Fig. 2
From: Changes in radiomic and radiologic features in meningiomas after radiation therapy
Workflow of the development and testing of a radiomics model. First, lesions were semiautomatically segmented on MRI scans for radiomic analysis. Second, a total of 1691 radiomics features were extracted. Third, in the training phase, the 10 most relevant features were selected with classic minimum redundancy maximum relevance. The random forest (RF) model was built and validated with the 10-fold cross-validation method. Fourth, in the test phase, the RF model was tested with a validation test set. AUC = area under the receiver operating characteristic curve