Establishment Of A Bagged-trees-based Model For Prediction Of Sentinel Lymph Node Metastasis For Early Breast Cancer Patients

Objective: According to several large clinical trials,sentinel lymph node biopsy has gradually replaced routine axillary lymph node dissection,which has become one of the routine surgical methods in early breast cancer patients.However,sentinel lymph node biopsy can still bring some complications,although the incidence is low but should not be ignored.Therefore,several research centers design and use mathematical models to predict the pathological state of sentinel lymph nodes before operation,however,there are differences in patient sources,surgical methods and pathological evaluation methods in various countries and regions,so the predictive variables included in each research center are also different.Our aim is to use the new algorithm to establish a widely used and high-precision sentinel lymph node metastasis Prediction model based on the clinical and pathological characteristics of early breast cancer patients,and to explore its clinical application prospects.Methods: The clinicopathological data of 633 patients with breast cancer who underwent sentinel lymph node(SLN)examination from January 2011 to December 2014 were retrospectively analyzed.Because of the unbalanced clinical data caused by the natural incidence of breast cancer,this will affect the accuracy of the algorithm and cause bias to the results,so we use the smote algorithm to sample the data to increase the balance of the data.For the first time,our study analyzed the shape of the tumour and the amount of mammary gland glands as a variable.At the same time,because the breasts as a three-dimensional organ,we break the routine according to the quadrant to group the location of the swelling method,and the use of vectors(the clock direction of the swelling and distance from the nipple)combined with the Quadrant method to analyze the primary location of the tumor,at the same time we use the method of simply using quadrant or vector for comparing.Predictive models are established by means of single factor analysis,logistic regression and smote-bagged-trees.In this study,SPSS 22.0 and matlab2018 a were used for data processing and R software was used for statistical analysis.The first type of error level is set at 0.05.First,logistic regression modeling is used to calculate the area under the ROC curve of the model.Then the bagged-tree algorithm is used to classify the samples,and the sentinel lymph node metastasis Prediction model is established by smote-bagged tree algorithm and 5-fold cross-verification method.We evaluated the clinical value of the model in early breast cancer patients by confusing matrix and area under the subject's operational characteristics(ROC)curve(AUC).Results: A single factor and multi-factor analysis of the dataset were carried out,and 4 independent predictors related to sentinel lymph node metastasis were obtained.Through the results of multi-factor analysis,We used logistic regression to established the model,in the modeling and validation groups,the area under the curve(AUC)were 0.660 and 0.580.The Sentinel lymph node metastasis Prediction model was established by smote-bagged tree method,which contained the following 12 variables: Age,Body mass index(BMI),quadrant,clock direction,distance of tumor from nipple,shape of tumor molybdenum target,mammary gland content,tumor size,ER,PR,HER-2,ki67.We use bagged tree to categorize samples and 5-fold cross-validation.Finally,the area under the ROC curve obtained by our model is 0.801,and the accuracy of the model is 70.3% by confusing matrix calculation.We used the vector combining quadrant method to analyze the original location of the tumor,which is more precise than simply using vector or quadrant(AUC 0.801 vs.0.791 vs.0.701,Accuracy 70.3 vs.70.3 vs.63.6%).Conclusion: This model can be more reliable and stable to help doctors to predict the sentinel lymph node metastasis in breast cancer patients before operation.For patients with low transfer risk,especially those who cannot tolerate sentinel lymph node biopsy operations,or who have higher requirements for the retention of postoperative limb sensory function,it is clinically possible to consider free sentinel lymph node biopsy in order to improve the quality of life of patients.For patients with high transfer risk,armpit treatment can be performed directly,such as ALND,axillary radiotherapy,especially for elderly or patients with poor foundation status,can greatly shorten the operation time.