The Value Of MRI Radiomics Of Glioma And Peritumoral Edema In Evaluating Tumor Recurrence

BACKGROUNDGlioma is the most common primary central nervous system tumor in adults,accounting for about 50%of intracranial tumors.The annual incidence rate of glioma in my country is（3-6.4）/100,000,and the annual death toll reaches 30,000.Glioma has the characteristics of strong invasiveness,invasive growth and high resistance to radiotherapy and chemotherapy.The high postoperative recurrence rate has become the main factor affecting the prognosis of patients.Therefore,how to predict tumor recurrence as soon as possible,and implement targeted treatment,has become the key to improving its efficacy.Current imaging techniques,such as PET-CT,MRI,etc.,play an important role in the prediction of postoperative recurrence of glioma,but are limited by multiple factors such as image resolution,evaluation parameters,and doctor’s experience.Imaging is still unable to meet the requirements of clinical precision treatment in the assessment of glioma recurrence.For example,studies have found that the peritumoral edema（PTE）displayed by traditional T2WI is related to the high aggressiveness and recurrence rate of gliomas,but there are also studies that disagree.In recent years,artificial intelligence technology with radiomics as the core has made major breakthroughs.Its characteristics of intelligence,multiple parameters,and objective quantification have made it possible to solve the above problems.Based on this,this study explores the value of glioma and PTE MRI imaging in the evaluation of tumor recurrence after surgery,aiming to improve the prediction of tumor recurrence and provide an effective evaluation tool for early treatment.OBJECTIVETo explore the value of MRI radiomics of glioma and peritumoral edema（PTE）in the assessment of tumor recurrence.METHODSThe clinical and MRI data of patients with glioma confirmed by surgery and pathology from January 2013 to December 2020 in Qilu Hospital of Shandong University were retrospectively analyzed,and 120 patients were finally selected,including 67 males and 53 females,aged 6-78 years old with an average of 45.4 years old.Postoperative pathology confirmed that there were 4,49,28 and 39 of grade Ⅰ,Ⅱ,Ⅲ,and Ⅳ gliomas,respectively.Among the above-mentioned patients,55 cases were confirmed by pathology or imaging follow-up after surgery,15 cases were confirmed by pathology after operation,40 cases were confirmed by image follow-up and follow-up;and other 65 cases had no recurrence,which were defined as recurrence group and control group.All patients’ axial T2WI and contrast-enhanced T1WI in DICOM format were uploaded to the Huiyi Huiying Radiology Cloud Radcloud platform for processing.According to the preoperative T2WI and T1WI enhanced images,the tumor and PTE were delineated in three-dimensional region of interest（ROI）.The ROI of PTE was delineated useing T2WI images,and the ROI of the tumor was delineated useing contrast-agent enhanced T1WI.And the intratuorous necrotic part was also included.Analysis of tumor+PTE was performed using the two ROIs above.The above-mentioned patients were randomly divided into training group and validation group at a ratio of 8:2,and the area under the receiver operating characteristic（ROC）curve（AUC）and accuracy matrix were used to compare and evaluate the training results of different imaging omics models.RESULTSAs for PTE,the KNN classifier scored the highest in the accuracy matrix:the AUC value,sensitivity,and specificity of the training group were 0.910,0.84,and 0.88,respectively,while the verification group were 0.916,0.82,and 0.93,respectively.For tumors,the LR classifier scored the highest in the accuracy matrix:the AUC value,sensitivity and specificity of the training group were 0.777,0.69,and 0.67,respectively,and the verification group were 0.758,0.82,and 0.92,respectively.When the tumor+PTE were combined,the LR classifier scored the highest in the accuracy matrix,the AUC value,sensitivity,and specificity of the training group were 0.977,0.88,and 0.89;meanwhile the verification group was 0.841,0.73,and 0.83 respectively.CONCLUSIONGlioma edema and tumor imaging omics characteristics have certain value in predicting postoperative recurrence of glioma,and the KNN omics model of PTE has the best performance.