Assessing The Prognosis Of High-grade Gliomas Based On Traditional And Radiomic Features Of Magnetic Resonance Imaging

Part 1 Assessing the prognosis of high-grade gliomas based on traditional features of magnetic resonance imagingObjectiveThe use of magnetic resonance imaging routine sequences summarizes and evaluates the predictive value of traditional imaging features for progression free survival of high-grade glioma patients.Materials and Methods1.Research Objects:This study was approved by the Ethical Committee of the Nanfang Hospital,Southern Medical University(approval number:NFEC-2017-154),the study subjects were from January 2009 to December 2016 in the Department of Neurosurgery,Nanfang Hospital of Southern Medical University.After surgery,the patients were diagnosed as high-grade glioma by pathology.A total of 203 patients were examined.2.Methods:Retrospective analysis of the MRI characteristics and clinical data of the above-mentioned high-grade glioma patients.Image characteristics include peritumoral edema,edema,necrosis,enhancement,cystic changes,satellite,multifocal and edema.At the midline,whether the tumor crosses the midline,the position of the midline shift,and the maximum diameter of the tumor;clinical data includes age,gender,Karnofsky score,and postoperative radiotherapy and chemotherapy.Analyze the impact of the above features on PFS.3.Statistical analysis:Using R-software(http://www.R-project.org),Kaplan-Meier survival analysis was used for single factor analysis.The effects of the above mentioned images and clinical features on PFS were analyzed one by one,and Log-rank test was used for survival rate comparison;P<0.05 was further integrated into multiple factors.For the analysis model,multivariate analysis was conducted using the Cox proportional hazards model and stepwise regression analysis was used to identify the characteristics of P<0.05 as independent influencing factors for PFS.The accuracy of the model was evaluated using the consistency index C-index.The following R packages were used:’survival’ package for Kaplan-Meier survival analysis;the ’rms’ package was used for the analysis of Cox proportional hazards;the ’Hmisc’ package was used to compare the C-index.Results1.Single factor analysis results:Postoperative radiotherapy,chemotherapy,tumor cystic changes,multifocal,enhancement,and necrosis on MR images had significant influence on patients’ PFS(P<0.05);2.Multivariate analysis results:In the Cox proportional hazards model constructed based on MR imaging characteristics,moderate necrosis(P=0.0173),severe necrosis(P=0.0022),and tumor cystic degeneration(P=0.0020)were considered to affect PFS of patients.Significant independent influence factors,the C-index of the constructed Cox comparative risk model was 0.6275;In the Cox proportional hazards model constructed with imaging combined with clinical data,severe tumor necrosis(P=0.0196),cystic degeneration(P=0.0018),multifocality(P=0.0310),and chemotherapy(P=0.0013)was a significant independent factor influencing PFS in patients.The C-index obtained from the multivariate Cox proportional hazards model analysis was 0.6546.Conclusion1.High-grade glioma patients undergoing radiotherapy,chemotherapy,and tumor accompanying cystic changes on MR images are favorable support factors for judging the prognosis of patients.On the other hand,the more tumor,necrosis,and enhancement of tumors on MR images are the unfavorable factors affecting the prognosis of high-grade glioma patients.Among them,postoperative chemotherapy,tumor associated with cystic degeneration,multifocal and severe necrosis are independent factors influencing the prognosis of patients.2.Based on the conventional sequence traditional imaging features of MR,there is a certain ability to predict the prognosis of high-grade glioma patients.After the clinical data of patients are combined,the predictive model is significantly improved.Part 2 Evaluation of the prognosis of high-grade gliomas based on the radiomics features of magnetic resonance imagingObjectiveTo investigate the feasibility of predicting the prognosis of patients with high-grade gliomas based on the radiomics features of magnetic resonance imaging.Materials and Methods1.Research Objects:Same as the first part.2.Methods:First,the subjects were randomly divided into two groups,namely the training set(n=153)and the validation set(n=50).The patients’ age,gender,KPS score,radiotherapy,chemotherapy,and examination were calculated between the training set and the validation set.Are statistical differences between machines and PFS and other general information.Then the research of image composition is carried out.The main flow is as follows:image acquisition,image segmentation,feature extraction,prediction model construction and model verification.3.Statistical Analysis:Statistical analysis was performed on the general data using Statistical Product and Service Solution 20.0.Chi-square test was used to compare the patient’s gender,KPS score,radiotherapy,and chemotherapy between the training set and the validation set.And check the machine for statistical differences,using an independent sample t-test to compare the age of the patients between the training set and the validation set and whether the PFS was statistically different.R software was used to establish the Cox proportional hazards model,and the C-index of the four ROIs was calculated.The first part of the R packages was used.P<0.05 was considered statistically significant.ResultThere was no statistically significant difference(P=0.0700-0.9999)between the training set and the validation set for patients’ age,gender,KPS score,radiotherapy treatment,chemotherapy,examination machine,and PFS results.By Pearson correlation analysis,we screened out the features of the four ROIs in the training set that were significantly associated with PFS(P<0.05):A total of 21 PFS-related features were identified in the tumor ROI of the CET1WI image(P=0.0017-0.0470).The C-index in the Cox proportional hazards model constructed on the basis of these 21 features was 0.7656;29 PFS-related features were screened in the ROI of the Flair sequence tumor region(P=0.0015-0.0473).The C-index of the Cox proportional hazards model was 0.7115;16 PFS-related features were screened out in the Flair sequence PBZ(P=0.0072-0.0464),and the corresponding Cox proportional hazards model had a C-index of 0.6302;the entire abnormal signal zone of the Flair sequence(PBZ+tumor),A total of 15 PFS-related features(P=0.0072-0.0422)were screened out in the regional ROI.Corresponding Cox proportional hazards model had a C-index of 0.6375.Validation set validation results show that the C-index in the Cox proportional hazards model based on CET1WI sequence tumor region,Flair sequence tumor region,Flair sequence tumor region,and Flair PBZ ROI features were 0.7705,0.7219,0.6940,and 0.6606,respectively.Conclusion:1.Quantitative excavation of MRI histology information in the tumor region and peritumoral edema region of the magnetic resonance image can obtain information related to the patient’s prognosis,among which more than 30%of the features have a significant correlation with the patient’s PFS.2.The Cox proportional hazards model based on magnetic resonance imaging has a certain ability to predict the prognosis of patients with gliomas.Among them,the C-index of the Cox regression model based on the CET1WI tumor region is the highest,and the model is stable.In addition,the Cox proportional hazards model based on magnetic resonance imaging has a higher significance than the Cox proportional hazard model based on traditional imaging features.