OBJECTIVETo assess the value of contrast enhanced MRI features for predicting epidermal growth factor receptor () gene amplification in glioblastoma multiforme (GBM) with radiomic method.

METHODSEighty patients withstatus examined GBM were retrospectively reviewed. The data were randomly divided into a training dataset (60%) and test dataset (40%). Texture features of each case were extracted from the enhanced region and the edema region in contrast enhanced MR images. Principal component analysis was used for dimension reduction. Random forest model, support vector machine model and neural network model were built. Area under the curve (AUC) of the receiver operating characteristics curve was used to assess the performance of models with test dataset.

RESULTSA total of 542 features were extracted from the enhanced region and the edema region. Forty-eight principal components were obtained, which accounted for 100% accumulation contribution rate, and the first 31 principal components were selected for models building, which accounted for 98.5% accumulation contribution rate. The values of AUCs were 0.74, 0.69 and 0.63 for random forest model, support vector machine model and neural network model in the test dataset, respectively.

CONCLUSIONSRadiomic method with proper model may have a potential role in predicting thegene status with enhanced MRI features derived from the enhanced region and the edema region in patients with glioblastoma multiforme.