The Clinical Value Of MR Habitat Imaging In Predicting The Molecular Classification And Prognosis Of High-grade Glioma

Background and Objective:High-grade glioma(HGG)is the most common primary malignant tumor in the central nervous system in adults,which has the biological characteristics of rapid progression,high invasiveness and poor prognosis.The mutation status of isocitrate dehydrogenase(IDH)gene in glioma increases our attention on molecular integration diagnosis.The molecular characteristics represented by IDH are intricately correlated with malignant outcome and poor prognosis of HGG patients.In addition to specific molecular features affecting prognosis,significant tumor heterogeneity and unique microenvironmental differences within HGG also support the formation and development of tumors,and are also key factors leading to drug resistance and recurrence.Therefore,a full understanding of tumor molecular characteristics and microenvironmental heterogeneity is helpful for making treatment decisions and prolonging patients’ survival time.As a highly heterogeneous and highly vascularized tumor,significant differences in internal signals and significant uneven enhancement can be observed in conventional MR sequences of HGG.In recent years,the improvement and development of MR function imaging has optimized the accurate diagnosis of HGG,which can reflect the potential pathophysiological significance and realize the real-time monitoring of biological behavior changes within tumors.For example,dynamic susceptibility contrast imaging(DSC)is the most commonly used diagnostic tool for evaluating hemodynamic changes in tumors.Its perfusion parameter,relative cerebral blood volume(rCBV),can reflect the situation of vascular proliferation and play an important role in predicting IDH mutation status and evaluating prognosis.Diffusion weighted imaging(DWI)can reflect the density of tumor cells by observing the restriction of free diffusion movement of water,thereby indirectly reflecting the degree of tumor cell proliferation.However,single function imaging can not fully and intuitively reflect the complex evolution process of tumor.The selection of combined multi-parameter imaging can cooperatively display different signal regions in tumor,which is helpful to explore the potential biological significance.At present,artificial intelligence algorithm has been widely used in imaging medicine,enabling multi-parameter MR data to be integrated and clearly displayed in the same space,enabling further comprehensive analysis of tumor habitat and tumor characteristics.In this study,conventional MRI sequences,DSC and DWI function imaging were used as the original input images to construct tumor habitats with different pathophysiological meanings,aiming to explore the value of habitat imaging in predicting IDH mutation status and finding tumor habitats associated with poor prognosis,so as to provide valuable clinical information for individualized precise treatment decision-making.Materials and methodsPart I: The value of MR Habitat Imaging in evaluating molecular typing of high-grade glioma1.Research objects: Retrospective analysis was performed on the imaging data of patients with HGG who met the fifth edition of the WHO Classification of Tumors of the Central Nervous System criteria(WHO CNS 5)after surgery and pathology between January 1,2016 and December 31,2021.A total of 78 patients were included(IDH wild-type: mutant-type = 60:18).All patients received routine MRI examination(T1WI,T2 WI,FLAIR),T1 WI contrast enhanced(T1CE),DSC,and DWI sequence scan at least simultaneously within one week before surgery.2.Habitat construction: The tumor enhancement,edema and necrotic areas were first distinguished according to routine MRI examination and T1 CE sequence,so as to construct Traditional Habitat(TH).And Vascular Habitat(VH)was constructed by superimposed DSC sequences on the basis of TH.Cellular Density Habitat(DH)was constructed by superimposing DWI sequences on TH.Finally,Combined Habitat(CH)was constructed by superimposing DSC and DWI sequences simultaneously on the basis of TH.3.Data collection: Clinical and imaging characteristics of patients with HGG were retrospectively collected and analyzed,including age,gender,tumor location,tumor enhancement area volume,edema area volume,necrotic area volume,the ratio of the volume of the above areas to the total volume,rCBVmedian and r ADCmedian of different habitats.4.Statistical analyses: The Chi-square test or Fisher’s exact test was used to analyze the constituent ratio of variables.The Independent Samples t-test or Mann-Whitney U test was used to analyze the differences between IDH mutant and wild-type HGG in parameter values of each habitat,age and tumor volume,using the Benjamini-Hochberg Procedure for controlling the false discovery rate in multiple comparisons.Receiver operating characteristic(ROC)curve analysis was applied to evaluate the value of differences in IDH mutation status.Then the Z test was applied to verify the differential effectiveness of precise tumor habitats constructed by multi-parametric MRI and TH.Multiple logistic regression was used to construct the combined ROC curve.Kendall correlation was used to determine the correlation between rCBVmedian and IDH mutation state of HGG in each habitat.Part II: The value of MR Habitat Imaging in predicting prognosis of glioblastoma1.Research objects: Retrospective analysis was performed on the imaging data of patients with glioblastoma(GBM)who met the fifth edition of the WHO Classification of Tumors of the Central Nervous System criteria(WHO CNS 5)after surgery and pathology between January 1,2016 and December 31,2021.A total of 58 patients were included.All patients received routine MRI examination(T1WI,T2 WI,FLAIR),T1 WI contrast enhanced(T1CE),DSC,and DWI sequence scan at least simultaneously within one week before surgery.2.Habitat construction: The tumor enhancement,edema and necrotic areas were first distinguished according to routine MRI examination and T1 CE sequence,so as to construct Traditional Habitat(TH).And Vascular Habitat(VH)was constructed by superimposed DSC sequences on the basis of TH.Finally,Combined Habitat(CH)was constructed by superimposing DSC and DWI sequences simultaneously on the basis of TH.3.Data collection: Clinical and imaging characteristics of GBM patients were retrospectively collected and analyzed,including age,gender,the extent of resection,adjuvant treatment,MGMT promoter methylation status,ATRX molecular status,rCBVmedian value of habitat,occurrence of endpoint events and calculating overall survival time.4.Statistical analyses: According to whether continuous variables satisfied normal distribution,Pearson or Spearman correlation analysis was used to determine the correlation between rCBVmedian and OS of GBM patients in each habitat.GBM patients were then converted into dichotomous variables according to the rCBVmedian of each habitat,i.e.,they were divided into high rCBVmedian group(rCBVmedian-High)and low rCBVmedian group(rCBVmedian-low).Kaplan-Merier method was used to draw the survival curve,and the OS difference between the two groups was tested by Log-rank.Finally,univariate Cox regression analysis was used to evaluate the risk factors of OS in GBM patients undergoing maximum safety resection and postoperative concurrent chemoradiotherapy.Multivariate Cox survival analysis was conducted based on the results of univariate Cox survival analysis.Results:Part I: The value of MR Habitat Imaging in evaluating molecular typing of high-grade glioma1.The age of IDH wild-type HGG patients was older than that of the mutant-type(P< 0.05).The difference of lesion location between the two groups was statistically significant(P < 0.05).The wild-type was more common in frontal lobe(30.0%)and parietal lobe(30.0%),while the mutant-type was more common in frontal lobe(66.7%).There were also statistical differences in total lesion volume,tumor enhancement area volume,ratio of component volume to total volume,rCBVmedian in Whole ET,VH1,VH3 and CH1-4 of HGG with different IDH mutation states(P < 0.05).2.ROC curve analysis showed that rCBVmedian at CH1 habitat had the highest AUC value of 0.815 for distinguishing IDH mutant and wild-type HGG.When cutoff value was4.81,the sensitivity and specificity of distinguishing IDH mutant and wild-type HGG were73.3% and 89.9%.The second habitat with a higher AUC value was VH1 habitat with an AUC value of 0.761.Logistic regression analysis was conducted to construct a joint index based on patient age,tumor site,total lesion volume,tumor enhancement area volume,ratio of component volume to total volume,and rCBVmedian value at Whole ET,VH1,VH3 and CH1-4.A higher AUC value of 0.965 was obtained,with a sensitivity of 95.0% and specificity of 88.9%.3.There was a positive correlation between rCBVmedian and IDH wild-type HGG patients in Whole ET,VH1,VH3 and CH1-4 habitat.In CH1 habitat,rCBVmedian and IDH mutation status were more correlated than in other habitats(r = 0.378,P < 0.05).Part II: The value of MR Habitat Imaging in predicting prognosis of glioblastoma1.rCBVmedian was negatively correlated with OS of GBM patients to varying degrees in habitat Whole ET,VH1,VH3,CH1 and CH2.In CH1 habitat,rCBVmedian was more correlated with OS of GBM patients than in other habitats,(r =-0.404,P = 0.002).2.In the Whole ET,VH1 and CH1 habitats,the OS of rCBVmedian-low and rCBVmedian-high groups was significantly different according to the conversion of rCBVmedian into binary variables(P < 0.05).In Whole ET habitat,the average OS of rCBVmedian-High group was about 279 days,which was significantly shorter than that of rCBVmedian-Low group(about 413 days),with a difference of about 4.47 months.In VH1 habitat,the average OS in the rCBVmedian-High group was about 257 days,which was significantly shorter than that in the rCBVmedian-Low group(440 days),with a difference of about 6.10 months.In the CH1 habitat,the mean OS of the rCBVmedian-High group was239 days,which was significantly shorter than that of the rCBVmedian-Low group(445days),with a difference of about 6.87 months.3.The analysis of GBM patients with maximum safe range of tumor resection and postoperative concurrent chemoradiotherapy showed that the rCBVmedian value of Whole ET,VH1,VH3,CH1,CH2 and CH3 habitats were risk factors for OS of GBM patients.According to the results of univariate Cox regression analysis,three multivariate Cox risk proportional models were constructed,and the rCBVmedian extracted from VH1,CH1 and VH1 habitats were independent risk factors in Model 1,Model 2 and Model 3,respectively.Conclusions:1.The rCBVmedian value obtained from CH1,the tumor subregion most likely to represent tumor cells and neovascularization proliferate actively,and VH1,the tumor subregion most likely to represent the high proliferation of neovascularization,had high diagnostic efficacy for distinguishing the IDH mutation status of HGG.It can be used as an imaging marker to predict the IDH mutation status of HGG before surgery,among which the former has higher sensitivity and specificity.On this basis,combined with the clinical information and imaging characteristics of patients,the diagnostic efficacy of IDH mutation status can be further improved.2.GBM perfusion heterogeneity and perfusion parameter values contained relevant information about patient survival.rCBVmedian at VH1 and CH1 habitats was an independent risk factor for poor prognosis in GBM patients.In addition,it was observed in Whole ET,VH1 and CH1 that patients with lower preoperative perfusion parameters had longer survival time,among which the difference in overall survival between the rCBVmedian-High group and the rCBVmedian-Low group was the longest according to the rCBVmedian at CH1 habitat.