The Value of CT-Based Radiomics for Preoperative Prediction Ki-67 Expression Level of the Esophagogastric Junction Adenocarcinoma
Abstract
Purpose
To investigate and improve the diagnostic performance of preoperative prediction of the Ki-67 expression index using CT-based radiomics.
Methods
The imaging and clinicopathological data of 100 esophagogastric junction adenocarcinoma patients who had a Ki-67 measurement and contrast-enhanced CT scan of the abdomen were collected. According to the expression index of Ki-67, they were divided into high expression (60 cases) and low expression (40 cases). According to the ratio of 7:3, the enrolled cases were randomly divided into the training set（70 cases）and the test set（30 cases）. The univariate analysis was used to compare the clinical and pathological features of the high and low expression groups of Ki-67. Tumor ROIs were manually segmented on axial contrast-enhanced CT venous phase images (CE-CT). Radiomics features were extracted from axial contrast-enhanced CT venous phase images based on the training set.
Results
The univariate analysis showed that age, gender, CT-reported lymph node status, CEA, CA125, CA199, CA724 and the major location of the esophagogastric junction adenocarcinoma had no statistically significant difference between the high and low expression groups (P＞0.05). The ROC curve showed that the AUC for preoperative prediction of Ki-67 expression level with the radiomics model was 0.812 and 0.795 in the training and test sets, respectively.
Conclusion
The efficiency of CT-based radiomics in predicting Ki-67 expression index in esophagogastric junction adenocarcinoma is ideal.