Preoperative Staging CT-based Radiomics Signature In Predicting Ki-67 Status In Breast Cancer

Background:Breast cancer is the most frequent malignancy and leading cause of cancer-related death in women.In the era of precision medicine,Ki-67 as a tumor biomarker plays a more important role in guiding the individualized clinical care for patients with breast cancer.Clinically,the Ki-67 status is mainly obtained by immunohistochemistry.This method is not only invasive,but also easy to ignore the heterogeneity of tumor,with certain limitations.Radiomics,as a non-invasive,comprehensive and objective evaluation tool,can provides a new idea for preoperative prediction of the Ki-67 status.Nowadays,the chest CT enhanced scan has become a routine examination for preoperative staging of breast cancer patients,on this basis,using radiomics to mining the valuable information contained in staging CT will not increase the radiation dose and financial burden of patients,therefore,it is expected to provide a non-invasive method for preoperative prediction of the Ki-67 status of breast cancer patients.Objective:This study aims to explore the value of preoperative staging CT-based radiomics signature for noninvasi ve prediction of Ki-67 status of breast cancer patients.Materials and Methods:In this study,245 invasive breast cancer patients who met the inclusion and exclusion criteria from May 2016 to December 2017 were retrospectively collected,all patients were performed routine staging enhanced chest CT before surgery.The corresponding clinicopathological data were collected from the patient’s electronic medical records.The dataset was divided into the primary cohort(145 patients,May 2016 to May 2017)and the validation cohort(100 patients,June 2017 to December 2017)according to examination time.In the training cohort,regions of interest(ROIs)were manually delineated around the tumor profile,radiomics features were extracted from the volume of interest(VOI)of tumor.Using the minimum redundancy maximum relevance algorithm and Boruta algorithm,we performed feature selection.Then,the radiomics signature was constructed by using the multivariable logistic regression.The predictive performance of radiomics signature was evaluated in the training cohort with respect to its discrimination,calibration,clinical usefulness by using ROC curve,Calibration curve,Decisionmaking curve.Finally,the radiomics signature was validated in the validation cohort.Result:After feature selection,eight key features were identified.The radiomics signature containing those key features were built to predict Ki-67 status in breast cancer.The radiomics signature shows favorable predictive performance in predicting Ki-67 status,with the area under curve(AUC)of 0.782(95%confidence interval[CI]:0.691-0.874)and 0.781(95%CI:0.686-0.876)in the training and validation cohort,respectively.Conclusion:The staging CT-based radiomics signature shows favorable performance in predicting Ki-67 status in breast cancer,can provide additional value of preoperative staging CT for clinical treatment decisions,it may serve as a noninvasive approach to facilitate the preoperative prediction of Ki-67 status in clinical practice.