Electroencephalography (EEG) is a non-invasive neurophysiological monitoring technique. It records the electrical activity of the cerebral cortex using electrodes placed on the scalp surface. Owing to its high safety, portability, and millisecond-level temporal resolution, EEG has been widely utilized in a variety of fields, including clinical diagnosis, brain-computer interfaces (BCIs), and cognitive neuroscience research. However, due to its microvolt-level amplitude, EEG is highly susceptible to various artifacts, including electrooculographic (EOG), electrocardiographic (ECG), electromyographic (EMG), and power line interference (PLI). These artifacts can obscure genuine neural activity and introduce spurious electrophysiological features. Consequently, they may compromise EEG signal quality, thereby reducing the reliability of downstream analyses. To address this issue, numerous EEG artifact removal methods have been developed, including both traditional denoising techniques and deep learning-based approaches. Traditional EEG denoising methods have long served as the primary solutions for artifact removal. Representative approaches include filtering, regression, and blind source separation. Although these methods have demonstrated effectiveness in specific scenarios, they suffer from several inherent limitations. Filtering assumes that artifacts and EEG signals can be separated in the frequency domain, but many artifacts, such as EOG and EMG, overlap with EEG spectra, which may lead to the loss of valuable neural information. Regression methods require high-quality artifact references to estimate and subtract contaminations, limiting their effectiveness in reference-free scenarios. Blind source separation can remove artifacts without external references, but it typically requires the number of EEG channels to exceed the number of sources, restricting its application in single- or low-channel EEG recordings. Deep learning-based EEG denoising methods address these limitations effectively. First, they learn the nonlinear mapping between contaminated and clean EEG directly from data in an end-to-end manner. This approach does not rely on assumptions about spectral separability, thereby preserving neural activity more completely. Second, the reference information is incorporated during the training phase, allowing the trained model to perform artifact removal independently without external references. Third, deep learning models can be flexibly designed to accommodate various recording setups, achieving robust denoising for both high-density and single-channel EEG. Collectively, these advantages enable deep learning-based methods to overcome the main challenges of traditional approaches, providing more accurate and reliable EEG signal recovery. The superior denoising performance of deep learning-based EEG denoising methods has attracted increasing attention in EEG artifact removal research. As a result, many deep learning-based denoising methods have been developed and successfully applied in neural engineering areas. However, a systematic review of the techniques and applications in this field is still lacking. To address this gap, this paper reviews recent advances in deep learning-based EEG denoising from four perspectives: technical principle, benchmark dataset, denoising model, and evaluation method. Representative applications in neural signal analysis and BCI decoding are also summarized. Furthermore, the advantage, existing challenge, and future research direction of deep learning-based EEG denoising are discussed. This review aims to provide valuable theoretical insights and technical guidance for researchers. It is also expected to promote further advances and broader applications of deep learning-based EEG denoising techniques.

Consciousness, which underpins human perception, cognition, and behavioural responses, has long been a major focus of neuroscience and medicine. Alterations in consciousness occur not only during reversible processes such as sleep and anaesthesia, but also in pathological conditions, including disorders of consciousness (DOC) and brain injury. With recent advances in the synchronized acquisition and analysis of multimodal physiological signals, such as electroencephalography (EEG) and electrocardiography (ECG), the dynamic interactions between the heart and brain have become increasingly apparent. A growing body of evidence indicates that heart-brain interactions are closely associated with the maintenance and transition of consciousness states, offering a novel perspective for elucidating their underlying mechanisms and supporting relevant clinical assessments. Current approaches to heart-brain interaction analysis for consciousness-state assessment can be broadly grouped into three categories: ECG-based methods, EEG-based methods, and heart-brain coupling analysis. These approaches characterize heart-brain interactions during changes in consciousness from the complementary perspectives of autonomic regulation, cortical interoceptive processing, and cross-system information transfer, respectively. Nevertheless, this field remains at an early stage. Most studies have focused on a single experimental context or a fixed analytical method, with insufficient systematic consideration of the applicability and limitations of different approaches. Moreover, the mechanisms underlying the emergence and alteration of consciousness remain incompletely understood. Accordingly, this review examines heart-brain interaction methods employed in recent consciousness-state assessment studies in accordance with progressively deeper analytical perspectives, and systematically evaluates the advantages and limitations of these methods across distinct consciousness-related conditions. Heart rate variability (HRV) is one of the classical ECG-based indices used to investigate heart-brain interactions. It refers to fluctuations in consecutive interbeat intervals over a specified time period and is regarded as a physiological measure of autonomic nervous system regulation that may indirectly reflect the strength of heart-brain interactions. Depending on the analytical dimension, HRV analysis is commonly divided into time-domain, frequency-domain, and complexity-index analyses. These approaches characterize autonomic activity in terms of temporal fluctuations, the distribution of frequency components, and nonlinear dynamical properties, respectively. Although extensive research supports their physiological interpretation, they cannot characterize directional interactions from the heart to the brain. Among EEG-based methods for heart-brain interaction analysis, EEG activity directly related to cardiac fluctuations has received particular attention. One of the most widely used measures is the heartbeat-evoked potential (HEP). HEP-based approaches are considered to directly reflect the capacity of the cerebral cortex to process cardiac afferent signals and may be especially advantageous for characterizing rapidly changing consciousness levels, such as those observed during sleep. However, they do not comprehensively capture dynamic bidirectional information exchange between the brain and heart and are susceptible to cardiac artefacts, respiration, and variation in preprocessing procedures. Heart-brain coupling analysis quantifies functional associations between ECG and EEG through cross-system interaction measures, thereby assessing dynamic interactions between the heart and brain. It emphasizes information transfer, feedback regulation, and cross-timescale coordination between the two systems, offering a more comprehensive perspective for investigating changes in consciousness during sleep, anaesthesia, and DOC. However, the neurophysiological mechanisms and physiological interpretation of individual computational indices have not yet been fully established, and the robustness of their results and their potential for clinical generalization require further validation. Finally, this review summarizes the commonalities and differences among the three classes of methods across research domains involving different consciousness states. It further argues that heart-brain interaction research requires deeper investigation at both the mechanistic level of consciousness and the level of clinical application. Current challenges and key issues are identified, and future directions and feasible pathways for heart-brain interaction research are discussed, with the aim of providing new perspectives for researchers in related fields.

Objective To explore the risk factors for 28-day mortality of sepsis-associated acute kidney injury(SA-AKI)patients and to develop a nomogram risk prediction model.Methods A retrospective cohort study was conducted,involving 184 patients with SA-AKI admitted to the intensive care unit(ICU)of the 940th Hospital of Joint Logistic Support Force of PLA between January 2017 and December 2022.Patients were categorized into survival(n=135)and non-survival(n=49)groups based on 28-day mortality.Clinical data were collected,and statistically significant risk factors were preliminarily screened.Multivariate stepwise logistic regression analysis was performed to identify independent risk factors for 28-day mortality of SA-AKI patients.A nomogram predictive model was constructed using these factors,and internally validated with the Bootstrap method.The receiver operating characteristic curve(ROC curve)was drawn,and the area under the ROC curve(AUC)was calculated to verify the predictive value and accuracy of the model.Results The 28-day mortality rate among 184 SA-AKI patients was 26.6％(49/184).Multivariate stepwise logistic regression analysis identified multiple organ dysfunction syndrome(MODS)(OR=16.393,95％CI 4.317-62.254,P＜0.001),high acute physiology and chronic health evaluation Ⅱ(APACHE Ⅱ)score(OR=1.097,95％CI 1.036-1.161,P=0.002),low oxygenation index(OR=0.992,95％CI 0.986-0.998,P=0.015),low neutrophil count(OR=0.912,95％CI 0.860-0.968,P=0.002)and low fibrinogen concentration(OR=0.733,95％CI 0.549-0.978,P=0.034)as independent risk factors.The prediction model equation was P=1/1+e-logit(P),logit(P)=-1.626+2.797×MODS+0.092×AP ACHE Ⅱ+(-0.311)×fibrinogen+(-0.092)×neutrophil count+(-0.008)×oxygenation index.Internal validation with 1000 Bootstrap resamples showed high consistency between predicted and actual values.ROC analysis showed an AUC of 0.911(95％CI 0.868-0.955,P＜0.05)for the model,with 93.9％sensitivity and 78.5％specificity at a cut-off of 0.194.The Hosmer-Lemeshow test confirmed good calibration(P=0.62),and decision-making curve analysis demonstrated clinical utility within the high-risk threshold range(0.1-0.9).Conclusions MODS,high APACHE Ⅱ score,low oxygenation index,low neutrophil count,and low fibrinogen concentration are independent risk factors for 28-day mortality in SA-AKI patients.The developed nomogram risk prediction model may provide important guidance for predicting 28-day mortality in SA-AKI patients.

Objective To establish a stable,reliable,and clinically relevant porcine model of endotoxin-induced acute respiratory distress syndrome(ARDS).Methods Ten 8-month-old male Bama minipigs were deeply sedated,followed by invasive mechanical ventilation and electrocardiographic monitoring.Lipopolysaccharide(LPS)was intravenously pumped at 600 μg/(kg·h)for 3 hours,then maintained at 15 μg/(kg·h)thereafter.Dynamic monitoring was performed at five time points after LPS injection(LPS 0,1,3,5,and 8 h),including arterial blood gas analysis and chest computed tomography(CT)scans.Pathological examination of lung tissues obtained via bronchoscopic biopsy(HE staining and transmission electron microscopy)was conducted.These indicators were comprehensively used to evaluate the success of the animal model.Results At 5 hours after LPS administration,8 minipigs developed symptoms such as skin cyanosis,elevated body temperature,and respiratory distress.The oxygenation index decreased to<300 mmHg.Chest CT scans showed diffuse pulmonary infiltrates.Histopathology revealed alveolar edema and hyaline membrane formation.Transmission electron microscopy demonstrated disruption of pulmonary blood-air barrier,depletion of lamellar bodies in type Ⅱ pneumocytes,inflammatory cell infiltration,and exudation of plasma proteins and fibrin.Compared with LPS 0 h,at LPS 8 h,the oxygenation index and arterial blood pH were significantly decreased(P<0.001),while blood lactic acid and serum potassium were significantly increased(P<0.05);serum calcium and base excess were significantly decreased(P<0.05),and the lung injury score based on HE-stained lung sections was significantly increased(P<0.01).Conclusion The porcine ARDS model established by continuous LPS injection can dynamically simulate the pathophysiological characteristics and typical pathological manifestations of clinical septic ARDS,making it an effective tool to study the pathogenesis,prevention,and treatment strategies of septic ARDS.

BACKGROUND: Knee osteoarthritis (OA) is still challenging to prevent or treat. Enhanced endoplasmic reticulum (ER) stress and increased pyroptosis in chondrocytes may be responsible for cartilage degeneration. This study aims to investigate the effect of ER stress on chondrocyte pyroptosis and the upstream regulatory mechanisms, which have rarely been reported. METHODS: The expression of the histone methyltransferase enhancer of zeste homolog 2 (EZH2), microRNA-142-3p (miR-142-3p), and high mobility group box 1 (HMGB1) and the levels of ER stress, pyroptosis, and metabolic markers in normal and OA chondrocytes were investigated by western blotting, quantitative polymerase chain reaction, immunohistochemistry, fluorescence in situ hybridization, fluorescein amidite-tyrosine-valine-alanine-aspartic acid-fluoromethyl ketone (FAM-YVAD-FMK)/Hoechst 33342/propidium iodide (PI) staining, lactate dehydrogenase (LDH) release assays, and cell viability assessments. The effects of EZH2, miR-142-3p, and HMGB1 on ER stress and pyroptosis and the hierarchical regulatory relationship between them were analyzed by chromatin immunoprecipitation, luciferase reporters, gain/loss-of-function assays, and rescue assays in interleukin (IL)-1β-induced OA chondrocytes. The mechanistic contribution of EZH2, miR-142-3p, and HMGB1 to chondrocyte ER stress and pyroptosis and therapeutic prospects were validated radiologically, histologically, and immunohistochemically in surgically induced OA rats. RESULTS: Increased EZH2 and HMGB1, decreased miR-142-3p, enhanced ER stress, and activated pyroptosis in chondrocytes were associated with OA occurrence and progression. EZH2 and HMGB1 exacerbated and miR-142-3p alleviated ER stress and pyroptosis in OA chondrocytes. EZH2 transcriptionally silenced miR-142-3p via H3K27 trimethylation, and miR-142-3p posttranscriptionally silenced HMGB1 by targeting the 3'-UTR of the HMGB1 gene. Moreover, ER stress mediated the effects of EZH2, miR-142-3p, and HMGB1 on chondrocyte pyroptosis. In vivo experiments mechanistically validated the hierarchical regulatory relationship between EZH2, miR-142-3p, and HMGB1 and their effects on chondrocyte ER stress and pyroptosis. CONCLUSIONS A novel EZH2/miR-142-3p/HMGB1 axis mediates chondrocyte pyroptosis and cartilage degeneration by regulating ER stress in OA, contributing novel mechanistic insights into OA pathogenesis and providing potential targets for future therapeutic research.
Enhancer of Zeste Homolog 2 Protein/genetics* ; Osteoarthritis, Knee/pathology* ; Chondrocytes/metabolism* ; Pyroptosis/physiology* ; HMGB1 Protein/genetics* ; MicroRNAs/metabolism* ; Endoplasmic Reticulum Stress/genetics* ; Humans ; Animals ; Rats ; Male ; Rats, Sprague-Dawley ; Middle Aged

Numerous c-mesenchymal-epithelial transition (c-MET) inhibitors have been reported as potential anticancer agents. However, most fail to enter clinical trials owing to poor efficacy or drug resistance. To date, the scaffold-based chemical space of small-molecule c-MET inhibitors has not been analyzed. In this study, we constructed the largest c-MET dataset, which included 2,278 molecules with different structures, by inhibiting the half maximal inhibitory concentration (IC₅₀) of kinase activity. No significant differences in drug-like properties were observed between active molecules (1,228) and inactive molecules (1,050), including chemical space coverage, physicochemical properties, and absorption, distribution, metabolism, excretion, and toxicity (ADMET) profiles. The higher chemical diversity of the active molecules was downscaled using t-distributed stochastic neighbor embedding (t-SNE) high-dimensional data. Further clustering and chemical space networks (CSNs) analyses revealed commonly used scaffolds for c-MET inhibitors, such as M5, M7, and M8. Activity cliffs and structural alerts were used to reveal "dead ends" and "safe bets" for c-MET, as well as dominant structural fragments consisting of pyridazinones, triazoles, and pyrazines. Finally, the decision tree model precisely indicated the key structural features required to constitute active c-MET inhibitor molecules, including at least three aromatic heterocycles, five aromatic nitrogen atoms, and eight nitrogen-oxygen atoms. Overall, our analyses revealed potential structure-activity relationship (SAR) patterns for c-MET inhibitors, which can inform the screening of new compounds and guide future optimization efforts.

Image 1.

Objective:To explore the effect of C1q like protein 4(C1ql4)on migration and invasion of breast cancer and its mechanism.Methods:Real-time fluorescence quantitative PCR was used to detect the expression level of C1ql4 mRNA in tumor tissues and paracancerous tissues from 50 cases of breast cancer,as well as the breast cells including T549 and MCF-7 breast cancer cells,and MCF-10A normal breast cells.The siRNA targeting C1ql4 gene(siC1ql4)was synthesized and transfected into BT549 breast cancer cells by liposome transfection,and the silencing efficiency was detected by Western blotting.MTT assay was used to detect the proliferative ability of the cells in each group,the invasive ability of the cells was detected by Transwell assay,and the migratory ability of the cells was detected by scratch healing assay.Western blotting was used to detect the proliferative ability of the cells in each group.The expression levels of Snail,Slug,nuclear factor-κB(NF-κB)and phospho-NF-κB(p-NF-κB)were detected.The immunofluorescence was used to localize the distribution of NF-κB.Real-time fluorescence PCR was used to detect the mRNA expression levels of the NF-κB target genes TNF-α and IL-1β.Results:The expression level of C1ql4 mRNA in breast cancer tissues was significantly higher than that in paracancerous tissues,and the expression level of C1ql4 mRNA in BT549 and MCF-7 breast cancer cells was significantly higher than that in MCF-10A normal breast cells(P<0.05).After silencing C1ql4 expression,the proliferation,invasion and migration abilities of BT549 cells were significantly decreased(all P<0.001);the mRNA and prorein expression levels of Snail and Slug were significantly down-regulated(both P<0.001),and the mRNA expression levels of TNFα and IL-1β were significantly reduced(both P<0.001).The amount of NF-κB entering the nucleus was reduced,and the ratio of p-NF-κB to NF-κB was reduced(both P<0.05).Conclusion:The C1ql4 gene may regulate migration and invasion of breast cancer cells and promote breast cancer progression through the NF-κB pathway.