HPV-dependent Squamous Cell Carcinoma Of The Cervix Prediction Model Based On Machine Learning

Objective: To explore the preoperative clinical data of patients with HPV infection,and to further select the predictors of HPV-dependent squamous cell carcinoma of the cervix,so as to construct the HPV-dependent squamous cell carcinoma of the cervix prediction model,in order to play a certain reference role for clinical decision-making.Methods: A total of 538 patients who were admitted to The Northern Theater Command General Hospital for cervical conization with a scalpel from January2019 to October 2022.were reviewed in this study.The differences in postoperative histology and pathology were compared,and variance analysis and test were performed for each factor to select the different influencing factors as follows: Age,Times,FIB,CA125,CA199,CEA,AFP,SCC.Based on these factors,six prediction models including GLM,RF,AdaBoost,SVM,LGBM and Gaussian were constructed by machine learning,and according to each model we have drawn one ROC curve.Finally we will chose the optimal one as the prediction model of our research after comprehensive comparison.Results :LGBM was the final prediction model of HPV-dependent squamous cell carcinoma of the cervix in this study.The accuracy of this model was 0.795,the sensitivity was 0.934,and the sensitivity was 0.359.Conclusions: Preoperative general information for women infected with HPV,Age,Times,FIB,CA125,CA199,CEA,AFP and SCC have the value of predicting HPVdependent squamous cell carcinoma of the cervix and can be used as the index of HPV-dependent squamous cell carcinoma of the cervix model construction.LGBM is the best model of HPV-dependent squamous cell carcinoma of the cervix prediction in this study.