Study On Prediction Model Of Breast Cancer Sentinel Lymph Node Metastasis Based On Ultrasonography Combined With Peripheral Blood T Cell Subsets

Objective:To explore new methods for predicting sentinel lymph node metastasis in breast cancer and establish predictive models.Methods:Patients with single breast cancer diagnosed in Yancheng First People’s Hospital from 2020-10-01 to 2022-12-31 with complete preoperative peripheral blood T cell subsets analysis data and sentinel lymph node biopsy were continuously included.The gold standard for diagnosis was sentinel lymph node biopsy histopathology.The cases were randomly assigned to the training cohort and the validation cohort in an 8:2 ratio.Mann-whitney U test was used to analyze the difference of T cell subsets in peripheral blood of patients with positive and negative sentinel lymph nodes.Features such as major diameter,minor diameter,marginal malignancy,internal echo.calcification and parallel to skin ratio were collected in conventional ultrasound images of primary breast cancer tumors.Regions of interest were segmented by two ultrasound diagnostic doctors independently by 3DSlicer software.The radiomics features were extracted by Py-radiomics software package.Then,consistency test.Mann-Whitney U test and LASSO_CV were used for further screening,and finally Rad-score was calculated.The peripheral blood T cell model,conventional ultrasound model and radiomics model were constructed by logistic regression,naive baves,Support vector machine and classification decision tree in the training cohort,the best performing models were used to construct the combination model by logistic regression.Finally,area under ROC curve(AUC)was used to evaluate the diagnosis effect of the model in validation cohort,and clinical decision curve was drawn to evaluate the practical application value.Results:A total of 199 cases were included in this study,with sentinel lymph node positivity accounting for 33%.The number of total peripheral blood T cell and CD4 positive T cell were significantly increased in patients with sentinel lymph nodes metastasis,while there was no statistical difference in number of CD8 positive T cell.There were no statistical differences in age.pathological type,positive rate and molecular typing between the randomly divided training cohort and the validation cohort(P>0.05).The results of the inter-group/intragroup correlation coefficient suggested that the two physicians had a good consistency in the radiomics features.The conventional ultrasound model and peripheral blood T cell model which were constructed by classification decision tree have the best performance.The AUC of the conventional ultrasound model in the training cohort and the validation cohort were 0.712 and 0.678.respectively.The radiomics model which were constructed by logistic regression have the best performance.The AUC of radiomics model in the training cohort and validation cohort were 0.772 and 0.681.respectively.The AUC of peripheral blood T cell model in training cohort and validation cohort were 0.814 and 0.685.respectively.The AUC of combination model in the training cohort and validation cohort were 0.912 and 0.778.respectively.Compared with conventional ultrasound models,the combination model significantly improved the net benefit for patients.Conclusions:Primary breast cancer ultrasound imaging model,conventional ultrasound model and peripheral blood T cell model all have the value of clinical applications to diagnose sentinel lymph node metastasis of breast cancer,and the multi-factor model has better diagnostic effect,and the model is simple,non-invasive,good repeatability,and it can provide a certain basis for selecting surgical methods in breast surgery.