlighting key translational constraints such as botanical heterogeneity, dose/formulation variability, and inconsistent microbiome endpoint standardisation, that must be addressed to strengthen human evidence and clinical relevance. Illustrative microbiome-mediated processes involve botanicals such as turmeric (curcumin), ginseng (ginsenosides), and green tea (catechins), though evidence strength varies by study design. Future progress requires standardised phytochemical characterisation, microbiome-stratified trials, and integration of multi-omics with artificial intelligence analytics to enhance mechanistic insight, identify responders, and enable personalised plant-based microbiome therapies.
Host dependency factors (HDF) are essential for viral replication and are promising targets for broad-spectrum antivirals. However, most work has focused on individual viruses or individual data types, limiting our understanding of shared host mechanisms across viruses. We developed a pan-viral framework that integrates multi-omics data-including genome-wide perturbation screens, single-cell transcriptomes and viral interactomes-and combines graph-based learning with classical machine-learning models to prioritize HDF for four RNA viruses (SARS-CoV-2, influenza A virus, dengue virus and Zika virus). Across viruses, the framework achieved high discrimination, with area under the receiver operating characteristic curve (ROC-AUC) greater than 0.90 on benchmark datasets, and identified a conserved signature of 118 genes shared by all four viruses and 427 genes shared by at least three. These genes converge on recurrent host programmes such as clathrin-mediated entry and endomembrane trafficking, nuclear transport, RNA processing and stress granules, and proteostasis and ubiquitin-proteasome signalling. The pan-viral signature generalizes beyond the training set, as genes shared by three or more viruses are strongly enriched among top-ranked Ebola virus candidates. We further provide a prioritized shortlist and an experimental validation roadmap to guide follow-up perturbation studies. Our integrative multi-omics and machine-learning approach outlines a prediction-based, data-driven map of pan-viral host liabilities and highlights tractable opportunities for host-directed therapy against diverse RNA viruses.
African swine fever virus (ASFV) is a large, complex DNA virus that causes a haemorrhagic disease with lethality rates approaching 100%. Mechanisms of virulence are still poorly understood, but loss of members of the multigene families (MGF) is commonly observed in naturally occurring attenuated virus strains, indicating that these proteins play an important role in disease pathogenesis. Here, a suppressor of cytokine signalling (SOCS)-box like motif was identified in proteins of the MGF505 cluster. SOCS-box motifs typically recruit the Cullin-RING-ligase (CRL) machinery, a superfamily of E3 ubiquitin ligases that target proteins for proteasomal degradation. Data presented show that MGF505-1R inhibits the host innate immune responses controlled by the activation of the transcription factors IRF3 and NFκB. MGF505‑1R expression was shown to correlate with a reduction in the protein levels of the transcriptional co‑activator p300. Targeted mutations of the SOCS box motif were shown to reverse the observed IRF3 and NFκB inhibition. The data support a role for MGF505-1R in hijacking the host ubiquitin machinery to evade the host immune responses.
To evaluate finerenone-associated adverse events (AEs) and to investigate the association between finerenone use and renal injury via data mining of the Food and Drug Administration Adverse Event Reporting System (FAERS). To minimize statistical bias, the data extraction period was set from database inception (2004) to provide a stable background for disproportionality analysis. Four disproportionality algorithms (ROR, PRR, BCPNN, and MGPS) and stricter case-screening methods were employed to improve analytical precision. Additionally, a clinical priority evaluation was conducted to rank clinical risks and surveillance levels for these AEs. Supplementary analysis was performed to assess the relationship between finerenone and renal injury, as well as associated risk factors. A total of 1316 finerenone-related reports were identified. 30 AEs were detected as significantly positive signals, with most being related to renal function (15 PTs, 50%), blood pressure (5 PTs, 16.67%), and blood potassium (4 PTs, 13.33%). Among them, blood glucose increased, blood creatine increased, and flank pain were new potential AEs. Acute kidney injury, hyperkalemia, renal impairment, glomerular filtration rate decreased, blood creatinineincreased, blood potassium increased, and hyponatremia exhibited moderate clinical priority levels and warrant further study. Signals reflecting renal injury were detected in patients regardless of baseline nephropathy. Male sex, taking more than 3 drugs, and using amlodipine may be risk factors for finerenone-related nephrotoxicity. These results highlight new finerenone-related AEs, provide ranked guidance for pharmacovigilance through clinical priority evaluation, and clarify factors that influence renal injury, providing guidance for individualized treatment and improved drug safety.
Age-related macular degeneration (AMD) is characterized by progressive retinal pigment epithelium (RPE) dysfunction driven by oxidative stress and chronic inflammation, in which NLRP3 inflammasome activation plays a critical role. Mesenchymal stem cells (MSCs) exhibit therapeutic potential, but their efficacy is limited by poor survival and reduced paracrine activity in hostile microenvironments. Here, we investigated whether three-dimensional (3D) spheroid culture enhances the protective effects of umbilical cord-derived MSCs (UC-MSCs) on RPE cells by promoting autophagy and suppressing inflammasome activation. Human UC-MSCs were cultured as 3D spheroids or conventional 2D monolayers and applied in sodium iodate (NaIO3)-induced oxidative injury models both in vitro and in vivo. Retinal morphology and function were assessed via histology and electroretinography, while NLRP3/caspase-1 activation, LC3-II/I ratios, and autophagy flux were quantified using immunofluorescence and Western blot. GO/KEGG enrichment was performed to identify pathways associated with 3D MSCs efficacy. Mechanistic involvement of autophagy was validated using 3-methyladenine (3-MA) and rapamycin. 3D MSCs formed compact spheroids exhibiting enhanced paracrine potential and significantly outperformed 2D MSCs in protecting RPE cells against NaIO3-induced injury. In vivo, 3D MSC treatment preserved retinal structure, reduced RPE cell loss, and improved retinal function. In vitro, co-culture with 3D MSCs markedly improved ARPE-19 viability, reduced apoptosis, and modulated autophagy-related marker expression, as evidenced by increased LC3-II/I ratios. 3D MSCs significantly inhibited NLRP3 inflammasome activation and pro-inflammatory cytokine release, effects reversed by 3-MA and further enhanced by rapamycin. 3D spheroid culture substantially augments the therapeutic efficacy of UC-MSCs by boosting autophagy and suppressing NLRP3 inflammasome signaling, resulting in enhanced protection of RPE cells from oxidative and inflammatory injury. These findings provide preclinical evidence supporting 3D MSCs as a promising therapeutic strategy for AMD.
Skin photoaging is predominantly induced by ultraviolet (UV) irradiation. Intense pulsed light (IPL) is a commonly employed non-ablative treatment for photoaging. However, the effects and mechanisms of IPL on UV-induced skin photoaging remain insufficiently understood. In this study, we aimed to examine the anti-photoaging effects of IPL and elucidate the underlying mechanisms. This study revealed that UV triggered extracellular signal-regulated kinases (ERK) together with c-jun NH2-terminal kinase (JNK), while selectively suppressed UV-induced ERK phosphorylation while activating JNK in human skin keratinocytes. The different ERK/JNK expression patterns induced by UV and IPL resulted in distinct c-fos/c-jun (activator protein 1) phosphorylation, cyclin D1 expression, and matrix metalloproteinase (MMP) secretion. In vivo, IPL inhibited MMP expression in guinea pig skin and promoted c-fos/c-jun phosphorylation, epidermal proliferation, and collagen remodeling. These findings indicated that ERK was involved in IPL rejuvenation by regulating c-fos, c-jun, cyclin D1, and MMPs, providing a potential target for skin rejuvenation.
Human papillomavirus (HPV) is a well-established oncogenic virus implicated in the development of several epithelial cancers, most notably cervical, anogenital, and oropharyngeal carcinomas. In contrast, neuroendocrine neoplasms (NENs)-a heterogeneous group of malignancies arising from neuroendocrine cells across various organ systems-have not traditionally been linked to HPV infection. In this study, we performed extensive genomic and transcriptomic profiling to compare HPV-positive NENs to HPV-positive non-NENs across anatomical sites, aiming to uncover biologically and clinically actionable differences. HPV16- and HPV18-positive tumors were identified from 101,343 solid tumors profiled at Caris Life Sciences (Phoenix, AZ) with DNA and RNA sequencing. Prevalence of pathogenic mutations and copy number amplifications were calculated. Fisher's exact/χ2 tests were applied appropriately with p-values adjusted for multiple comparisons (p < 0.05). HPV positivity was most frequent in cervical carcinomas (70%, 1200/1716). Importantly, 6% (96/1620) of NENs from all tissues were positive for HPV16 or HPV18. Among HPV-positive NENs, 93% were high-grade compared to 54% observed in HPV-negative NENs (p < 0.001), highlighting a strong association between HPV and tumor aggressiveness in this subset. Analysis of HPV-associated sites (cervix, anorectal region, and head and neck) revealed that HPV-positive NENs possess distinct genomic and transcriptomic landscapes compared to HPV-positive non-NENs. Notably, interferon signaling was significantly suppressed in HPV-positive NENs, suggesting a unique tumor-immune microenvironment. Our findings indicate that HPV-positive NENs form a distinct subset with unique genomic features, including reduced interferon signaling, compared to HPV-positive non-NENs. Thus, future studies focused on evaluating HPV status, along with genomic and transcriptomic characteristics, may uncover biologically and clinically actionable distinctions for this rare yet clinically significant tumor subgroup. Not applicable.
Optically pumped magnetometers magnetoencephalography (OPM-MEG) have demonstrated their value in the diagnosis and mapping of epilepsy, as well as their advantages in pediatric applications. We present a case of 8-year-old boy with drug-resistant epilepsy, whose epileptogenic lesion is in left Broca's region. The boy underwent language function evaluation and localization by on-scalp OPM-MEG before surgery. Dipole clusters and Dipole Density of epileptogenic signals by OPM-MEG were located in the left inferior frontal gyrus, though language verbal generation mapping of OPM-MEG signals were mainly located in left frontal orbital gyrus, indicating a localization of the language function area. Seizure freedom and no loss of language function were achieved after MRI-guided laser interstitial thermal therapy. This article underscores the feasibility of using OPM-MEG to record abnormal discharges of seizure and assess language function area in children, especially in drug-resistant epilepsy surgery involving brain functional areas.
Itch is a complex noxious sensation associated with many skin and systemic conditions, which varies in intensity and quality across different body regions. Despite its prevalence, the molecular and cellular mechanisms underlying regional itch differences remain poorly understood. Investigating the neural basis of regional itch differences, we identified a functional divergence in neuropeptide signaling and central circuit engagement between the trigeminal and spinal systems, which was independent of peripheral innervation density. Utilizing a combination of behavioral, pharmacological, genetic, and molecular assays, we identified a unique population of trigeminal (TG) neurons that facilitate specialized itch-pain coding. Our results indicate that while histamine receptors HRH1 and HRH3 are both involved in mediating mixed itch-and-pain sensations, the specific activity of Substance P (SP)- and Somatostatin (SST)-expressing neurons orchestrates this transition in the cheek. This behavioral shift is mediated by a central mechanism wherein sensory neurons activation recruits distinct nociceptive circuits within the brainstem. In brief, these findings provide insights into the molecular and cellular mechanisms underlying regional itch differences, highlighting the importance of considering anatomical location when developing targeted treatments.
Cold preservation is a critical logistical step in liver transplantation but induces ischemia-reperfusion injury (IRI), a key driver of early graft dysfunction. While bulk tissue assays capture global damage, they obscure the cell-type-specific transcriptional programs engaged during hypothermic storage. We utilized a multicellular human liver-on-chip model comprising Patient-Derived Organoids (PDOs), hepatic stellate cells (HSCs), liver sinusoidal endothelial cells (LSECs), and macrophages. Chips were exposed to 24-h static cold storage using either the clinical standard University of Wisconsin (UW) solution or a hyperbranched polyglycerol (HPG)-based formulation, followed by normothermic reperfusion. Single-cell RNA sequencing (scRNA-seq) was performed to map transcriptional trajectories across the preservation-reperfusion axis. We identified candidate solution-dependent transcriptional differences across cell types. PDOs from UW-preserved chips showed comparatively higher mean expression of inflammatory and oxidative stress-associated transcripts (IFI27, SAA1, HMOX1) and mitochondrially-encoded genes (MT-ND5) relative to HPG-preserved samples, which retained comparatively higher expression of homeostatic epithelial markers (EPCAM, KRT18). HSCs and LSECs in the UW group showed comparatively elevated expression of fibrosis-associated (COL1A1, TAGLN) and endothelial adhesion (ICAM1) transcripts. Ligand-receptor interaction modelling identified candidate inflammatory communication axes, including chemokine signaling interactions (CXCL1, CCL20) between macrophages and epithelial compartments, with higher predicted activity under UW preservation. This study provides an exploratory, high-resolution map of cell-type-specific transcriptional patterns associated with hypothermic preservation in a liver-on-chip model. Our findings suggest that preservation solution chemistry is associated with distinct transcriptional signatures spanning stress response, mitochondrial, and intercellular signaling pathways. Transcriptional patterns in HPG-preserved cells were consistent with comparatively attenuated injury responses; however, these observations are hypothesis-generating and require independent biological replication and functional validation, including metabolic flux assays and ROS production measurements before conclusions regarding mitochondrial protection or clinical preservation efficacy can be drawn.
The increasing application of time-series analysis in fields like biomedical engineering or telecommunications emphasizes the need for high-quality data to train and evaluate advanced machine learning models. Acquiring temporal data at suitable resolutions is often limited by ethical, economic, or practical constraints. We introduce CoSiBD (Complex Signal Benchmark Dataset for Super-Resolution), a synthetic dataset designed for reproducible time-series super-resolution research. CoSiBD provides 2,500 high-resolution signals (N = 5, 000 samples each over a reference domain τ ∈ [0, 4π]) with aligned low-resolution versions at four levels (150, 250, 500, and 1,000 samples) obtained via uniform decimation. Signals are generated with diverse non-stationary behaviors through piecewise frequency modulation and spline-based amplitude envelopes, and provides both clean and noisy variants. Signals are distributed as NumPy arrays, plain text, and JSON, with comprehensive metadata describing segment structure, generation parameters, and seeds for full reproducibility. Technical validation analyzes spectral properties and reports baseline SR benchmarking and transfer experiments on EEG and speech data.
High-altitude pulmonary hypertension (HAPH), classified as Group 3 pulmonary hypertension, is a significant threat to the health of high-altitude populations. The scarcity of studies in diverse populations has become a research bottleneck, limiting diagnostic and therapeutic advances. In this first proteomic study focusing on the eastern Pamir Plateau (Kizilsu Kyrgyz Autonomous Prefecture, Xinjiang), plasma samples were analyzed using data-independent acquisition (DIA) mass spectrometry. Differential expression analysis in parallel with weighted gene co-expression network analysis was performed to identify core pathways and hub proteins, and gene set enrichment analysis was used for quality assessment. Integrative analysis of the two methods was used to select candidates for validation by enzyme-linked immunosorbent assay (ELISA) in an independent cohort. Among > 1400 detected proteins, 123 were differentially expressed and 45 were identified as hub proteins significantly associated with HAPH. Extracellular matrix (ECM) remodeling- and angiogenesis-related proteins were upregulated, whereas proteins related to enzyme activity, iron metabolism, and inflammatory responses were downregulated. Integrative analysis identified 23 core proteins, with ECM-receptor interaction and TGF-β/Smad signaling identified as key pathways. ELISA confirmed that plasma levels of THBS2, LOXL1, and POSTN were significantly elevated in patients with HAPH (P < 0.05). Among these, THBS2 and LOXL1 levels were positively correlated with mPAP (THBS2: r = 0.389, 95% CI: 0.034-0.657, P = 0.033; LOXL1: r = 0.457, 95% CI: 0.115-0.701, P = 0.011). ECM remodeling is closely associated with HAPH in this indigenous high-altitude population. THBS2, LOXL1, and POSTN show potential as biomarkers and therapeutic targets.
Cross-study inconsistencies in autism spectrum disorder (ASD) blood microRNA biomarker studies suggest that methodological heterogeneity may substantially limit reproducibility. We conducted an exploratory meta-analysis of publicly available ASD blood miRNA datasets from the Gene Expression Omnibus, applying rigorous inclusion criteria and standardized analytical protocols. Three datasets were included (GSE89596, GSE67979, GSE222046) comprising 614 miRNAs across 90 participants (45 ASD, 45 controls). Random-effects meta-analysis was performed using Hedges' g effect sizes, with comprehensive heterogeneity assessment and leave-one-dataset-out cross-validation. No miRNAs survived multiple testing correction (Benjamini-Hochberg FDR < 0.05), though seven candidate signals showed consistent evidence with unadjusted p < 0.01 and large effect sizes. These candidates demonstrated near-zero between-study heterogeneity and consistent directionality across validation analyses. Potential age-related and platform-related differences were observed, with near-zero correlation between adult and pediatric effect sizes (Kendall's τ = -0.022); however, these two sources of variability were fully confounded in the available data and could not be separated. Some miRNAs exhibited extreme between-study variability (I² > 80%), indicating substantial methodological differences. Cross-validation revealed that excluding the single adult dataset reduced sign consistency from 89.9% to 68.9%. Our findings suggest that age-related and methodological factors, including technical platform differences, may contribute to limited reproducibility in ASD blood miRNA research, and that blood-derived signals should be interpreted as potentially reflecting peripheral physiological states rather than central disease mechanisms. A supplementary cross-tissue analysis using post-mortem prefrontal cortex data (GSE59286; n = 45) provided direct empirical support for this interpretation: the majority of blood candidate miRNAs showed no corresponding expression in brain tissue, with only hsa-miR-29c-5p demonstrating directional concordance across both tissues. These findings suggest that age stratification, platform harmonization, and cross-tissue validation should be considered essential prerequisites for reliable ASD miRNA biomarker discovery, rather than optional refinements.
Bone remodelling is essential for maintaining skeletal integrity by preserving the balance between bone formation and resorption, with excessive osteoclast activity contributing to osteoporosis. Osteocytes act as central regulators of osteoclastogenesis through mechanically sensitive paracrine signals, yet the influence of osteoblasts and their mesenchymal precursors remains less defined. Extracellular vesicles (EVs) have recently emerged as mediators of bone cell communication, although their role in osteoclast regulation are still underexplored. This study demonstrates that mesenchymal-derived bone cells inhibit osteoclastogenesis through an EV-dependent mechanism shaped by their differentiation stage and mechanical environment. Mechanically stimulated osteocyte-derived EVs showed the strongest anti-catabolic response. Notably, we identify miR-150-5p as a mechano-responsive miRNA enriched within osteocyte EVs, capable of inducing a dose-dependent reduction in osteoclastogenesis. Transcriptomic analyses reveal that EV treatment and miR-150-5p delivery induce substantial transcriptional changes in osteoclast precursors, including downregulation of shared target genes linked to bone remodelling. Overall, we highlight mechanically activated osteocytes as key regulators of osteoclastogenesis through an EV-mediated mechanism, in which miR-150-5p represents a promising candidate contributor within the broader EV cargo landscape, highlighting their potential for future cell-free therapeutic strategies.
Non-invasive health monitoring has recently gained a lot of consideration in the modern healthcare system, because it has the potential to diagnose diseases earlier and can monitor patients in a remote manner. This research presents a hybrid approach in healthcare monitoring by integrating vocal and lung abnormality detection using a multinetwork model. The model utilizes multiple data sources and Mel Frequency Cepstral Coefficients (MFCCs) to capture the frequency spectrum of the signal. A multinetwork model developed for disease identification is made up of hybrid deep learning networks, which consist of Convolutional Neural Networks (CNN) and Bi-directional Recurrent Neural Networks (BiRNN) referred as the Convolutional Bi-directional Recurrent Neural Network (CBiRNN). These CBiRNN models process both the vocal and lung datasets in parallel and feed the predicted results into the ensemble model for comprehensive evaluation. The experimental results show that the proposed CBiRNN model achieves 92% accuracy in voice disorder detection and 98% accuracy in respiratory disorder detection, while the ensemble model attains 98% accuracy for both voice and lung prediction. This innovative multimodal processing technique demonstrates significant potential in advancing health monitoring systems, offering a pathway to more accurate and reliable diagnostic tools.
Early life exposure to common pathogens and a high pathogen burden during childhood can have long-term effects on immune development and overall health. These infections can trigger molecular changes, including alterations in gene expression and DNA methylation (DNAm), which regulate immune and metabolic pathways. Our aim was to identify biological processes underlying differential patterns of DNAm and gene expression in whole blood by infection status in European children. In the Rhea (Greece) and INMA (Spain) cohorts, serum/plasma samples collected at mean ages of 4 and 8 years were analyzed by multiplex serology to measure IgG against 14 antigens from 9 pathogens, and blood collected at a mean age of 8 years was used for DNAm and gene expression profiling. Epigenome- and transcriptome-wide analyses were conducted to assess association with childhood infections. A total of 290 unique CpGs were significantly associated with pathogen outcomes: 265 with seropositivity, 111 with first exposure timing, and one with viral burden. Cytomegalovirus (CMV) exposure accounted for the largest number of both epigenetic (n = 325) and transcriptomic (n = 8) associations. A total of 89 CMV-related CpGs had been described before in adults, and among novel ones, 54 showed consistent effects in adults. CMV-related CpGs were enriched for SUZ12 targets linked to morphogenesis, oxidative stress, and cognition. A previously developed CMV episcore in adults predicted serologically assessed CMV infection at 4 and 8 years of age, with area under the curve values ranging from 0.74 to 0.78 (95% CI 0.68-0.83). We identified novel DNAm and gene expression signatures of common childhood infections, particularly CMV, implicating immune and morphogenesis pathways. A subset of CMV-related DNAm signals showed consistent associations with those reported previously in adults, suggesting similar molecular effects across ages.
Accurate detection and segmentation of moving objects constitute a fundamental challenge in computer vision, particularly for intelligent video surveillance systems operating under variable illumination, dynamic backgrounds, and environmental noise. This paper presents a fully unsupervised dual-phase motion analysis framework that effectively combines statistical independence modeling and geometric contour evolution to achieve high-precision motion detection and segmentation. In the first phase, an enhanced Fast Independent Component Analysis (Fast-ICA) algorithm is employed to perform statistical decomposition of video sequences, exploiting temporal independence to distinguish moving foregrounds from static backgrounds. This process generates an initial motion mask with strong robustness to illumination variation and noise artifacts. In the second phase, a hybrid level set segmentation model integrating the global Chan-Vese formulation and a locally adaptive Yezzi-based energy function refines object boundaries through an adaptive energy minimization process. A stabilization term and a self-regulating convergence criterion are further incorporated to ensure contour smoothness, numerical stability, and resilience to topological changes. Comprehensive experiments conducted on the CDNet-2014 benchmark dataset demonstrate that the proposed method achieves an average recall of 0.9613, precision of 0.9089, and F-measure of 0.9310, outperforming several state-of-the-art supervised, semi-supervised and unsupervised background subtraction algorithms. The proposed Fast-ICA-Level Set fusion framework thus provides a robust, adaptive, and computationally efficient solution for real-world intelligent surveillance and autonomous visual monitoring applications.
Reliable detection of bolt loosening in safety-critical infrastructure requires monitoring that captures microsecond-scale transients. However, processing 1 MHz vibration signals at the edge presents a fundamental dilemma: standard downsampling can smear sparse diagnostic impulses, whereas full-bandwidth processing is often computationally prohibitive. Here we propose PGRF-Net, a physics-guided deep learning framework that reconciles transient preservation with extreme compression. PGRF-Net integrates (1) a physics-guided resampling operator for high-rate vibration signals (e.g., 150 : 1 compression); (2) a heterogeneous gradient-spectral tensor representation that augments time-frequency information with morphology-aware channels; and (3) an asymmetric fusion module over two representations derived from the same signal (pseudo-image representation + waveform representation). On a six-class 1 MHz PVDF vibration dataset (109,668 segments; temporal-split test = 16,131), PGRF-Net reaches a best-run clean test accuracy of 95.12% under a block-wise temporal split. A strict file-disjoint split is also reported as a cross-scene hold-out protocol. On an independent Zenodo benchmark (3-fold file-disjoint CV, 9 runs per experiment), we evaluate the proposed pipeline together with three feature-engineering controls (Exp 501-503). These results support a practical compression-learning pipeline for industrial monitoring where both transient fidelity and computational efficiency are required.
Seizure forecasting and affective state analysis using EEG-ECG data play a pivotal role in advancing neurological and mental health monitoring. However, existing methods such as Fed-Transformer, Res-1D CNN, and Fed-ESD suffer from privacy risks, inefficient feature extraction, and high computational overhead, limiting their effectiveness in real-world applications. To overcome these challenges, this study proposes NeuroFedSense, a novel Federated Learning-enabled Privacy-Preserving Framework that integrates a Temporal Convolutional Network (TCN) with an Attention Mechanism for accurate seizure forecasting and affective state analysis using EEG-ECG data, ensuring enhanced feature selection, interpretability and efficient decentralized training. The model leverages adaptive attention-based optimization and weighted feature selection to improve classification performance while ensuring data privacy. Implemented using TensorFlow, NeuroFedSense achieves 99.54% accuracy, 99.62% precision, 99.34% recall, and a 99.46% F1-score, outperforming Fed-Transformer (97.10% accuracy), Res-1D CNN (81.62% accuracy), and FML (99.10% accuracy). The ROC-AUC score of 0.99 further establishes its superiority over competing models. Additionally, the federated approach reduces energy consumption per node by 30% and optimizes communication efficiency by minimizing data transmission by 15% over 100 rounds. By ensuring high accuracy, improved privacy, reduced computational overhead, and enhanced energy efficiency, NeuroFedSense sets a new benchmark for decentralized, real-time seizure prediction and affective state monitoring. These findings underscore its potential for deployment in intelligent, privacy-preserving healthcare applications, addressing critical challenges in remote neurological monitoring.