The Study Of Identification Of Glioma Postoperative Recurrence And Radiation Necrosis By MR Perfusion Imaging And Radiomics

Part I The study of identification of glioma postoperative recurrence and radiation necrosis by perfusion imaging based on 3D-ASL and DSC-PWIObjective:To explore the value of 3D-ASL and DSC-PWI in the diagnosis and differential diagnosis of glioma postoperative recurrence and radiation necrosis.Materials and Methods:47 patients with glioma who underwent postoperative MRI follow-up with abnormal enhanced lesions were selected.25 cases of postoperative recurrence and 22 cases of radiation necrosis were confirmed by the second operation pathology or long-term follow-up clinical diagnosis.All cases underwent T1 Weighted contrast-enhanced and 3D-ASL,33 cases underwent DSC-PWI.The values of hemodynamic parameters related to focal reactivity and normal contralateral image were measured,including ASL-CBF,DSC-CBF,DSC-CBV,DSC-MTT and DSC-TTP.Then the perfusion value of the focus was standardized(mean perfusion value of the focus/normal part of the contralateral mirror image).The ratio was the mean relative perfusion ratio of the focus,ASL-rCBF,DSC-rCBF,DSC-rCBV,DSC-rMTT and DSC-rTTP respectively.The differences of the hemodynamic parameters of the newly increased abnormal focus in the two groups were compared by t test,P<0.05 was considered statistically significantResults:Postoperative recurrence of glioma and radiation necrosis were enhanced on CE-T1WI showed garland-like,flocculent or irregular ring-shaped reinforcement ASL-CBF,DSC-CBF,and DSC-CBV perfusion in newly-enhanced lesions in patients with postoperative recurrence of gliomas were higher than those in the contralateral side Except that CE-T1WI enhanced areas in 2 patients with radiation necrosis presented high perfusion in ASL-CBF,DSC-CBF and DSC-CBV,the perfusion of ASL-CBF,DSC-CBF and DSC-CBV in other radiation necrosis patients showed equal-low performance compared with the contralateral side.The hemodynamic parameters of ASL-rCBF,DSC-rCBF and DSC-rCBV in the postoperative recurrence group of glioma were significantly higher than those in the radiation necrosis group(P<0.01),the diagnostic efficiency is the highest with the ASL-rCBF critical value 2.08.While,there were no statistical differences for DSC-rMTT and DSC-rTTP parameters between the postoperative recurrence and the radiation necrosis(P=0.58,0.37)Conclusions:Both 3D-ASL and DSC-PWI can distinguish postoperative recurrence and radiation necrosis of glioma with excellent performance,and the efficacy of ASL-rCBF is slightly higher than that of DSC-rCBF and DSC-rCBV in differentiating the postoperative recurrence and radiation necrosis of gliomaPart II The study of radiomics based on contrast-enhanced MRI images in the identification of postoperative recurrence and radiation necrosis of gliomaObjective:To explore the value of radiomics based on contrast-enhanced MRI images in the identification of postoperative recurrence and radiation necrosis of gliomaMaterials and Methods:58 patients(71 lesions)with glioma who underwent postoperative MRI follow-up with abnormal enhanced lesions were selected.30 cases(39 lesions)of postoperative recurrence and 28 cases(32 lesions)of radiation necrosis were confirmed by the second operation pathology or long-term follow-up examination.49 lesions(27 postoperative recurrence,22 radiation necrosis)were assigned into the training set,the remaining 22 lesions(12 postoperative recurrence,10 radiation necrosis into the validation set,according to a ratio of 7:3.By using Darwin intelligent scientific research platform,the newly added lesions with abnormal enhancement on the contrast-enhanced MRI(CE-T1WI)axial images were manually outlined.The feature selection was carried out through the three steps of F-test feature screening,LASSO dimension reduction algorithm and support vector machine iterative screening.Finally,the SVM algorithm was used to build the model.The diagnostic efficacy of the diagnostic model was evaluated by ROC curve.The differences of the characteristic parameters between the postoperative recurrence group and radiation necrosis group were compared by t test,P<0.05 was considered statistically significant.Results:A total of 1223 quantitative image feature parameters were extracted,100 features were coarse screened by F-test,7 features were screened by LASSO dimension reduction algorithm,and the support vector machine iterative filter was further optimized to 4,including skewness,joint entropy,MCC and strength.When SVM with linear kernel function was adopted,the accuracy,sensitivity and specificity of the training set were 83.7%,86.4%and 81.5%respectively,and the accuracy,sensitivity and specificity of the verification set were 86.4%,80.0%and 91.7%respectively When using SVM with radial basis function,the accuracy,sensitivity and specificity of the training set were 85.7%,90.9%and 81.5%respectively,and the accuracy,sensitivity and specificity of the verification set were 86.4%,90.0%and 83.3%respectivelyConclusions:The SVM based radiomics has a high accuracy in the identification of postoperative recurrence and radiation necrosis of glioma.The SVM model based on radial basis function is more effective than linear kernel function.