The application of AV diagnostic criteria based on gram stain smear in the automatic detection of vaginal mi-croecology
Objective To integrate the combined diagnostic criteria of aerobic vaginitis(AV)based on gram stain smear combined with clinical features into the automatic detection system of vaginal microecology,verify its accuracy and simplify the experimental operation.Methods Based on AV combined diagnostic criteria,five evaluation indexes including white blood cell count,proportion of basal cells,lactobacillus grade,background colony and clinical character-istics(vaginal pH value,vaginal mucosal congestion and yellow secretions)were integrated into SHTARS Comet-60au high magnification microscopy analysis system.All samples in this study were obtained from January-August 2023 at out-patient clinics of Beijing Obstetrics and Gynecology Hospital,Capital Medical University for gynecological vaginal micro-ecology evaluation system testing.A total of 1184 samples were used for the model development of AV automatic diagno-sis and machine learning was carried out.AV was diagnosed when 0～2 points were scored for each index and ≥4 points were used.Clinical gold standard test results were used as the final diagnostic results,and 517 cases were used to validate the model to evaluate the accuracy of automated diagnosis.Results In the model building stage,the over-all consistency rate of Lactobacillus grading index was 73.8％,the consistency rate of white blood cell count was 68.6％,the consistency rate of background bacteria population was 32.7％,and the proportion of basal cells was 57.0％.The AV negative agreement rate was 98.3％,and the overall agreement rate of the two diagnostic methods was 96.5％,with no difference in the AV detection rate.In the further clinical verification stage,the overall agreement rate of the two detection methods was 87.02％,and the overall agreement rate of all indicators from high to low was the proportion of parasbasal cells(94.6％),background bacterial community(70.8％),lactobacillus classification(69.6％),white blood cell count(59.6％),and the agreement rate of all indicators were＞95％.Conclusion AV automatic detection based on gram stained smears combined with clinical characteristics of AV joint diagnostic criteria is feasible,AV automatic detec-tion and clinical gold standard detection overall consistency is high,but the sensitivity needs to be improved,a larger sample size in clinical application to optimize and verify the model.
国家科技期刊平台
NSTL
万方数据
