Analysis of preoperative nutritional and immune indicators and risk prediction model of nomogram in female breast cancer patients

JIANG Min; HUANG Yue; WANG Zun

doi:10.13820/j.cnki.gdyx.20241288

Abstract：Objective The characteristics of preoperative nutritional and immune indicators in female breast cancer patients were analyzed,and three breast cancer prediction models were constructed and the optimal models were se-lected.Methods A total of 450 female patients with breast mass were selected and divided into breast cancer group and benign control group according to the pathological results,and their basic information and laboratory results at the time of admission were collected.Their BMI,inflammation-related indicators and prognostic nutritional index(PNI)were calcu-lated,and a breast cancer prediction model based on three machine learning algorithms,including random forest(RF),extreme gradient boosting(XGBoost)and logistic regression,was constructed and evaluated.Results Univariate analy-sis showed that the BMI,neutrophil to lymphocyte ratio(NLR),between the two groups,Systemic Inflammatory Response Index(SIRI),PNI,lymphocyte to monocyte ratio(LMR),mass length were significant differences(P＜0.05).Among the three models,the XGBoost model had the best prediction effect,and the area under the ROC curve(AUC)for predic-ting breast cancer was 0.903 and 0.809 in the training group and test group,respectively(The test group demonstrated an accuracy of 0.76,with a sensitivity of 0.691,specificity of 0.823,and an F1-score of 0.729).The calibration curve showed that the predictive probability of the model for breast cancer has a good fit with the actual probability.The order of importance of risk factors for developing breast cancer is:mass length,LMR、PNI、BMI、SIRI、NLR.Conclusion In this study,a machine learning model based on XGBoost algorithm was constructed,which has good reference value in the pre-diction of breast cancer.