The Application Of MRI And Its Related Radiomics In The Non-invasive Evaluation Of Tumor Grades And Molecular Features Of Cerebral Gliomas

Part I.Amide Proton Transfer Imaging Allows Detection of Glioma Grades and Tumor Proliferation: Comparison with Ki-67 Expression and Proton MR Spectroscopy ImagingBackground and purpose: Prognosis in glioma depends strongly on tumor grade and proliferation.In this prospective study of patients with untreated primary cerebral gliomas,we investigated whether amide proton transfer-weighted imaging could reveal tumor proliferation and reliably distinguish low-grade from high-grade gliomas compared with Ki-67 expression and proton MR spectroscopy imaging.Patients and methods: This perspective study included 42 patients with low-grade(n = 28)or high-grade(n = 14)glioma,all of whom underwent conventional MR imaging,proton MR spectroscopy imaging,and amide proton transfer-weighted imaging on the same 3T scanner within 2 weeks before surgery.We assessed metabolites of choline and N-acetylaspartate from proton MR spectroscopy imaging and the asymmetric magnetization transfer ratio at 3.5 ppm from amide proton transfer-weighted imaging and compared them with histopathologic grade and immunohistochemical expression of Ki-67 in the resected specimens.Results: The asymmetric magnetization transfer ratio at 3.5 ppm values measured by different readers showed good concordance and were significantly higher in high-grade gliomas than in low-grade gliomas(3.61% ± 0.155 versus 2.64% ± 0.185,P =.0016),with sensitivity and specificity values of 92.9% and 71.4%,respectively,at a cutoff value of 2.93%.The asymmetric magnetization transfer ratio at 3.5 ppm values correlated with tumor grade(r = 0.506,P =.0006)and Ki-67 labeling index(r = 0.502,P =.002).For all patients,the asymmetric magnetization transfer ratio at 3.5 ppm correlated positively with choline(r = 0.43,P =.009)and choline/N-acetylaspartate ratio(r = 0.42,P =.01)and negatively with N-acetylaspartate(r =-0.455,P =.005).These correlations held for patients with low-grade gliomas versus those with high-grade gliomas,but the correlation coefficients were higher in high-grade gliomas(choline: r = 0.547,P =.053;N-acetylaspartate: r =-0.644,P =.017;choline/N-acetylaspartate: r = 0.583,P =.036).Conclsion: The asymmetric magnetization transfer ratio at 3.5 ppm may serve as a potential biomarker not only for assessing proliferation,but also for predicting histopathologic grades in gliomas.Part II.Radiomics based on multicontrast MRI can precisely differentiate glioma subtypes and predict tumour-proliferative behaviourPurpose: To explore the feasibility and diagnostic performance of radiomics based on anatomical,diffusion and perfusion MRI in differentiating among glioma subtypes and predicting tumour proliferation.Methods: Two hundred twenty pathology-confirmed gliomas and ten contrasts were included in the retrospective analysis.After being registered to T2 FLAIR images and resampling to 1 mm3 isotropically,431 radiomics features were extracted from each contrast map within a semi-automatic defined tumour volume.For single-contrast and the combination of all contrasts,correlations between the radiomics features and pathological biomarkers were revealed by partial correlation analysis,and multivariate models were built to identify the best predictive models with adjusted 0.632+ bootstrap AUC.Results: In univariate analysis,both non-wavelet and wavelet radiomics features were correlated significantly with tumour grade and the Ki-67 labelling index.The max R was 0.557(p=2.04E-14)in T1 C for tumour grade and 0.395(p=2.33E-07)in ADC for Ki-67.In the multivariate analysis,the combination of all-contrast radiomics features had the highest AUCs in both differentiating among glioma subtypes and predicting proliferation compared with those in single-contrast images.For low-/high-grade gliomas,the best AUC was 0.911.In differentiating among glioma subtypes,the best AUC was 0.896 for grades II-III,0.997 for grades II-IV,and 0.881 for grades III-IV.In predicting proliferation levels,multicontrast features led to an AUC of 0.936.Conclusion: Multicontrast radiomics supplies complementary information on both geometric characters and molecular biological traits,which correlated significantly with tumour grade and proliferation.Combining all-contrast radiomics models might precisely predict glioma biological behaviour,which may be attributed to presurgical personal diagnosis.Part III.The application of radiomics based on T1 C and T2 WI in the prediction of tumor grades and 1p/19 q co-deletion status in lower grade gliomasBackgroud and purpose: To explore the feasibility of T1 contrast-enhanced MRI and T2-weighted MRI imaging in predicting 1p / 19 q co-deletion status and histopathology grade of lower grade glioma.Moreover,comparison of the correlation between the two commonly used magnetic resonance imaging modalities and tumor pathology,molecular features was conducted,so did the predictive performances of obtained models based on their radiomics features.Materials and Methods: A total of 159 patients were enrolled in the study.All patients underwent T1 contrast-enhanced MRI(T1C)and T2-weighted MRI(T2WI)scanning.All data were obtained from the Cancer Imaging Archive(TCIA)and three manually drawn regions of interest were obtained to calculate radiomics features.A total of 431 radiomics features,including first-order statistical features(histogram features),features based on shape and size,texture features,and wavelet features of bandwidth 1-8.In univariate analysis,Pearson's correlation analysis was used to explore the correlation between imaging features and histology/genetic markers;the independent two sample T test was used to determine significant different features in different morphological and genetic status.In the multivariate analysis,the adjusted 0.632+ bootstrap method was used to reduce features,and the linear models combining different radiomics features were constructed by logistic regression to predict the different histological and molecular features of gliomas.Results: In uni-variate analysis,T1 C and T2 WI radiomics features were significantly correlated with tumor grade and 1p/19 q gene molecular types.In the grading,T1 C has more features than in T2 WI,and the maximum correlation coefficients are 0.27 and 0.18,respectively.However,there are no related features of the two modalities in correlation analysis and difference analysis that can be tested by Bonfernoni correction.In correlation analysis with 1p/19 q status,the maximum correlation coefficient of T2 WI is 0.50(P=1.69E-11),which is more significant than that of T1C(R=0.34,P=1.26E-05),and the number of positive features passed by Bonfernoni correction is 24 and 1.In the difference analysis,the number of features with significant differences between tumor 1p/19 q co-deletion is 137 and 47,respectively.And the number of features through multiple comparisons is 33 and 6,respectively.In the multivariate modeling analysis,T2WI showed better predictive performance than T1 C in distinguishing between different gliomas grade and 1p/19 q co-deletion mutations,with higher accuracy(AUCs in tumor grading: 0.88 vs 0.739;AUCs in the prediction of 1p/19 q co-deletion: 0.879 vs 0.75).Conclusion: Radiomics features based on T1 C and T2 WI is significantly correlated with the grades of lower-grade glioma and the status of 1p/19 q mutation.The prediction models based on radiomics features extracted from two modalities can effectively distinguish lower grade glioma types Compared with T1 C,T2WI showed obvious advantages in sub-type prediction of gliomas with higher predictive performances.