Objective To address the challenges of accurately capturing critical features in small-sample diabetic retinopathy(DR)recognition models in real-world applications,and the overly smooth distribution of true and false feature coefficients,we propose an enhanced small-sample DR recognition method based on an improved capsule network.Methods Firstly,the method enhances image feature representation by removing redundant boundary information and employing discrete wavelet transform based on the Haar wavelet function,thereby high-lighting critical pathological features.Secondly,the convolutional layer of the capsule network is optimized through a multi-branch architecture to extract multi-scale features from retinal images,while incorporating a convolutional block attention module that is subsequently fed into the capsule layer.Finally,the sigmoid function replaces the soft-max function in dynamic routing,thereby improving the model's robustness.Result The enhanced neural network model achieved an accuracy of 98.62%on the Kaggle public dataset following a rigorous selection and preprocessing procedure.Conclusion The enhanced capsule network demonstrated superior precision in identifying diabetic reti-nopathy within small sample sizes compared to other state-of-the-art algorithms currently available.

Objective To address the challenges of accurately capturing critical features in small-sample diabetic retinopathy(DR)recognition models in real-world applications,and the overly smooth distribution of true and false feature coefficients,we propose an enhanced small-sample DR recognition method based on an improved capsule network.Methods Firstly,the method enhances image feature representation by removing redundant boundary information and employing discrete wavelet transform based on the Haar wavelet function,thereby high-lighting critical pathological features.Secondly,the convolutional layer of the capsule network is optimized through a multi-branch architecture to extract multi-scale features from retinal images,while incorporating a convolutional block attention module that is subsequently fed into the capsule layer.Finally,the sigmoid function replaces the soft-max function in dynamic routing,thereby improving the model's robustness.Result The enhanced neural network model achieved an accuracy of 98.62%on the Kaggle public dataset following a rigorous selection and preprocessing procedure.Conclusion The enhanced capsule network demonstrated superior precision in identifying diabetic reti-nopathy within small sample sizes compared to other state-of-the-art algorithms currently available.

Background/Aims: Acute-on-chronic liver failure (ACLF) is a catastrophic illness. Few studies investigated the prognostic value of vitamin D-binding protein (VDBP) for hepatitis B virus (HBV)-related ACLF (HBV-ACLF) resulted in conflicting results. Methods: Two prospective HBV-ACLF cohorts (n=287 and n=119) were enrolled to assess and validate the prognostic performance of VDBP. Results: VDBP levels in the non-survivors were significantly lower than in the survivors (P<0.001). Multivariate Cox regression demonstrated that VDBP was an independent prognostic factor for HBV-ACLF. The VDBP level at admission gradually decreased as the number of failed organs increased (P<0.001), and it was closely related to coagulation failure. The areas under the receiver operating characteristic curve (AUCs) of the Child-Pugh-VDBP and chronic liver failuresequential organ failure assessment (CLIF–SOFA)-VDBP scores were significantly higher than those of Child-Pugh (P<0.001) and CLIF-SOFA (P=0.0013). The AUCs of model for end-stage liver disease (MELD)-VDBP were significantly higher than those of MELD (P= 0.0384) only in the case of cirrhotic HBV-ACLF patients. Similar results were validated using an external multicenter HBV-ACLF cohort. By longitudinal observation, the VDBP levels gradually increased in survivors (P=0.026) and gradually decreased in non-survivors (P<0.001). Additionally, the VDBP levels were found to be significantly decreased in the deterioration group (P=0.012) and tended to be decreased in the fluctuation group (P=0.055). In contrast, they showed a significant increase in the improvement group (P=0.036). Conclusions The VDBP was a promising prognostic biomarker for HBV-ACLF. Sequential measurement of circulating VDBP shows value for the monitoring of ACLF progression.

Objective:To discuss the effective of artesunate in the treatment of coronavirus disease 2019 (COVID-19).Methods:Using prospective method, the 43 cases of confirmed COVID-19 patients in Nanning Fourth People's Hospital from January 22nd to February 15th in 2020 were enrolled and divided into routine treatment group ( n = 25) and artesunate treatment group ( n = 18) by odd-even rule after admission. According to the guidelines, the routine treatment group was recommended to receive lopinavir/ritonavir 500 mg + α-aerosolized interferon 500×10 4 U, twice daily; the artesunate treatment group was given artesunate 60 mg, twice daily besides the routine treatment, for 10 days in both groups. During the treatment period, the pharynx swab test of 2019 novel coronavirus (2019-nCoV) nucleic acid was carried out every 2 days, and the routine blood test, liver and kidney functions, blood coagulation function and myocardial enzymes were re-examined. Chest CT was checked every 3 days after the treatment, and re-examined every 5 days after the condition was improved. The routine blood test and biochemical results of two groups were observed, and the efficacy evaluation was performed by monitoring the time for significant improvement of symptoms, negative conversion time of throat swab virus nucleic acid, lung lesion absorption time, adverse drug reactions and the length of hospital stay of the two groups. Results:There were no significant differences between the two groups in terms of gender, age, body weight, routine blood test and biochemical results before treatment. In artesunate treatment group, the time for significant improvement of symptoms (days: 3.33±1.91 vs. 4.84±2.19), negative conversion time of 2019-nCoV nucleic acid (days: 4.72±2.16 vs. 6.68±3.76), lung lesion absorption starting time (days: 5.39±2.36 vs. 7.48±3.78), lung lesion absorption greater than 70% time (days: 14.11±4.16 vs. 17.04±4.42) and the length of hospital stay (days: 16.56±3.71 vs. 18.04±3.97) were significantly shorter than those in routine treatment group, with significant differences (all P < 0.05). The incidence of adverse drug reactions in two groups had no significant difference (72.2% vs. 80.0%, P > 0.05). Conclusion:Artesunate can shorten the treatment time of COVID-19, improve prognosis and eliminate pathogens, with fewer adverse reactions and a good application prospect.

OBJECTIVE:To investigate the anti-HBV activities of deoxynojirimycin(DNJ)derivatives N—benzyl—1—DNJ(P-DNJ)and N—nonyl—1—DNJ(NN-DNJ)in vitro.METHODS:Human hepatoma carcinoma cell lines HepG2 2,2,15,which were transfected from HBV DNA,were taken as target cells,cells were cultured with different concentrations of test drugs, with HBsAg,HBeAg and HBV DNA in the cultured supernatant determined by time-resolved immunofluorometric assay and fluorescent quantitation PCR assay on day 6th and 10th day,meanwhile,with the cytotoxicity of DNJ derivatives determined by MTT assay.RESULTS:No cytotoxic effects was noted when the test concentration of P-DNJ and NN-DNJ was within 5～125?g?mL-1,HBsAg and HBeAg levels and HBV DNA secretion decreased significantly at a concentration of 125?g?mL-1.CONCLUSION:P-DNJ and NN-DNJ showed anti-HBV activity in the in vitro cell culture experiment.