Objective Exploring the value of a multisequence peritumor intratumor radiomics model for predicting path-ological grading of gliomas.Methods A retrospective analysis was conducted on MRI data from 153 patients with patho-logically confirmed gliomas,including 50 cases of low-grade gliomas and 103 cases of high-grade gliomas.Preprocessed T1WI,T2WI,T2FL AIR,and T1C imaging data were imported into 3D Slicer for region of interest delineation.The regions were automatically expanded to 2mm,3mm,4mm,5mm,and 6mm around the tumor.Radiomic features were extracted using pyradiomics in Python.Feature selection was performed using t-tests,Pearson correlation analysis,and the Least Absolute Shrinkage and Selection Operator(LASSO)model.Six classifiers were employed to build peritumoral and intratumoral ra-diomic models separately.After selecting the optimal peritumoral model,the best peritumoral features(all with a peritumoral margin of 3 mm)for each sequence were combined with intratumoral features to construct an intratumoral+peritumoral model.The performance of different prediction models was evaluated using the receiver operating characteristic(ROC)curve and the area under the curve(AUC).The DeLong test was used to compare the differences in AUC between different models.Results The patients gender and age were not significantly associated with the pathological grading of gliomas(P＞0.05).Among the constructed predictive models,the model using both peritumoral and intratumoral imaging features across four sequences combined with the MLP classifier demonstrated the best predictive performance,achieving an AUC of 0.987 in the test set,with an accuracy of 0.935,specificity of 0.929,and sensitivity of 0.906.Conclusion The peritu-moral radiomics model also exhibited excellent predictive performance,comparable to that of the intratumoral radiomics model,and was able to enhance the predictive performance of the intratumoral radiomics model.The combined peritumoral and intratumoral radiomics provides a more accurate non-invasive method for preoperatively determining the pathological grade of gliomas,offering critical information for the development of individualized and precise treatment plans,as well as for prognosis evaluation in clinical practice.
GliomaPathological gradeRadiomicsPeritumoral brain spaceCombinatorial model
NETL
NSTL
万方数据
