Exploring the influencing factors in urine cytology detection rates in urothelial carcinoma:an interpretable machine learning approach
Objective:Based on interpretable machine learning algorithms,this study analyzes the key clinical and pathological factors affecting the positive rate of acridine orange fluorescence staining of exfoliated urinary cells in the diagnosis of bladder cancer,aiming to optimize its diagnostic efficacy in bladder cancer detection.Methods:We selected 737 patients diagnosed with urothelial carcinoma who visited the Second Hospital of Tianjin Medical University from August 2019 to August 2022.We collected data on urine cytology results,concurrent urinalysis,urine sediment microscopy results,and postoperative pathology findings from the same period,including tumor location,number,grade,and Ki-67.We then used six machine learning models to verify the correlation between these variables and positive urine cytology results.Finally,we employed the Shapley Additive Explanation(SHAP)method to analyze the interpretability and importance of the optimal model,aiming to identify variables related to the positive detection rate of urine cytology among the clinical and pathological factors mentioned above.Results:Among the six machine learning algorithms,the random forest model performed the best.Interpretability analysis indicated that,in the urinalysis and sediment microscopy categories,the factors related to the positive de-tection rate,ranked by importance from highest to lowest,were the number of leukocytes observed under micros-copy,the number of non-squamous cells observed under microscopy,and the red blood cell count.Finally,while the importance of patients'age and gender was relatively low,they were still correlated with the positive detection rate of urine cytology.Tumor number and location showed no significant correlation with the positive detection rate of urine cy-tology.Conclusion:This study explored the clinical and pathological factors influencing the positive rate of acridine orange fluorescence staining in urine cytology.It enhanced the understanding of sensitive factors affecting urine cytology testing and provided guidance for the application scenarios of acridine orange fluorescence staining technology in urine cytology.
urine cytologymachine learningurothelial carcinomabladder cancerroutine urine test
NETL
NSTL
万方数据
