Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the prospective associations between physical activity and incident ischemic stroke in adults.Methods:Data of China Kadoorie Biobank study in Tongxiang of Zhejiang were used. After excluding participants with cancers, strokes, heart diseases and diabetes at baseline study, a total of 53 916 participants aged 30-79 years were included in the final analysis. The participants were divided into 5 groups according to the quintiles of their physical activity level. Cox proportional hazard regression models was used to calculate the hazard ratios ( HR) for the analysis on the association between baseline physical activity level and risk for ischemic stroke. Results:The total physical activity level in the participants was (30.63±15.25) metabolic equivalent (MET)-h/d, and it was higher in men [(31.04±15.48) MET-h/d] than that in women [(30.33±15.07) MET-h/d] ( P<0.001). In 595 526 person-years of the follow-up (average 11.4 years), a total of 1 138 men and 1 082 women were newly diagnosed with ischemic stroke. Compared to participants with the lowest physical activity level (<16.17 MET-h/d), after adjusting for socio-demographic factors, lifestyle, BMI, waist circumference, and SBP, the HRs for the risk for ischemic stroke in those with moderate low physical activity level (16.17-24.94 MET-h/d), moderate physical activity level (24.95-35.63 MET-h/d), moderate high physical activity level (35.64-43.86 MET-h/d) and the highest physical activity level (≥43.87 MET-h/d) were 0.93 (95% CI: 0.83-1.04), 0.87 (95% CI: 0.76-0.98), 0.82 (95% CI: 0.71-0.95) and 0.76 (95% CI: 0.64-0.89), respectively. Conclusion:Improving physical activity level has an effect on reducing the risk for ischemic stroke.

Paroxysmal nocturnal hemoglobinuria (PNH) is an acquired clonal hematopoietic stem cell disease caused by abnormal expression of glycosylphosphatidylinositol (GPI) on the cell membrane due to mutations in the phosphatidylinositol glycan class A(PIGA) gene. It is commonly characterized by intravascular hemolysis, repeated thrombosis, and bone marrow failure, as well as multiple systemic involvement symptoms such as renal dysfunction, pulmonary hypertension, swallowing difficulties, chest pain, abdominal pain, and erectile dysfunction. Due to the rarity of PNH and its strong heterogeneity in clinical manifestations, multidisciplinary collaboration is often required for diagnosis and treatment. Peking Union Medical College Hospital, relying on the rare disease diagnosis and treatment platform, has invited multidisciplinary clinical experts to form a unified opinion on the diagnosis and treatment of PNH, and formulated the Expert Consensus of Multidisciplinary Diagnosis and Treatment for Paroxysmal Nocturnal Hemoglobinuria (2024), with the hope of promoting the standardization of PNH diagnosis and treatment.

Objective To investigate the clinical characteristics and prognosis of pneumococcal meningitis(PM),and drug sensitivity of Streptococcus pneumoniae(SP)isolates in Chinese children.Methods A retrospective analysis was conducted on clinical information,laboratory data,and microbiological data of 160 hospitalized children under 15 years old with PM from January 2019 to December 2020 in 33 tertiary hospitals across the country.Results Among the 160 children with PM,there were 103 males and 57 females.The age ranged from 15 days to 15 years,with 109 cases(68.1% )aged 3 months to under 3 years.SP strains were isolated from 95 cases(59.4% )in cerebrospinal fluid cultures and from 57 cases(35.6% )in blood cultures.The positive rates of SP detection by cerebrospinal fluid metagenomic next-generation sequencing and cerebrospinal fluid SP antigen testing were 40% (35/87)and 27% (21/78),respectively.Fifty-five cases(34.4% )had one or more risk factors for purulent meningitis,113 cases(70.6% )had one or more extra-cranial infectious foci,and 18 cases(11.3% )had underlying diseases.The most common clinical symptoms were fever(147 cases,91.9% ),followed by lethargy(98 cases,61.3% )and vomiting(61 cases,38.1% ).Sixty-nine cases(43.1% )experienced intracranial complications during hospitalization,with subdural effusion and/or empyema being the most common complication[43 cases(26.9% )],followed by hydrocephalus in 24 cases(15.0% ),brain abscess in 23 cases(14.4% ),and cerebral hemorrhage in 8 cases(5.0% ).Subdural effusion and/or empyema and hydrocephalus mainly occurred in children under 1 year old,with rates of 91% (39/43)and 83% (20/24),respectively.SP strains exhibited complete sensitivity to vancomycin(100% ,75/75),linezolid(100% ,56/56),and meropenem(100% ,6/6).High sensitivity rates were also observed for levofloxacin(81% ,22/27),moxifloxacin(82% ,14/17),rifampicin(96% ,25/26),and chloramphenicol(91% ,21/23).However,low sensitivity rates were found for penicillin(16% ,11/68)and clindamycin(6% ,1/17),and SP strains were completely resistant to erythromycin(100% ,31/31).The rates of discharge with cure and improvement were 22.5% (36/160)and 66.2% (106/160),respectively,while 18 cases(11.3% )had adverse outcomes.Conclusions Pediatric PM is more common in children aged 3 months to under 3 years.Intracranial complications are more frequently observed in children under 1 year old.Fever is the most common clinical manifestation of PM,and subdural effusion/emphysema and hydrocephalus are the most frequent complications.Non-culture detection methods for cerebrospinal fluid can improve pathogen detection rates.Adverse outcomes can be noted in more than 10% of PM cases.SP strains are high sensitivity to vancomycin,linezolid,meropenem,levofloxacin,moxifloxacin,rifampicin,and chloramphenicol.[Chinese Journal of Contemporary Pediatrics,2024,26(2):131-138]

Objective To analyze risk factors and pathogenic characteristics of intestinal colonization of carbape-nem-resistant Enterobacterales(CRE)in patients from intensive care unit(ICU).Methods A total of 392 ICU pa-tients who underwent intestinal CRE screening in a tertiary hospital in Changzhou from March to December,2023 were divided into the colonization group(n=42)and the non-colonization group(n=350)according to the screening results.Clinical data of patients,including age,gender,underlying diseases,malignant tumors,radiotherapy,chemotherapy,infection before the last screening,antimicrobial use,and invasive procedures were collected for the analysis on risk factors and pathogenicity.Results Among 42 patients with positive CRE screening results,44 CRE strains were detected,mainly Klebsiella pneumoniae(65.91％),followed by Escherichia coli(15.91％)and En-terobacter cloacae(13.64％).The average time from admission in ICU to positive screening results of intestinal CRE in the colonization group was 14 days.Long term use of carbapenem antibiotics(OR=1.47,95％CI:1.31-1.65),mechanical ventilation(OR=1.14,95％CI:1.06-1.22),and Enterobacterales infection(OR=10.10,95％CI:3.28-32.09)were independent risk factors for intestinal CRE colonization.Patients who received carbap-enem antibiotics for ≥15 days(x2=167.52,P＜0.001)and those who received mechanical ventilation for ≥15 days(x2=101.03,P＜0.001)had higher risks for intestinal CRE colonization.Conclusion In clinical practice,it is necessary to improve pathogen detection,treat Enterobacterales infection timely,choose carbapenem antibiotics carefully,shorten treatment course,actively evaluate indications for mechanical ventilation,and wean off ventilator timely.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.