Objective: To develop a dual-branch deep learning framework for accurate multi-label classification of fundus diseases, addressing the key limitations of insufficient complementary feature extraction and inadequate cross-modal feature fusion in existing automated diagnostic methods. Methods: The fundus multi-label classification dataset with 12 disease categories (FMLC-12) dataset was constructed by integrating complementary samples from Ocular Disease Intelligent Recognition (ODIR) and Retinal Fundus Multi-Disease Image Dataset (RFMiD), yielding 6 936 fundus images across 12 retinal pathology categories, and the framework was validated on both FMLC-12 and ODIR. Inspired by the holistic multi-regional assessment principle of the Five Wheels theory in traditional Chinese medicine (TCM) ophthalmology, the dual-branch multi-label network (DBMNet) was developed as a novel framework integrating complementary visual feature extraction with pathological correlation modeling. The architecture employed a TransNeXt backbone within a dual-branch design: one branch processed red-green-blue (RGB) images to capture color-dependent features, such as vascular patterns and lesion morphology, while the other processed grayscale-converted images to enhance subtle textural details and contrast variations. A feature interaction module (FIM) effectively integrated the multi-scale features from both branches. Comprehensive ablation studies were conducted to evaluate the contributions of the dual-branch architecture and the FIM. The performance of DBMNet was compared against four state-of-the-art methods, including EfficientNet Ensemble, transfer learning-based convolutional neural network (CNN), BFENet, and EyeDeep-Net, using mean average precision (mAP), F1-score, and Cohen's kappa coefficient. Results: The dual-branch architecture improved mAP by 15.44 percentage points over the single-branch TransNeXt baseline, increasing from 34.41% to 44.24%, and the addition of FIM further boosted mAP to 49.85%. On FMLC-12, DBMNet achieved an mAP of 49.85%, a Cohen’s kappa coefficient of 62.14%, and an F1-score of 70.21%. Compared with BFENet (mAP: 45.42%, kappa: 46.64%, F1-score: 71.34%), DBMNet outperformed it by 4.43 percentage points in mAP and 15.50 percentage points in kappa, while BFENet achieved a marginally higher F1-score. On ODIR, DBMNet achieved an F1-score of 85.50%, comparable to state-of-the-art methods. Conclusion DBMNet effectively integrates RGB and grayscale visual modalities through a dual-branch architecture, significantly improving multi-label fundus disease classification. The framework not only addresses the issue of insufficient feature fusion in existing methods but also demonstrates outstanding performance in balancing detection across both common and rare diseases, providing a promising and clinically applicable pathway for standardized, intelligent fundus disease classification.

Health economics evidence plays an important role in linking clinical value evidence with health resource allocation decisions in the development of clinical practice guidelines. It can not only effectively balance clinical effectiveness and economic feasibility but also avoid forming "idealized" recommendations that are detached from the affordability of the healthcare system or the burden-bearing capacity of patients. To promote guideline developers to use health economics evidence more standardizedly and fully, this paper conducts an in-depth analysis of the current application status, existing challenges, access channels, and application processes of health economics evidence in current guidelines, and on this basis, puts forward considerations and suggestions for strengthening and standardizing the application of health economics evidence in China's clinical practice guidelines.

BACKGROUND:Although researchers have noted that fibroblast growth factor receptor 1 shows great potential in rheumatoid arthritis bone destruction,there is a lack of reviews related to the potential mechanisms of fibroblast growth factor receptor 1 in rheumatoid arthritis bone destruction. OBJECTIVE:To comprehensively analyze the mechanism of fibroblast growth factor receptor 1 in bone destruction in rheumatoid arthritis by reviewing the relevant literature at both home and abroad. METHODS:We searched the CNKI database using the Chinese search terms"fibroblast growth factor receptor 1,rheumatoid arthritis,bone destruction,bone cells,osteoblasts,osteoclasts,chondrocytes,macrophages,synovial fibroblasts,T cells,vascular endothelial cells."PubMed database was searched using the English search terms"fibroblast growth factor receptor 1,rheumatoid arthritis,bone destruction,osteocytes,osteoblasts,osteoclasts,chondrocytes,macrophages,synovial fibroblasts,T cells,endothelial cells."The search period focused on April 1992 to January 2024.After screening the literature by reading titles,abstracts,and full texts,a total of 82 articles were finally included for review according to inclusion and exclusion criteria. RESULTS AND CONCLUSION:Fibroblast growth factor receptor 1 was found to be widely expressed in bone tissue-associated cells,including osteoblasts,osteoclasts,and osteoclasts.Fibroblast growth factor receptor 1 affects bone remodeling and homeostasis by regulating the function of these cells,as well as promoting the onset and progression of bone destruction in rheumatoid arthritis.Fibroblast growth factor receptor 1 is involved in the inflammatory response of synovial fibroblasts and macrophages and regulates angiogenesis of endothelial cells in synovial tissues.Fibroblast growth factor receptor 1 promotes bone destruction in several ways.Fibroblast growth factor receptor 1 may be a potential causative agent of bone destruction in rheumatoid arthritis and provides a reference for further research on its therapeutic targets.

Resistance to poly adenosine diphosphate(ADP)-ribose polymerase inhibitor(PARPi)presents a considerable obstacle in the treatment of ovarian cancer.F-box and tryptophan-aspartic(WD)repeat domain containing 11(FBXW11)modulates the ubiquitination of growth-and invasion-related factors in lung cancer,colorectal cancer,and osteosarcoma.The function of FBXW11 in PARPi therapy is still ambiguous.In this study,RNA sequencing(RNA-seq)showed that FBXW11 expression was raised in ovarian cancer cells that had been treated with PARPi.FBXW11 was abnormally expressed at low levels in high-grade serous ovarian cancer(HGSOC)tissues,and low levels of FBXW11 were associated with shorter overall survival(OS)and progression-free survival(PFS)in HGSOC patients.Overexpressing FBXW11 made ovarian cancer more sensitive to PARPi,while knocking down FBXW11 made it less sensitive.The four-dimensional(4D)label-free quantitative proteomic analysis revealed that FBXW11 targeted S100 calcium binding protein A11(S100A11)and promoted its degradation through ubiquiti-nation.The increased degradation of S100A11 led to less efficient DNA damage repair,which in turn contributed to increased PARPi-induced DNA damage.The role of FBXW11 in promoting PARPi sensitivity was also confirmed in xenograft mouse models.In summary,our study confirms that FBXW11 promotes the susceptibility of ovarian cancer cells to PARPi via affecting S10OA11-mediated DNA damage repair.

Background Some studies have suggested that exposure to multiple metals is closely linked to the development of metabolic syndrome (MS) in the populations, but the effect of aluminum exposure on MS remains unclear. Objective To analyze the prevalence and influencing factors of MS among employees with aluminum exposure in Shanxi Province. Methods Cluster sampling was employed to survey male frontline workers in an aluminum factory in Shanxi Province. Data on general demographic information, lifestyle, occupational history, medical history, and family history of chronic diseases were collected through questionnaires. The concentration of fasting blood glucose was determined using the glucose oxidase technique, and blood lipid levels were determined using the peroxidase method. Serum aluminum levels were detected using inductively coupled plasma-mass spectrometry (ICP-MS), and blood biochemical indicators were measured using the peroxidase method. Based on the China's 2020 diagnostic criteria for MS, the participants were and divided into an MS group anda non-MS group. Variables with statistical significance in univariate analysis were included to construct a logistic regression model. Results A cohort of 312 workers participated in this research, with 84 individuals diagnosed with MS, yielding a prevalence rate of 26.92%. The logistic regression model revealed that body mass index (BMI)≥24.0 kg·m⁻² (OR=1.967, 95%CI: 1.057, 3.659), alcohol consumption (OR=1.883, 95%CI: 1.063, 3.336), experiencing major life event (OR=3.886, 95%CI: 1.509, 10.008), family history of hypertension (OR=2.112, 95%CI: 1.162, 3.837), serum aluminum concentration (OR=1.024, 95%CI: 1.012, 1.035), alanine aminotransferase (ALT) level (OR=1.032, 95%CI: 1.011, 1.054), and white blood cell (WBC) count (OR=1.210, 95%CI: 1.001, 1.465) were significant influencing factors for MS. Conclusion BMI≥24.0 kg·m⁻², alcohol consumption, experiencing major life event, family history of hypertension, elevated serum aluminum concentration, increased ALT level, and elevated WBC count are risk factors for MS among occupationally aluminum-exposed workers.

To characterize the genomes of different emm-type group A Streptococcus (GAS), their virulence genes and drug resistance profiles in Tianjin City from 2011 to 2024. After PCR, a total of 42 strains with different years and emm types were selected for whole genome sequencing and multi-locus sequence typing (MLST), and the core genomes were used to generate a phylogenetic tree, after which the virulence genes and resistance genes were identified and analyzed, followed by the drug susceptibility test. In this study, the GAS strains were dominated by emm1 (50.0%) and emm12 (40.4%), and the MLST phenotypes were categorized into six types: ST36 (40.4%), ST1274 (26.1%), ST28 (23.8%), ST921 (4.7%), ST46 (2.3%), and ST403 (2.3%). There was a high consistency between their emm-types and ST types. A total of 68 virulence genes were detected in the genomes of 42 GAS strains, involving functional genes encoding exotoxin, bacterial adhesion, extracellular enzymes, etc. The virulence genes they carried were significantly different between emm1-type and emm12-type strains, such as speA. At the same time, the carrying rates of some virulence genes in the same emm-type strains changed with time, such as hyl. The resistance genes were basically the same among different emm-type strains except for the vanSE gene detected in all emm12 strains. The results of drug sensitivity showed that the GAS strains isolated in Tianjin City from 2011 to 2024 were sensitive to penicillin, cefazolin, chloramphenicol, vancomycin, and levofloxacin, while the resistance rates to erythromycin, azithromycin, clarithromycin, and clindamycin ranged from 88.5% to 100.0%, and there was a certain degree of consistency between the resistance phenotypes and the detected resistance genes. Overall, the main emm types and evolutionary features of GAS in Tianjin City from 2011 to 2024 were consistent with the dominant types in China, and the carrying rate of virulence genes and drug resistance genes differed significantly among different emm-type strains, and there were continuous evolution and variation in the prevalence of virulence genes in GAS.

Objective: To improve the efficiency and standardization of preoperative anemia diagnosis and treatment by establishing a systematic intelligent management platform for preoperative anemia. Methods: A multidisciplinary collaborative model was adopted to develop a preoperative anemia management system that integrates intelligent early warning, standardized treatment pathways, and quality control. The system utilizes natural language processing technology to automatically capture laboratory data and establish evidence-based medical decision support functions. A pre-post study design was employed to compare changes in preoperative anemia screening rates, preoperative anemia intervention rates, reasonable use of iron supplements, and perioperative red blood cell transfusion rates before and after system implementation. Results: After system implementation, the standardization of anemia diagnosis and treatment significantly improved: 1) Screening effectiveness: The anemia screening rate increased to 50.00% (an increase of 27.24%); 2) Intervention effectiveness: The anemia treatment rate rose to 56.30% (an increase of 14.02%); 3) Treatment standardization: The reasonable use rate of iron supplements increased to 55.33% (an increase of 21.02%); the red blood cell transfusion rate decreased to 18.29% (a decrease of 4.07%), and the amount of red blood cell transfusions was reduced by 291 units. Conclusion: This system achieves full-process management of preoperative anemia through information technology, significantly enhancing the standardization of diagnosis and treatment as well as intervention effectiveness, providing an effective solution for perioperative anemia management.

Objective: To investigate the influencing factors for cognitive function among aluminum workers, so as to provide the basis for intervention and prevention of cognitive function among aluminum-exposed populations. Methods: From July to August 2019, male aluminum workers in the electrolytic aluminum workshop of an aluminum factory in Shanxi Province were selected using the cluster sampling method. Demographic information, prevalence of chronic diseases, lifestyle behaviors, night shifts, and sleep quality were collected through questionnaire surveys. Blood aluminum levels were measured using inductively coupled plasma-mass spectrometry. Cognitive function was investigated using the Montreal Cognitive Assessment. Factors affecting cognitive function among aluminum workers were analyzed by a quantile regression model. Results: A total of 142 aluminum workers were surveyed, including 57 workers aged 20 to <40 years (40.14%) and 85 workers aged 40 to 60 years (59.86%). The median blood aluminum level was 38.23 (interquartile range, 21.82) μg/L. The median cognitive function score was 24.00 (interquartile range, 3.00) points. Quantile regression analysis revealed that older age (βQ5=-0.186, 95%CI: -0.269 to -0.102), lower educational level (βQ5=1.933, 95%CI: 1.029 to 2.838; βQ10=1.743, 95%CI: 0.480 to 3.006; βQ50=1.038, 95%CI: 0.141 to 1.935; βQ75=1.006, 95%CI: 0.437 to 1.575; βQ90=1.111, 95%CI: 0.291 to 1.930), smoking (βQ5=-2.056, 95%CI: -3.264 to -0.849), alcohol consumption (βQ5=-1.821, 95%CI: -3.247 to -0.396) and higher blood aluminum level (βQ5=-0.075, 95%CI: -0.110 to -0.040; βQ10=-0.078, 95%CI: -0.127 to -0.029; βQ50=-0.075, 95%CI: -0.110 to -0.040; βQ75=-0.057, 95%CI: -0.079 to -0.035; βQ90=-0.067, 95%CI: -0.099 to -0.035) were associated with cognitive function decline among aluminum workers. Conclusions Educational level and blood aluminum level are the main factors affecting the cognitive function among aluminum workers. Among those with lower cognitive function scores, age, smoking and alcohol consumption are also associated with cognitive function.

Objective To explore the effect of back-pushing manipulation on motor ability and skele-tal muscle mitochondrial function in rats with chronic fatigue syndrome(CFS).Methods Twenty-four male Sprague-Dawley rats were randomly divided into a blank group of 8 and a modeling group of 16.The CFS rat model was established using forced weight-bearing swimming combined with chronic stress stimulation for 21 days.After successful modeling,the modeling group was further divided into the modeling control(MC)group and the back-pushing manipulation(BPM)group,each of 8.In the back-pushing manipulation group,all rats were given daily 20-minute back pushing for 14 consecu-tive days after modeling.Then,all groups were recorded semi-quantitative scores of general conditions(SQS-GC),body mass,and exhaustion swimming time(EST).Moreover,the grasping ability of limbs was assessed by using the mesh screen test(MST).After the intervention,the tissue of the erector spinae muscle was taken to detect the adenosine triphosphate(ATP)content using the biochemi-cal method,and the activities of mitochondrial respiratory chain complexes Ⅰ to Ⅳ were detected by using the spectrophotometric method.Results After modeling,compared with the blank group,the SQS-GC of the MC group and BPM groups were higher(P<0.01),while EST was shorter and the MST score was lower(P<0.01 or P<0.05).However,after intervention,compared with the MC group,the SQS-GC of the BPM group was lower(P<0.01),while EST and the MST score was higher(P<0.01 or P<0.05).Compared with the blank group,the ATP content of the modelling group was significantly lower(P<0.01),and the activities of mitochondrial respiratory chain complexes I,II,III,and IV de-creased significantly(P<0.01).However,after intervention,all the above values of the BPM group in-creased significantly,compared with the MC group(P<0.01 or P<0.05).Conclusion The back-pushing manipulation can improve the mitochondrial function and energy metabolism,as well as the exercise ca-pacity,and alleviate fatigue of CFS rats,which may be related to the improvement of the activities of mitochondrial respiratory chain complexes I,II,III,and IV,and the increase of ATP content in the skeletal muscles.

Network pharmacology has gained widespread application in drug discovery, particularly in traditional Chinese medicine (TCM) research, which is characterized by its "multi-component, multi-target, and multi-pathway" nature. Through the integration of network biology, TCM network pharmacology enables systematic evaluation of therapeutic efficacy and detailed elucidation of action mechanisms, establishing a novel research paradigm for TCM modernization. The rapid advancement of machine learning, particularly revolutionary deep learning methods, has substantially enhanced artificial intelligence (AI) technology, offering significant potential to advance TCM network pharmacology research. This paper describes the methodology of TCM network pharmacology, encompassing ingredient identification, network construction, network analysis, and experimental validation. Furthermore, it summarizes key strategies for constructing various networks and analyzing constructed networks using AI methods. Finally, it addresses challenges and future directions regarding cell-cell communication (CCC)-based network construction, analysis, and validation, providing valuable insights for TCM network pharmacology.
Medicine, Chinese Traditional/methods* ; Artificial Intelligence ; Network Pharmacology/methods* ; Humans ; Drugs, Chinese Herbal/chemistry* ; Drug Discovery