摘要
Abstract
Objective To explore the value of a combined detection model of six serum tumor markers and artificial intelligence algorithm in the diagnosis of lung cancer.Methods A retrospective study was conducted on 374 patients with pulmonary diseases admitted to Hangzhou Cancer Hospital from January 2020 to December 2022.Among them,191 patients had lung cancer(lung cancer group),and 183 patients had benign pulmonary lesions(benign lesions group).Two hundred healthy subjects underwent physical examination during the same period were selected as the control group.The levels of neuronspecific enolase(NSE),cytokeratin fragments 21-1(CYFRA21-1),squamous cell carcinoma antigen(SCC),carcinoma embryonic antigen(CEA),carbohydrate antigen 19-9(CA19-9),and carbohydrate antigen 125(CA125)in the serum of each group were detected by chemiluminescence method.The correlation analysis of these six serum tumor markers with lung cancer and the screening of important features for modeling were carried out,and finally the important features for modeling were screened out and included in the subsequent modeling.The lung cancer groups and the control group were divided into a modeling cohort(n=274)and a testing cohort(n=117)at a ratio of 7:3.Eight different combined detection models were constructed in the modeling cohort and the testing cohort by combining eight different artificial intelligence algorithms.The combined detection model with the optimal efficacy was selected,and the diagnostic efficacy of this model for the lung cancer group,early lung cancer group,benign lesions group,and control group was tested in the overall cohort(n=574).Results All six serum tumor markers could be used as important features for establishing a model to distinguish between the lung cancer group and the control group.After screening models based on different artificial intelligence algorithms in the testing cohort,the combined detection model based on the stochastic gradient boosting(SGB)algorithm showed to be with the optimal efficacy.The efficacy of this model was further evaluated in the overall cohort,which showed an area under the curve(AUC)of 0.874,and the sensitivity,specificity,and accuracy of 0.789,0.800,and 0.795,respectively for distinguishing between the lung cancer group and the control group;and an AUC of 0.800,and the sensitivity,specificity,and accuracy of 0.958,0.525,and 0.688,respectively for distinguishing between the early lung cancer group and the control group.The AUC of the model could still remain at 0.700 when distinguishing between the lung cancer group and the benign lesions group.Conclusion The combined detection model of six serum tumor markers and the artificial intelligence algorithm SGB has certain application value in the auxiliary diagnosis of lung cancer,which makes up for the deficiency of a single tumor marker and improves the diagnostic efficacy of lung cancer.关键词
肺癌/肿瘤标志物/人工智能算法/联合检测Key words
Lung cancer/Tumor markers/Artificial intelligence algorithm/Combined detection
