Value Of MRI Radiomics In The Differential Diagnosis Of High-grade Glioma And Single Brain Metastasis

Objective:To study MRI radiomics parameters is of great value in differentiating high-grade glioma from solitary brain metastases.Methods:Firstly,67 patients with high-grade gliomas and 61 patients with single brain metastases who were admitted to the First Affiliated Hospital of Shanxi Medical University were selected in the postoperative pathological results,and then their T1WI-contrast enhanced images were selected to segment the region of interest,a total of1034 radiomic features were extracted,and the features useful for the differential diagnosis of the two were selected by variance analysis,Pearson correlation coefficients and recursive feature elimination methods,and the features with strong correlation were used in logistic regression,support vector machine and random forest three classifier models,the classification performance of the model is measured in the training set using accuracy,sensitivity and specificity,and the area under the receiver operating characteristic curve.Apply the trained classifier model to the test set.Five-fold cross-validation was used to evaluate the predictive performance of the model.Results:Five salient features were selected for the three classifiers,and the three classifiers showed that the AUCs of the training set were logistic regression 0.9028,support vector machine 0.9361,and random forest 0.8750,of which the support vector machine classifier had the best performance,Its AUC in the test group was 0.8352.Conclusion:MRI radiomics is multi-parameter,and has high sensitivity,specificity and accuracy in identifying and diagnosing high-grade gliomas and single brain metastases.Therefore,a high-throughput computer-aided diagnosis system can be constructed,and it has great clinical value.