In recent years, generative artificial intelligence (AI) technologies have achieved remarkable progress in the early screening, risk prediction, disease progression assessment, and clinical trial design of glaucoma.Using advanced algorithms, such as generative adversarial networks, variational autoencoders, and diffusion models, researchers have synthesized high-quality structural images of the optic disc, macular region, and retinal nerve fiber layer, which effectively alleviates the limitations of scarce clinical imaging data and label imbalance.These methods have substantially improved the accuracy and generalization of deep learning models in visual field defect prediction, structure-function mapping, and longitudinal disease progression simulation.Meanwhile, multimodal generative approaches that integrate imaging data, visual field tests, and clinical features have facilitated individualized prediction of glaucoma progression.In addition, large language models have shown preliminary potential in ophthalmic image interpretation, clinical text information extraction, and decision support, providing new insights into intelligent ophthalmic diagnosis and treatment.However, the clinical implementation of generative AI in glaucoma faces challenges.The pathological authenticity and cross-device consistency of generated images require further validation, which may affect the reliability of early glaucoma detection.The heterogeneous characteristics of different glaucoma subtypes, such as open-angle and angle-closure glaucoma, also limit the generalization of synthetic data.Moreover, issues related to model interpretability (" black-box" nature), artifact generation, data privacy, and ethical governance remain key barriers to clinical translation.In the future, it is expected that establishing large-scale training frameworks that incorporate multicenter, multimodal, and multiethnic datasets will enhance model robustness and clinical applicability.Furthermore, generative AI may contribute to remote ophthalmic care and personalized precision therapy by enhancing low-quality image, reconstructing missing data, and simulating dynamic disease courses.This article reviews the current applications, core technologies, and challenges of generative AI in glaucoma diagnosis and management, and discusses its future directions and translational potential in clinical ophthalmology.

Objective To construct a high-precision statistical shape model of the humerus and systematically describe its anatomical variation patterns.Methods In this study,a statistical shape model of the humerus was constructed using the three-dimensional model data of 60 collected humerus cases.The principal component analysis method was adopted to reveal the main patterns of humerus anatomical variations and their contribution rates.Results The results showed that the first five principal components(PC01-PC05)collectively explained 96.6%of the total anatomical variations.Among them,PC01 and PC02 were the main components,contributing 66.6%and 23.5%of the variations respectively.PC01 mainly reflects the scaling effect of the overall size(length/width)of the humerus,while PC02 reveals the length variation characteristics independent of the overall scaling,which may reflect individualized differences.The subsequent principal components(PC03-PC05)depicted the local morphological characteristics and fine changes of the proximal and distal humerus.Conclusion The statistical shape model constructed in this study provides a reliable digital basis for personalized prosthesis design,surgical planning and biomechanical simulation.

In recent years, generative artificial intelligence (AI) technologies have achieved remarkable progress in the early screening, risk prediction, disease progression assessment, and clinical trial design of glaucoma.Using advanced algorithms, such as generative adversarial networks, variational autoencoders, and diffusion models, researchers have synthesized high-quality structural images of the optic disc, macular region, and retinal nerve fiber layer, which effectively alleviates the limitations of scarce clinical imaging data and label imbalance.These methods have substantially improved the accuracy and generalization of deep learning models in visual field defect prediction, structure-function mapping, and longitudinal disease progression simulation.Meanwhile, multimodal generative approaches that integrate imaging data, visual field tests, and clinical features have facilitated individualized prediction of glaucoma progression.In addition, large language models have shown preliminary potential in ophthalmic image interpretation, clinical text information extraction, and decision support, providing new insights into intelligent ophthalmic diagnosis and treatment.However, the clinical implementation of generative AI in glaucoma faces challenges.The pathological authenticity and cross-device consistency of generated images require further validation, which may affect the reliability of early glaucoma detection.The heterogeneous characteristics of different glaucoma subtypes, such as open-angle and angle-closure glaucoma, also limit the generalization of synthetic data.Moreover, issues related to model interpretability (" black-box" nature), artifact generation, data privacy, and ethical governance remain key barriers to clinical translation.In the future, it is expected that establishing large-scale training frameworks that incorporate multicenter, multimodal, and multiethnic datasets will enhance model robustness and clinical applicability.Furthermore, generative AI may contribute to remote ophthalmic care and personalized precision therapy by enhancing low-quality image, reconstructing missing data, and simulating dynamic disease courses.This article reviews the current applications, core technologies, and challenges of generative AI in glaucoma diagnosis and management, and discusses its future directions and translational potential in clinical ophthalmology.

Objective To construct a high-precision statistical shape model of the humerus and systematically describe its anatomical variation patterns.Methods In this study,a statistical shape model of the humerus was constructed using the three-dimensional model data of 60 collected humerus cases.The principal component analysis method was adopted to reveal the main patterns of humerus anatomical variations and their contribution rates.Results The results showed that the first five principal components(PC01-PC05)collectively explained 96.6%of the total anatomical variations.Among them,PC01 and PC02 were the main components,contributing 66.6%and 23.5%of the variations respectively.PC01 mainly reflects the scaling effect of the overall size(length/width)of the humerus,while PC02 reveals the length variation characteristics independent of the overall scaling,which may reflect individualized differences.The subsequent principal components(PC03-PC05)depicted the local morphological characteristics and fine changes of the proximal and distal humerus.Conclusion The statistical shape model constructed in this study provides a reliable digital basis for personalized prosthesis design,surgical planning and biomechanical simulation.

Objective To understand the genotyping of human immunodeficiency virus (HIV) co-infected hepatitis C virus (HCV) patients in Yunnan Province, and to analyze the differences in viral load, biochemical indicators, and blood routine indicators among different genotypes, in order to provide a laboratory basis for the diagnosis and clinical treatment of HIV/HCV co-infected patients. Methods From November 2022 to June 2023, the serum samples and basic information of patients diagnosed with HIV/HCV co-infection were collected in the antiviral outpatient clinic of Yunnan Provincial Hospital of Infectious Diseases. The HCV viral load was detected by one-step qRT-PCR amplification, the positive samples were sequenced, and genotyping was determined based on NS5 gene sequence. The differences in biochemical and blood routine indexes between HIV patients co-infected with different HCV genotypes and low/high viral loads were analyzed. Results A total of 126 HIV/HCV co-infected patients were collected, including 20 HCV genotype 1 (15.9%), 91 HCV genotype 3 (72.2%), and 15 HCV genotype 6 (11.9%). The maximum and minimum viral load of the three HCV genotypes were as follows: HCV type 1 (1.0×108, 4.8×104 IU/mL), HCV type 3 (2.2×108, 2.9×102 IU/mL), and HCV type 6 (8.1×107, 6.8×104 IU/mL). The results showed that there was no significant difference between HIV co-infection with different genotypes of HCV and three HIV treatment schemes, including nucleoside reverse transcriptase inhibitors+integrase strand transfer inhibitors (NRTIs+INSTIs), nucleoside reverse transcriptase inhibitors+non-nucleoside reverse transcriptase inhibitors (NRTIs+NNRTIs) and nucleoside reverse transcriptase inhibitors+protease inhibitor (NRTIs+PLs), and the viral load of patients (P>0.05). The analysis of biochemical indexes such as total bilirubin (TBIL), direct bilirubin (DBIL), alanine aminotransferase (ALT), aspartate aminotransferase (AST), creatinine (CREA), and blood routine indexes such as white blood cell (WBC), red blood cell (RBC), hemoglobin (HGB), platelet (PLT), mean corpuscular hemoglobin (MCH) and mean corpuscular hemoglobin concentration (MCHC) among different HCV genotypes and low/high viral loads showed that there was no significant difference in biochemical indexes and blood routine indexes between low/high viral loads of HIV co-infected HCV patients (P>0.05); however, the biochemical indicators TBIL, IBIL and MCHC were significantly different statistically between patients with genotype 3 HCV infection and those with genotype 1 HCV infection (P<0.05), while other biochemical and blood routine indexes were not statistically different among different HCV genotypes (P>0.05). Conclusions There are six subtypes of HCV co-infection in HIV patients in Kunming, Yunnan Province, including three genes of genotype 1, 3, and 6. Among them, genotype 3 HCV is the main prevalent genetic virus among HIV co-infected populations. The TBIL, IBIL and MCHC values of HIV patients co-infected with HCV type 3 are different from those infected with HCV type 1.

Objective To investigate the application value of tenofovir alafenamide fumarate (TAF) in elderly patients with chronic hepatitis B (CHB) and its influence on bones and kidneys. Methods A total of 36 CHB patients, aged ≥60 years, who received TAF antiviral therapy in Qingdao Municipal Hospital, The Affiliated Hospital of Qingdao University, Qingdao Sixth People's Hospital, Chengyang People's Hospital, and Jimo People's Hospital from June 2021 to October 2022 were enrolled in this study, and all patients received TAF (25 mg/d) antiviral therapy. Related data were collected at baseline and weeks 24 and 48 of treatment, including virological indicators, biochemical parameters, urinary protein electrophoresis indices, transient elastography (FibroScan), and bone mineral density. Virological indicators included high-sensitivity HBV DNA quantification; biochemical parameters included total bilirubin, direct bilirubin (DBil), indirect bilirubin (IBil), alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase, gamma-glutamyl transpeptidase, total bile acid (TBA), glucose, blood urea nitrogen, creatinine, estimated glomerular filtration rate, and cystatin C (Cys C); urinary protein electrophoresis indices included urinary β2 microglobulin (β2-MG), urinary retinol (URBP), and urinary α1 microspherin (α1-MG). The paired t -test was used for comparison of normally distributed continuous data before and after treatment, and the Wilcoxon signed-rank test was used for comparison of non-normally distributed continuous data before and after treatment; the chi-square test or the Fisher's exact test was used for comparison of categorical data. Results A total of 36 CHB patients completed 24 weeks of follow-up. The complete virological response rate after 24 weeks of treatment was higher than that at baseline [83.3% (30/36) vs 77.8% (28/36), χ 2 =0.36, P =0.55], and there were significant reductions in DBil ( t =-2.42, P =0.02) and Cys C ( t =-4.34, P < 0.001) from baseline to week 24. A total of 18 CHB patients completed 48 weeks of follow-up. The complete virological response rate after 48 weeks of treatment was higher than that at baseline (94.4% vs 77.8%, χ 2 =2.22, P =0.34), and there were significant increases in IBil ( t =2.43, P =0.03), TBA ( Z =-2.24, P =0.03), and bone mineral density T score of lumbar vertebra ( t =2.92, P = 0.01) and femoral neck ( t =2.42, P =0.03) and a significant reduction in liver stiffness measurement ( t =-2.31, P =0.03). There were no significant changes in β2-MG, URBP, and α1-MG after treatment (all P > 0.05). Conclusion TAF has a good antiviral effect in CHB patients aged ≥60 years and can help more CHB patients achieve complete virological response, without causing damage to the kidney, and it can also improve bone mineral density and liver fibrosis degree.

Objective To investigate the effect of long-chain non-coding RNA(lncRNA)AP000892.6 on the migration,inva-sion and epithelial-mesenchymal transition of bladder cancer cells and the related molecular mechanism.Methods The expression of AP000892.6 in bladder cancer and adjacent tissues was analyzed using GEPIA dataset.The expression levels of AP000892.6 in bladder cancer cell lines 647V,T24,UMUC3,5637,RT4 and normal bladder epithelial cells SV-HUC-1 were detected by real-time quanti-tative polymerase chain reaction(RT-qPCR).The AP000892.6 overexpression plasmid or negative control plasmid was transfected into T24 cells,and the experiment was divided into AP000892.6 group and NC group.The plasmid transfection efficiency was verified by RT-qPCR.The migration and invasion abilities of transfected T24 cells were detected by cell scratch method and Transwell assay,re-spectively.The downstream target gene of AP000892.6 was predicted and validated using bioinformatics method and dual-luciferase re-porter gene experiment.The expression of AP000892.6downstream target gene was detected by RT-qPCR.Western blot was used to de-tect the expression of epithelial phenotype proteins E-cadherin,ZO-1 and mesenchymal phenotype proteins Vimentin,N-cadherin and Twist.Results Compared with adjacent tissues,the expression of AP000892.6 was decreased in bladder cancer tissues(P＜0.01).Compared with SV-HUC-1 cells,the expression of AP000892.6 was decreased in bladder cancer cell lines(P＜0.05),and the rela-tive expression level of AP000892.6 in T24 cells was the lowest(P＜0.01).Compared with the NC group,overexpression of AP000892.6 inhibited the migration(P＜0.01)and invasive ability(P＜0.01)of bladder cancer T24 cells.Dual-luciferase reporter gene analysis proved that miR-616-5p was the target gene of AP000892.6(P＜0.01).Compared with the NC group,AP000892.6 negatively regu-lated the expression of miR-616-5p(P＜0.01).Compared with the NC group,the expression of epithelial phenotype proteins E-cad-herin and ZO-1 were up-regulated in T24 cells of AP000892.6group,and the expression of mesenchymal phenotype proteins Vimentin,N-cadherin and Twist were down-regulated.Conclusion AP000892.6 inhibits the migration,invasion and epithelial-mesenchymal transition process of bladder cancer T24 cells by targeting miR-616-5p.

Objective:To confirm the reuse of mismatched regenerated motor axons of brachial plexus and explore the effect of target organs on their regeneration in a rat model.Methods:This study was carried out between January 2021 and December 2021 at the research laboratory of the Department of Microsurgery, Orthopaedic Trauma and Hand Surgery, the First Affiliated Hospital of Sun Yat-sen University. Animals were randomly assigned into 2 groups, as a regeneration group (RGen) with 5 subgroups and a reuse group (RUs) with 3 subgroups. There were 6 rats per subgroup with 42 rats in total. It was observed that in the groups of RGen1-4, after the transection and suture of the musculocutaneous nerve, the motor axons of the proximal end could accurately grow into the distal corresponding endoneural tube. It was also observed that in the mismatched regenerated group, motor axons were the axons that grew into the endoneurial tube of the lateral forearm cutaneous nerve (LFCN), and other non-target organ contacts were made to the regenerated nerves after mismatch. It was specifically further divided into RGen1, the group without an organ for nerve to make contact with; RGen2, the group with skin as the target organ with nerves contact by neurorrhaphy; RGen3, the group with skin as the target organ with originally reserved natural nerve contact; RGen4, the group with muscle as the target organ with nerves contact by neurorrhaphy and RGen5, a control group. After 8 weeks, the positive area (PA), mean density (MD) and integral optical density (IOD) were measured, with AChE and ChAT fluorescence staining of the medial branch of LFCN, to evaluate the regenerated nerves after mismatch. Of the RUs group, firstly, the innervating branches of the flexor carpi radialis (FCR) were dissected and exposed, then further assigned according to initially innervated FCR (RUs1), contacted with regenerated nerves after mismatch (RUs2) and denervated (RUs3), respectively. After 8 weeks, compound muscle action potential (CMAP) and wet weight ratio of FCR were taken. Masson staining of FCR was also performed to evaluate muscle reinnervation by the regenerated nerves after mismatch. Data analysis with One-Way ANOVA and Bonferroni 0.05 indicated a statistically significant difference.Results:In the RGen groups, after AChE staining, the PA, MD and IOD of RGen3 and RGen4 were higher than that of RGen1 and RGen5, and PA of RGen4 were higher than that of RGen2, with a statistically significant difference ( P<0.05). After ChAT staining, the values of PA and IOD of RGen3 and RGen4 were higher than that of RGen1 and RGen5, and PA of RGen4 were higher than that of RGen2, with a statistically significant difference ( P<0.05). In the RUs, electrophysiological assessment showed that no CMAP was observed in RUs3, there was no significant difference in Latency of RUs1 and RUs2. The difference was statistically significant ( P<0.05). Wet weight rate of muscle of RUs1 (98.91%±3.86%) was higher than that of RUs3 (86.67%±4.68%) with a statistically significant difference ( P<0.01), but no significant difference when compared with RU2 (92.74%±3.88%). Masson staining showed that the CVF value of RUs2 (8.61%±1.16%) was significantly higher than that of RUs1 (3.17%±0.76%), and statistic significantly lower than that of RUs3 (16.44%±2.26%)( P<0.01). Conclusion:Target organ contact can promote the regenerated nerves after mismatched regeneration, and the muscle target organs exhibit greater facilitation than the cutaneous target organs. Besides, regenerated nerves after mismatch can establish effective innervation with muscle target organs, comfirming their effective reuse.

ObjectiveTo propose an estimation method of lower limb joints torques based on Gaussian process to achieve the accurate estimation of the lower limb joint torques in human gait. MethodsAccording to the characteristics of the natural gait joint torques curve, the squared exponential kernel function was selected to explore the interrelationship between the joint angle and the joint torques. A data fusion model based on Gaussian process was established. The lower limb joint angle was used as the input of the model, and the output was the joint torque. ResultsOne healthy subject walked on the gait running platform with a walking speed of 0.8 m/s. Three joint-torque experiments were conducted using the proposed Gaussian process. The results showed that most of the predicted values fell within the confidence intervals and 89% of the r² values were greater than 0.8. ConclusionThis method could achieve accurate estimation of joint torques. The potential application of this research is to optimize exoskeleton robots, control active prosthesis, and adjust the joint torque of humanoid robots.

Objective:To investigate the preoperative risk factors affecting early extremity blood supply after repair of major arterial injury so as to provide clues for prevention of limb ischemia.Methods:The clinical data were retrospectively analyzed of the 139 patients (140 extremities) with major extremity arterial injury who had been admitted to Department of Microsurgery, Orthopaedic Trauma and Hand Surgery, The First Hospital Affiliated to Sun Yat-sen University from January 2003 to December 2019. There were 112 males and 27 females, with a mean age of 30 (20, 44) years. The primary outcome was the early status of blood supply to the injured extremity (48 hours after surgery). Univariate analysis was conducted of such factors as gender, age, ischemia time, injury mechanism, injury site, fracture, soft tissue lesion, and duration of surgery. The significant factors ( P<0.1) were then analyzed by logistic regression, and P<0.05 was considered statistically significant. Results:Ischemia happened in 44 (31.4%, 44/140) extremities within 48 hours after surgery. There were significant differences in injury mechanism, ischemia time, fracture, and soft tissue lesion between patients with and without postoperative extremity ischemia ( P<0.05). Logistic regression analysis indicated that blunt injury ( OR=5.639, 95% CI: 1.068 to 29.761, P=0.042) and soft tissue lesion ( OR=12.568, 95% CI: 3.402 to 46.431, P<0.001) were significant preoperative risk factors affecting the early blood supply after repair of major extremity arterial injury. Conclusion:As blunt injury and soft tissue defect are preoperative risk factors for early extremity ischemia after repair of major extremity arterial injury, surgeons should pay more attention to them when assessing patients and making repair protocols.