Multi-parametric MR Imaging Based Radiomics And Histopathology Of Peritumoral Heterogeneity In Glioblastoma

Chapter 1: Multi-parametric MR Imaging and Radiomics of Peritumoral Heterogeneity in GlioblastomaPurpose: To explore the feasibility of multi-parametric MR imaging and radiomics in predicting future tumor recurrence in the peritumoral edema regions for patients with glioblastoma(GBM)at baseline.Methods: Fifty-night patients with recurrent GBM who underwent multi-parametric MR imaging,including conventional MR imaging,T2-FLAIR,DWI,DSC-PWI,or MRS were enrolled.The area with future GBM recurrence were identified through coregistration of preoperative and post-recurrent MR images.Regions of interest were placed in the edema regions with and without future tumor recurrence to calculate r FLAIR,ADC value,r ADC,r CBV,Cho/NAA,Cho/Cr and NAA/Cr.The radiomic characteristics of CBV map were extracted.The two-sample t test,Mann-Whitney U test,ROC curve,and logistic regression analysis were used for statistical analysis.Results: Significant lower r FLAIR,ADC value,r ADC,and NAA/Cr but higher r CBV,Cho/NAA,and Cho/Cr were found in the edema regions with future tumor recurrence than in the edema regions without recurrence(all P< 0.05).The threshold values ?0.89 for r CBV provide the optimal performance for predicting the edema regions with future tumor recurrence,with a sensitivity,specificity,and accuracy of 84.7%,83.6%,and 85.8%,respectively.The combination of r FALIR,ADC,and r CBV can provide better predictive performance than r CBV(P = 0.015).The radiomics analysis of CBV map showed that the AUC of verification set and test set were 0.930 and 0.980,respectively.The sensitivity and specificity of prediction model in test set were 100% and 93.3%,respectively.Conclusion: T2-FLAIR,DWI,DSC-PWI and MRS are useful in the prediction of the edema regions with future tumor recurrence for patients with GBM.The combined use of r FLAIR,ADC,and r CBV can effectively estimate the risk of future tumor recurrence.The radiomic characteristics of CBV map are an effective method to predict the risk of future tumor recurrence.Chapter 2: Histopathological Basis of Multi-parametric MR Imaging of Peritumoral Heterogeneity in GlioblastomaPurpose: To explore the histopathological differences in different sub-regions of peritumoral edema regions based on multi-parametric MR imaging and the relationship between multi-parametric MR imaging and histopathology.Methods: Thirty patients with GBM who underwent multi-parametric MR imaging including conventional MR,DWI and DSC-PWI were enrolled in this study.The high-risk and non-high-risk area in the peritumoral edema regions were segmented based on the prediction model of the first part and the r FLAIR,ADC value and r CBV in different sub-regions of the peritumor were calculated,respectively.The peritumoral edema regions were extended resected to obtain tissue specimens.HE staining and IHC were used to analyze the tumor cell density,ki-67,CD34 and CD105 in different sub-regions.The two-sample t test,Mann-Whitney U test and linear regression were used for statistical analysis.Results: The r FLAIR and ADC values in the high-risk area were significantly lower than those in the non-high-risk area(P < 0.05),and the r CBV values were significantly higher than those in the non-high-risk are(P < 0.05).The tumor cell density,CD34 and CD105 expression levels in high-risk areas were significantly higher than those in non-high-risk areas(P < 0.05),and the ki-67 was not statistically significant(P > 0.05).In high-risk areas,there was a positive correlation between r CBV and tumor cell density,CD34 and CD105(P < 0.05),and the correlation coefficients were 0.591,0.592 and 0.697,respectively;ADC value was negatively correlated with tumor cell density,CD34 and CD105(P < 0.05),with correlation coefficients of-0.738,-0.503 and-0.500,respectively;r FLAIR was negatively correlated with tumor cell density and CD105(P < 0.05),with correlation coefficients of-0.409 and-0.417,respectively.In non-high-risk areas,r CBV was positively correlated with CD34 and CD105(P < 0.05),and the correlation coefficients were 0.544 and 0.619,respectively;ADC value and r FLAIR were negatively correlated with tumor cell density,with correlation coefficients of-0.470 and-0.337,respectively.Conclusion: The tumor cell density and microvascular proliferation in the high-risk areas with GBM were significantly higher than that in the non-high-risk areas.Multi-parametric MR imaging in peritumoral edema regions with GBM can accurately reflect the heterogeneity of tumor cell distribution and vascular proliferation.Chapter 3: Effect of Peritumoral Heterogeneity on the Prognosis of Patients with GlioblastomaPurpose: To explore the relationship between multi-parametric MR imaging and histopathology in peritumoral edema regions and the prognosis of patients with GBM in gross total resection group and extended resection group,and to compare the effect of different surgical methods on the prognosis of patients.Methods: Progression-free survival(PFS)in fifty-nine patients with GBM and overall survival(OS)in thirty-three patients with GBM after gross total resection were obtained.PFS in nineteen patients with GBM after extended resection were obtained.MR and histopathological parameters were measured in the peritumoral edema regions of the gross total resection group and extended resection group.Cox regression model was used to analyze the relationship between different parameters and the prognosis of patients with GBM.Kaplan-meier method was used to compare the effect of different surgical methods of gross total resection and extended resection on the prognosis of patients.Results: ADC value and r CBV in the peritumoral recurrence area of the gross total resection group were independent risk factors for patients' PFS and OS(P< 0.05).The HR values of ADC in predicting the patients' PFS and OS were 6.442 and 7.295,respectively.The HR values of r CBV in predicting the patients' PFS and OS were 2.827 and 9.230,respectively.Tumor cell density was an independent risk factor for PFS in the high-risk area of the extended resection group(P < 0.05),and the HR value of tumor cell density in predicting PFS was 1.566.The difference of PFS between the extended resection group and the gross total resection group was statistically significant(P< 0.05),and the median PFS of the two groups was 8.3 and 6.5 months,respectively.Conclusion: GBM with elevated r CBV or reduced ADC values in the peritumoral edema regions after gross total resection have shorter PFS and OS.After extended resection,the tumor cell density in the high-risk area increased,leading to shorter PFS.The recurrence time of GBM with gross total resection was shorter than that of patients with extended resection.