Preoperative Assessment Of Ki-67 Status In Gastric Cancer With CT-based Radiomics Approach

Background:Gastric cancer is one of the most common malignant tumors in the digestive system.At present,the clinical treatment of gastric cancer is based on pathological diagnosis,and the treatment decision is determined by stage and prognosis.Due to the strong heterogeneity of gastric cancer prognosis,the exploration of specific prognostic biomarkers has always been a research hotspot.Among them,Ki-67 is considered as a potential biomarker for evaluating the biological behavior of gastric cancer.Its expression status is closely related to the occurrence,development and prognosis of gastric cancer.Accurate prediction of Ki-67 expression status has a guiding role in clinical decision-making.The immunohistochemical method of routine clinical Ki-67 detection is invasive and cannot dynamically detect its changes during treatment in a timely manner.In recent years,the popular radiomics can deeply explore the potential molecular information of images,which can provide a feasible direction for preoperative non-invasive prediction of Ki-67 expression.Objective:Using CT images to construct hand-crafted features and deep learning features,and fusion of clinical risk factors to build a radiomics model,to explore the diagnostic efficacy of the model on the expression of Ki-67 in patients with gastric cancer before surgery.Methods:The imaging and clinicopathological data of four hundred and sixtyeight gastric cancer patients who underwent contrast-enhanced CT scan of the upper abdomen and postoperative Ki-67 measurement from January 2009 to January 2019,were retrospectively analyzed.The enrolled patients were randomly divided into the training set(310 patients)and validation set(158 patients),and divided into high and low expression group according to the expression level of Ki-67(training and validation set:177 and 133 patients,79 and 79 patients,respectively).Based on the CT images of training set,hand-crafted features and deep learning features were respectively extracted and selected.Then a radiomics signature was developed.Next,multivariate logistic regression analysis was carried out in combination with clinical information to build a radiomics model that can individually predict the Ki-67 expression level of gastric cancer before surgery.The predictive ability of the signatures and fusion model for Ki-67 status were measured with receiver operating characteristic(ROC)curve in training and validation set respectively,and the area under the curve(AUC)was calculated.The calibration curve was used to evaluate the degree of fitting between the predicted Ki-67 status of the combined prediction model and the postoperative pathological reality.Decision curve analysis(DCA)was used to evaluate the threshold probability of net benefit of combined model.Two independent samples t test or chi-square test were used to analysis the clinicopathologic information and Rad-score between the high and low expression groups.Results:Twenty selected radiomics features(9 handcrafted features,11 deep learning features)were obviously associated with the Ki-67 level in gastric cancer.The radiomics signature was incorporated with the clinical risk factors(age)obtained by multivariate logistic regression analysis to build an imaging combined clinical prediction model,which was then visualized as a nomogram.The ROC curve showed that the AUC for predicting Ki-67 level with the radiomics only were 0.637(95%CI:0.570-0.704)and 0.724(95%CI:0.641-0.807)in the training dataset and validation dataset.The combined model with clinical risk factors improved the prediction of the Ki-67 level before surgery,and achieved the AUC of 0.656(95%CI:0.589～0.724)and 0.733(95%CI:0.650～0.816)in the training dataset and validation dataset.The calibration curves showed that the combined model has a better fit in the validation dataset than in the training dataset.The DCA showed the radiomics nomogram was clinically useful.Conclusion:Hand-crafted and deep learning labels based on CT images combined with clinical risk factors to establish a combined prediction model can be used as an atraumatic tool for preoperative evaluation of the expression of Ki-67 in gastric cancer patients,which is conducive to assist clinical decision-making and improve patient prognosis.