| Glioma is the most common primary tumor in the central system,accounting for about 45%of all intracranial tumors,and more than half of them are highly malignant.According to the latest statistics data of primary brain and other central nervous system tumors diagnosed in the United States from 2012 to 2016 from the Central Brain Tumor Registry of the United States(CBTRUS)in 2019,gliomas accounted for 25.5%of primary brain and other central nervous system tumors.In recent years,statistics have found that the incidence rate and mortality rate of glioma are increasing gradually,and now it has seriously endangered human health.The World Health Organization(WHO)published the histopathological classification of glioma in 2007,which divided glioma into four grades.Grade ? and ? glioma belong to low grade glioma,and grade ?and ? glioma belong to high grade glioma.The higher the histopathological grade of tumor,the greater the degree of clinical malignancy,the stronger the infiltration of surrounding tissue,the higher the recurrence rate of tumor after treatment,and the worse the prognosis of patients.In the course of clinical treatment,the main methods are surgical treatment,radiotherapy and chemotherapy.According to different histopathological grading of glioma,the choice of surgical treatment,the size of resection range and the length of radiation therapy time,the extent of the radiation range and the dose of chemotherapy drug and other treatments are also different,so it is of great clinical significance for clinicians to make reasonable treatment plan and evaluate the prognosis of patients to predict the histopathological grade of glioma before determining the treatment plan for patients.In the grading of gliomas,grade I gliomas are relatively rare in clinic.The imaging manifestations of grade ? gliomas are relatively typical because of their complex internal structure,which is easy for clinical and imaging physicians to diagnose them through images.The imaging manifestations of grade ? and grade ? gliomas are similar,which is difficult to differentiate.Therefore,the imaging differentiation of grade ? and grade ? gliomas is the difficulty and key points of medical imaging diagnosis.With the deepening of research,it has been found that the same histopathological type of glioma has different genetic characteristics,such as isocitrate dehydrogenase(IDH)mutation.In the 2016 edition of WHO central nervous system tumor classification,glioma genotypes were added on the basis of traditional histopathological,the most important of which is IDH genotyping,that is,glioma is divided into IDH mutant type glioma and wild-type glioma.Studies have found that IDH gene mutations mainly occur in low grade gliomas and secondary glioblastomas.At present,many scholars at home and abroad have shown that the IDH genotyping of low grade gliomas has a certain impact on the prognosis of patients with low grade gliomas.That is,the treatment prognosis of IDH mutant low grade gliomas is better than that of wild-type low grade gliomas,mainly because the mutation of IDH gene enhances the sensitivity of chemotherapy.In clinic practice,low grade gliomas are more common in grade II,the treatment of low grade glioma is given priority to with surgical resection,and the IDH mutant of low grade gliomas with chemotherapy,in order to reduce the recurrence of gliomas,and improve the effect of treatment.Therefore,preoperative non-invasive prediction of IDH genotyping of low grade glioma is of great clinical significance for the treatment of clinical patientsIn recent years,the organic integration of big data technology and medical image aided diagnosis has produced the research method of radiomics.Radiomics is the use of data descriptive algorithm,using automatic or semi-automatic analysis methods to extract massive image features from different images to quantify tumor and other major diseases.It can be used for quantitative evaluation of tumor heterogeneity,and then used for disease classification and diagnosis,guiding prognosis,and has important clinical value.With the development of magnetic resonance imaging(MRI),MRI has become the best noninvasive examination method for the diagnosis of nervous system diseases.Plain scan,diffusion weighted imaging and contrast enhanced scan are the most commonly used scanning methods in MRI.Magnetic resonance diffusion kurtosis imaging(DKI)is a new technology from magnetic resonance diffusion imaging,which is used to monitor the diffusion motion of non Gaussian water molecules in human tissues.It is based on diffusion weighted imaging(DWI)and diffusion tensor imaging(DTI).DKI can be more sensitive than DWI and DTI to reflect the complexity of tissue microstructure.DKI can use fractional anisotropy(FA),mean diffusion(MD)and mean kurtosis(MK)and other parameters can quantitatively describe the non Gaussian distribution of intracellular and extracellular water molecules diffusion and quantify the microstructure changes of pathological tissues,which can reflect the heterogeneity of glioma from the functional micro level.At present,there are still some deficiencies in the understanding of magnetic resonance imaging technology in glioma grading and genotyping.This study attempts to classify and identify the high and low grades of gliomas from the perspective of multimodal magnetic resonance radiomics,and identify the IDH subtypes of low grade gliomas through the image analysis of magnetic resonance DKI,so as to provides reference information and basis,and formulate individualized treatment for patients with gliomas by clinicians.To sum up,this study is divided into two parts.Part ?:Application of multimodal MRI imaging in glioma gradingBackground and purpose:According to the histopathological grading of glioma published by the World Health Organization in 2007,it is divided into four grades:Grade ? and grade ? glioma are classified as low grade glioma,grade ? and grade ? glioma are classified as high grade glioma.The treatment schemes of different grades of glioma are different,and the grading of glioma before treatment is particularly important.Grade ? gliomas are relatively rare in clinic.The imaging manifestations of grade ? gliomas are relatively typical because of their complex internal structure,so it is easy to diagnose in images.The imaging manifestations of grade ? and grade ? gliomas are relatively similar,so it is difficult to distinguish them.Therefore,the imaging differentiation of grade ? and grade ? gliomas has become the difficulty and focus of medical imaging diagnosis.MRI is the most commonly used examination method in clinic.The purpose of this study is to explore the value of multimodal MRI radiomics in the differential diagnosis of grade ? and grade? gliomas by analyzing the radiomics of multimodal MRI,extracting the radiomic features,establishing the statistical model of radiomics.Materials and methods:According to the inclusion and exclusion criteria,a total of 180 patients with gliomas,including 78 grade ? gliomas and 102 grade ? gliomas were selected.All patients underwent routine MRI scan within 2 weeks before operation.The examination equipment was Siemens Skyra 3.0T superconducting MRI scanner.The routine MRI scan sequences included T1 weighted image(T1WI),T2 weighted image(T2WI),T2 fluid attenuated inversion recovery(T2FLAIR),diffusion weighted imagingand Contrast Enhanced T1 Weighted Imaging(CET1WI).The DICOM format of the collected MRI image(T1 WI,T2WI,T2FLAIR,DWI,ADC,CET1WI)is imported into the radcloud of Huiyi Huiying,and then the region of interest of the imported image is outlined,and the region of interest is the tumor parenchyma area with the largest cross section.Huiyi Huiying platform was used to extract the radiomics characteristics of each MRI sequence region of interest,data analysis and radiomics statistical analysis were carried out at the same time.In the statistical analysis,the univariate analysis of variance,least absolute shrinkage and selection operator(LASSO)regression algorithm were used to reduce the dimension of features and gradually screen the icomic features.Then 80%of the data were randomly assigned to the training group for training,and 20%of the data were assigned to the validation group for data validation.For the training dataset,support vector machine(SVM)and logistic regression(LR)classifiers were used to construct the model based on multimodal MRI radiomics,and the validation set data was used to verify the model.In this study,the performance of the classifier is evaluated by four indicators:P(accuracy=true positive/(true positive+false positive)),R(recall=true positive/(true positive+false negative)),F1 score(F1 score=p*R*2/(P+R)),and support(total number of test sets).Receiver operating characteristic(ROC)curve was used to illustrate the predictive performance of radiomics.Result:1409 imaging features were extracted from each region of interest of each MRI sequence,and 8454 imaging features were extracted from 6 MRI sequences.After one-way ANOVA and lasso regression screening,the remaining seven features with the most diagnostic value.When SVM classifier is used for training,the area under curve of ROC curve of training set(area under curve,AUC was 0.962(95%CI:0.86-1.00),sensitivity was 0.94,specificity was 0.89 for grade ? glioma,sensitivity was 0.89,specificity was 0.94 for grade ?glioma,AUC was 0.76(95%CI:0.44-1.00),sensitivity was 0.40,specificity was 0.80 for grade ? glioma,sensitivity was 0.80,specificity was 0.40 for grade ? glioma.When LR classifier was used for training,AUC of training set was 0.955(95%Cl:0.85-1.00),sensitivity of grade ? glioma was 0.88,specificity was 0.94,sensitivity of grade ? glioma was 0.94,specificity was 0.88;AUC of validation data set was 0.68(95%CI:0.85-1.00)The sensitivity and specificity were 0.60 and 0.80 for grade ? gliomas and 0.80 and 0.60 for grade ? gliomas,respectively.When the SVM classifier is used for training,the accuracy,recall,F1 score and support of the training set are 0.88,0.94,0.91,16(grade ? glioma)and 0.94,0.89,0.91,18(grade ? glioma)respectively,while the accuracy,recall,F1 score and support of the verification set are 0.67,0.40,0.50,5(grade ?glioma)and 0.57,0.80,0.67,5(grade ? glioma)respectively.Using LR classifier training,the accuracy,recall,F1 score and support of training set were 0.93,0.88,0.90,16(grade ? glioma),0.89,0.94,0.92 and 18(grade ? glioma),respectively,while the accuracy,recall,F1 score and support of verification set were 0.75,0.60,0.67 and 5(grade ? glioma)and 0.67,0.80,0.73 and 5(grade? glioma).Conclusion:In this study,through the analysis process of multimodal MRI radiomics,the typical image signs were analyzed and integrated,and the multimodal radiomics of preoperative grading model in gliomas was constructed,which can provide reference for imaging physicians to use conventional MRI examination methods to diagnose grade ? and grade ? gliomas;at the same time,it can reduce the invasive examination of patients,and provide effective information for clinicians to help them make diagnosis and individualized treatment plan and prognosis evaluation.Part ?:The value of DKI in predicting the genotyping of low grade gliomasBackground and purpose:IDH1 gene mutation has a significant correlation with the prognosis of low grade gliomas,and the treatment regimens of low grade gliomas with different IDH1 genotypes are also different.The purpose of this study was to evaluate the value of DKI in predicting IDH1 genotyping in low grade gliomas.Materials and methods:The data of low grade gliomas were collected.All patients were scanned with 3T MRI before operation,and plane echo imaging(b=0,1000,2000,30 diffusion gradient directions)was used for DKI scanning.The tumor focus area,peripheral white matter(PWM)and contralateral normal appearing white matter(cNAWM)were selected as the regions of interest.The MD,MK and FA values of each region of interest were calculated by DKI imaging method.Independent sample t test was used to compare the differences of DKI parameters between IDH1 mutant and IDH1 wild-type groups,and paired sample t test was used to analyze the differences of DKI parameters between PWM and cNAWM.Result:According to the inclusion and exclusion criteria,38 patients with low grade gliomas were selected,including 10 patients with IDH1 wild-type(6 males,4 females,aged 32-67 years)and 28 patients with IDH1 mutant(16 males,12 females,aged 35-69 years).The MK value of wild-type glioma was significantly higher than that of mutant glioma,the MD value was significantly lower,and the FA value was not significantly different between them.ROC curve analysis showed that the predictive value of MK value was higher than that of MD value for IDH1 genotyping,and their AUC were 0.88 and 0.86,respectively.Paired t-test showed that there were significant differences in MD and MK values between pWM and cNAWM in IDH1 wild-type and IDH1 mutant gliomas(P<0.001).Interestingly,there was a significant difference in FA values between PWM and cNAWM in wild-type gliomas,but no difference in mutant gliomas.Conclusion:Compared with IDH1 mutant glioma,the MK value of IDH1 wild-type glioma was higher and the MD value was lower;in IDH1 wild-type and mutant glioma,the MD value and MK value of PWM and cNAWM were significantly different,the FA value of IDH1 wild-type glioma was lower than that of cNAWM,and there was no difference in IDH1 mutant glioma.MK value of tumor focus was more sensitive to IDH1 wild-type and mutant low grade gliomas than MD value.The change of FA value in PWM region was also helpful to distinguish the genotypes.This finding suggests that DKI parameters are helpful to distinguish IDH1 genotypes in low grade glioma. |
PDF Full Text Request