MRI-based radiomics model for predicting endometrial cancer with high tumor mutation burden.

Abstract

Purpose: To evaluate the performance of MRI-based radiomics in predicting endometrial cancer (EC) with a high tumor mutation burden (TMB-H).

Methods: A total of 122 patients with pathologically confirmed EC (40 TMB-H, 82 non-TMB-H) were included in this retrospective study. Patients were randomly divided into training and testing cohorts in a ratio of 7:3. Radiomics features were extracted from sagittal T2-weighted images and contrast-enhanced T1-weighted images. Then, the logistic regression (LR), random forest (RF), and support vector machine (SVM) algorithms were used to construct radiomics models. The area under the receiver operating characteristic curve (AUC) was calculated to evaluate the diagnostic performance of each model, and decision curve analysis was used to determine their clinical application value.

Results: Four radiomics features were selected to build the radiomics models. The three models had similar performance, achieving 0.771 (LR), 0.892 (RF), and 0.738 (SVM) in the training cohort, and 0.787 (LR), 0.798 (RF), and 0.777 (SVM) in the testing cohort. The decision curve demonstrated the good clinical application value of the LR model.

Conclusions: The MRI-based radiomics models demonstrated moderate predictive ability for TMB-H EC and thus may be a tool for preoperative, noninvasive prediction of TMB-H EC.