The Research Of MRI Radiomics For Glioma-associated Epilepsy

Glioma-associated epilepsy(GAE)is a common symptomatic diagnosis in glioma patients,with 50%of glioma patients and 89%of low-grade glioma patients experiencing seizures.It is currently believed that the occurrence of GAE may be attributed to various factors,including tumor location,peritumoral edema,changes in genetic background and microenvironment.GAE greatly reduces the quality of life of patients,and some patients may experience status epilepticus and multiple epilepsy seizures,even coma.Raising awareness of GAE,early prediction and appropriate treatment are essential to protect neurocognitive function and improve patients’quality of life.In glioma patients with epileptic seizures,in addition to tumor resection,the control of epilepsy is also an important goal.According to the frequency of seizures in patients with brain tumors,preventive use of antiepileptic drugs is an important measure in clinical treatment.It is generally recommended to start antiepileptic drugs when epilepsy is diagnosed preoperatively because of the high risk of recurrent seizures.For patients with younger age,temporal lobe gliomas and gliomas involving the cortex,although there is no preoperative epilepsy,preventive antiepileptic drugs can be used appropriately according to the surgical risk and individualized situation.Radiomics is a high-throughput data analysis method that extracts a large number of quantitative features from medical images and identifies features relevant to target information in an objective,reproducible,and noninvasive manner.Glioma radiomics has been used to study glioma grade,histological subtype and genotype,as well as to predict tumor proliferation and patient prognosis.Magnetic resonance single-sequence-based radiomics has been proven to have reliable performance in predicting the occurrence and types of GAE in glioma patients,but current research at home and abroad mainly focuses on LGG-related epilepsy,and in the prediction of GAE in high-grade gliomas The application has not yet been reported.The purpose of this study was to observe the relationship between clinical imaging manifestations and GAE,establish and verify the efficacy of radiomics models and nomograms to predict GAE,explore the imaging signs that affect the occurrence of GAE,and the effect of GAE on the prognosis of patients.This research is divided into the following three parts:Part 1 Radiomics for prediction of epilepsy in patients with frontal gliomaBackground and PurposeThe frontal lobe is a predisposing location for gliomas and a high incidence of GAE.To investigate the association between radiomic features and frontal gliomas associated epilepsy(GAE)and propose a reliable radiomics-based model to predict frontal GAE.Materials and methodsThis retrospective study consecutively enrolled 166 adult patients with frontal glioma(111 in the primary cohort and 55 in the testing cohort).A total 1130 features were extracted from T2-fluid attenuated inversion recovery images,including first-order statistics,3D shape,texture,and wavelet features.Regions of interest,including the entire tumor and peritumoral edema,were drawn manually.Pearson correlation coefficient,10-fold cross-validation,area under curve(AUC)analysis,and support vector machine were adopted to select the most relevant features to build a clinical model,a radiomic model and a clinic-radiomic model for GAE.The receiver operating characteristic curve(ROC)and AUC were used to evaluate the models’ classification performance in each cohort,and Delong’s test was used to compare the models’ performance.Two-sided t-test and Fisher’s exact test were used to compare the clinical variables.Statistical analysis was performed using SPSS software(version 22.0;IBM,Armonk,New York),and P<0.05 set as threshold for significant.ResultsThe classification accuracy of seven scout models,except the Wavelet first order model(0.793)and Wavelet texture model(0.784),was<0.75 in cross-validation.The clinic-radiomic model,including 17 magnetic resonance imaging-based features selected among the 1130 radiomic features and two clinical features(patient age and tumor grade),achieved better discriminative performance for GAE prediction in both the training(AUC=0.886,95%confidence interval[CI]=0.819-0.940)and testing cohorts(AUC=0.836,95%CI=0.707-0.937)than radiomic model(P=0.008)with 82.0 and 78.2%accuracy,respectively.ConclusionsRadiomics analysis can non-invasively predict GAE,thus allowing adequate treatment of frontal glioma.The clinical-radiomics model may enable a more precise prediction of frontal GAE.Further,age and pathology grade are important risk factors for GAE.Part 2 Radiomics nomogram may improve the prediction of epilepsy in patients with WHO Ⅱ-Ⅳ grade cerebral gliomasBackground and PurposeTo investigate the association between clinic-radiological features and glioma associated epilepsy(GAE),develop and validate a radiomics nomogram for predicting GAE in whole cerebral WHO grade Ⅱ～Ⅳ gliomas.Materials and MethodsThis retrospective study consecutively enrolled 380 adult patients with cerebral glioma(266 in the training cohort and 114 in the testing cohort).A total of 1130 features were extracted from T2-fluid attenuated inversion recovery images.Regions of interest,including the entire tumor and peritumoral edema,were drawn manually.The semantic radiological characteristics were assessed by a radiologist with 15 years of experience in neuro-oncology.Different combinations of algorithms for feature reduction,feature selection and classification were test to determine the most relevant features to build a clinic-radiological model,radiomic signature and a combined model for GAE.The combined model using the radiomics signature and three clinical characteristics was visualized as a radiomics nomogram.The receiver operating characteristic curve(ROC)and AUC were used to evaluate the model classification performance in both the training and testing cohorts.The Mc-Nemar test and Delong test were used to compare model performance.Two-sided t-test and Fisher’s exact test were used to compare the clinic-radiological signs between the GAE and non-GAE groups.Statistical analysis was performed using SPSS software(version 22.0;IBM,Armonk,New York),and P<0.05 was regarded as statistically significant.ResultsMale and young patients with glioma had a higher risk of epilepsy than female patients(P=0.002)and older patients(P<0.001).In the GAE group tumors were more likely to be in the left hemisphere,especially the left frontal lobe.The combined model reached the highest AUC in the testing cohort(training cohort,0.911[95%CI,0.878-0.942];testing cohort,0.866[95%CI,0.790-0.929]).The Mc-Nemar test revealed that the difference between accuracy of the clinic-radiological model,radiomic signature and combined model in predicting GAE in the testing cohorts(p>0.05)was not significantly different.The DeLong tests showed that the difference between the radiomic signature and the combined model’s performance was significant(p<0.05).ConclusionThe radiomics nomogram predicted seizures of patients with glioma noninvasively,simply and practically way.Compared with the radiomics models,comprehensive clinic-radiological imaging signs observed by the necked eye have reasonable and nondiscriminatory performance in prediction of GAE.Part 3 Effect of isocitrate dehydrogenase mutation in cerebral glioma-associated epilepsy and tumor recurrenceBackground and PurposeTo investigate the role of isocitrate dehydrogenase(IDH)genotype in cerebral glioma associated epilepsy(GAE)and tumor recurrence,and to develop and validate a radiomics model for predicting IDH genotype.Materials and MethodsThis retrospective study consecutively enrolled 823 adult patients with cerebral glioma and annoted and analysis the MR imaging signature of 751 patients.303 patients with presurgary GAE were obtained,and the epilepsy free or not after postsurgary had been followed up.Tumor recurrence and survival status of 523 patients in 18 months after surgery had been followed up.A total of 1130 features were extracted from T2-fluid attenuated inversion recovery images.Regions of interest,including the entire tumor and peritumoral edema,were drawn manually.The semantic radiological characteristics were assessed by a radiologist with 15 years of experience in neuro-oncology.Different combinations of algorithms for feature reduction,feature selection and classification were tested to determine the most relevant features to build a clinic-radiological model,radiomic signature and a combined model for IDH.The combined model using the radiomics signature and three clinical characteristics was visualized as a radiomics nomogram.The receiver operating characteristic curve(ROC)and AUC were used to evaluate the model classification performance in both the training and testing cohorts.The Delong test were used to compare model performance.Two-sided t-test and Fisher’s exact test were used to compare the clinic-radiological signs between the GAE and non-GAE groups.Logistic regression and Cox regression were used to analyze risk factors for tumor recurrence and seizures.Statistical analysis was performed using SPSS software(version 22.0;IBM,Armonk,New York),and P<0.05 was regarded as statistically significant.ResultsIDH had an important effect on tumor recurrence and preoperative GAE in the low-grade group,but not in the high-grade group.As a comprehensive risk factor,tumor grade is an independent risk factor for tumor recurrence.The influence of preoperative imaging features,radiomics features and surgical pathological indicators on postoperative epilepsy cannot accurately predict the control status of postoperative epilepsy.The clinical radiomics model can be used to accurately predict the IDH gene status(AUC=0.942),which is more efficient than the radiomics feature model(AUC=0.785),and the difference is statistically significant(P=0.01).The clinicradiological model(AUC=0.929)was better than the radiomics feature model,and the difference was not statistically significant(P=0.037);at the same time,the difference between the clinical radiomics model and the clinical radiological model was not statistically significant(P=0.876).Cox regression showed that the final screened model contained only tumor grades,P=0.13,HR=1.66(95%CI,1.11-2.49),indicating that tumor grade was an independent risk factor for tumor recurrence.Multivariate logistic regression analysis showed that the highest HR was the presence or absence of necrosis(HR=4.843,95%CI,1.54-15.2).In patients with low-grade glioma,the highest HR was the presence or absence of necrosis(HR=6.443,95%CI,1.83-22.62).In patients with high-grade gliomas,the model fitting information was obtained(P>0.99),and the significance of judging the model was not significant.ConclusionTumor necrosis was an independent risk factor for preoperative GAE,and IDH wild is a high risk factor for postoperative tumor recurrence in low-grade gliomas.Tumor grade was a comprehensive risk factor for preoperative GAE and postoperative tumor recurrence in all cohort.The role of IDH mutation statu in the analysis of postoperative GAE control should be combine more detailed intraoperative and postoperative information.