Spectral Analysis Based On MRI Hemodynamic Habitat Imaging Predicts Molecular Characteristics And Prognosis In High-grade Glioma

BackgroundHigh grade glioma(HGG)is the most common primary malignant brain tumor in central nervous system tumors,including grade 3 and 4 gliomas according to the 2021 world health organization(WHO)classification of central nervous system.Despite recent improvements in treatment,the prognosis for HGG has not significantly improved,possibly due to significant heterogeneity among tumors and unique tumor microenvironments(TME),resulting in complex metabolic changes in tumors.Quantitative spectral analysis of tumor metabolites(choline,creatine,etc.)can objectively reflect differences in the biological characteristics within tumors.Two-dimensional 1H multivoxel chemical shift imaging(CSI)has potential advantages in analyzing metabolic heterogeneity in HGG due to its wide coverage and ability to obtain a multispectral matrix.However,traditional spectral analysis is insufficient to accurately quantify local metabolic differences in tumors due to limitations such as inadequate spatial resolution and low signal-to-noise ratio.Tumors are composed of multiple significantly different and competing cell subpopulations,forming a habitat similar to different species occupying a specific habitat in an ecosystem.Habitat imaging identifies functionally related tumor subregions(habitats)by clustering voxels with similar radiological characteristics,revealing subtle differences in captured tissue features inside the tumor.Combining hemodynamic-related habitat imaging can further enhance the ability to describe tissue characteristics by breaking through the limitations of traditional spectral analysis and reflecting spatial metabolic differences within the tumor.Hemodynamic tissue signatures(HTS)is a more mature vascular flow habitat segmentation strategy that incorporates tumor perfusion parameters based on MRI anatomical segmentation,dividing the tumor into four vascular habitats: high-angiogenic enhancing tumor habitats(HAT),low-angiogenic enhancing tumor habitats(LAT),infiltrated peripheral edema habitats(IPE),and vasogenic peripheral edema habitats(VPE).Quantifying tumor metabolism based on habitat imaging that retains spatial and blood flow information more accurately describes metabolic differences in HGG patients with different molecular subtypes than directly quantifying metabolite levels in enhancing or edematous areas,establishing a link between tumor metabolic levels and progression-free survival(PFS)and overall survival time(OS),predicting HGG molecular features and prognosis.ObjectivePoor prognosis in HGG is associated with abnormal tumor metabolism.This study evaluates spatial metabolic differences in the tumor through quantitative spectral analysis of vascular habitats,to predict HGG molecular characteristics and evaluate patient prognosis.Material and methods1.Clinical dataThe study included 110 HGG patients from January 2016 to December 2020,including56 males and 54 females with an average age of 50.54±13.19 years.There were 27IDH-mutant cases and 83 IDH-wildtype cases.PFS and OS information of HGG patients were collected through electronic medical records and telephone follow-up.In the PFS analysis,there were 55 cases with recurrence,34 cases without recurrence,and 21 cases lost to follow-up.In the OS analysis,there were 44 deaths,50 survivors,and 16 lost to follow-up.2.MRI scanningA German Siemens Verio 3.0 T superconducting magnetic resonance imaging system and an 8-channel head coil were used for routine T1-weighted imaging(T1WI),T2-weighted imaging(T2WI),T2-weighted fluid attenuation inversion recovery(Flair),T1-weighted contrast-enhanced imaging(T1CE),dynamic susceptibility contrast perfusion-weighted imaging(DSC-PWI),and two-dimensional 1H multivoxel CSI.3.Image processing and habitat constructionAll image data of included patients were exported in DICOM format from the PACS system,converted to NII format files,and analyzed by Oncohabitats for HTS.CSI voxels within each habitat were independently read by two radiologists,and the selected CSI voxels were fitted to the relative volume of the HTS habitat after approval,simulating the metabolic ratio of vascular habitats.4.Statistical analysisWeighted least squares(WLS)were used to simulate the metabolic ratios(Cho/Cr and Cho/NAA)of HTS habitats.Pearson correlation coefficients were used to validate the correlation between metabolic ratios in each habitat and perfusion indices independent of WLS fitting.The metabolic ratios of HGG patients were divided into IDH mutant and IDH wildtype groups,with the metabolic ratio of the enhanced area as the control group.Independent sample t-tests were used to evaluate whether there were differences in metabolic ratios among different habitat groups and the enhanced area group.ROC curve analysis was performed to investigate the predictive power of metabolic ratio in HAT,metabolic ratio in LAT,and metabolic ratio in enhanced area for IDH status,including determination of optimal thresholds and calculation of specificity and sensitivity.Z-tests were used to compare differences in AUC values.Uni-and multivariate Cox regression analyses were used to evaluate factors affecting the prognosis of HGG patients based on metabolic ratios in different habitats.R software was used to construct column charts for predicting the median PFS of HGG patients.According to the median metabolic ratio in each habitat,patients were classified into two groups for survival analysis using Kaplan-Meier curves,with the metabolic ratio of the enhanced area as the control group.Differences in OS between groups were evaluated using Log-rank tests.Results1.Construction and validation of metabolic ratios in vascular habitatsWLS showed a linear relationship between the relative volumes of HAT and LAT(the ratio of vascular habitat to tumor volume)and the corresponding metabolic ratios(p<0.05),indicating that the metabolic ratios of HAT and LAT can be fitted based on the relative volume of the habitat,thus completing the metabolic ratios of vascular habitats.Pearson correlation coefficient showed a positive correlation between the metabolic ratios of vascular habitats and perfusion indices independent of WLS fitting,with Cho/Cr in HAT as the best(r CBV,r=0.517),verifying the reliability of metabolic ratios in vascular habitats.2.Predictive power of metabolic ratios in vascular habitats for IDH statusIndependent sample t-tests showed significant differences in metabolic ratios between IDH mutant and wildtype groups based on HAT and LAT metabolic ratios.Taking HAT as an example,there were significant differences in both Cho/Cr(IDH mutant: Cho/Cr=2.29±0.24,IDH wildtype: Cho/Cr =2.58±0.33,t=4.099,p<0.05,95% CI=0.147-0.422)and Cho/NAA(IDH mutant: Cho/NAA=4.47±0.92,IDH wildtype: Cho/NAA=4.93±0.72,t=2.736,p<0.05).ROC curves showed that the optimal diagnostic thresholds of Cho/Cr and Cho/NAA for predicting IDH mutant were 2.56(AUC=0.749,Spec.=85.2%,Sens.=59.0%)and 3.92(AUC=0.693,Spec.=70.4%,Sens.=71.1%),respectively,in HAT.Z-tests showed that the predictive power of HAT metabolic ratios was significantly stronger than that of metabolic ratios in the enhancing area(Cho/Cr: p<0.05,Z=2.824,95% CI=0.067-0.371;Cho/NAA: p<0.05,Z=1.965,95% CI=0.000-0.341).3.Prognostic efficacy of metabolic ratios in vascular habitats for HGG patientsCox regression analysis showed that both Cho/Cr and Cho/NAA in HAT and LAT were risk factors affecting patient prognosis,and Cho/Cr in HAT was an independent risk factor.Nomograms for predicting the median PFS of HGG patients and Kaplan-Meier survival curves for evaluating OS of HGG patients were constructed.The nomograms of Cho/Cr and Cho/NAA showed that their predictive C-indexes were 0.747 and 0.769,respectively,through 1000 bootstrap validations,indicating good predictive efficacy.Kaplan-Meier survival curves showed that there was a significant difference in OS between the high and low metabolic groups classified by metabolic ratios in vascular habitats,with a time difference of 9 months(p<0.05).ConclusionSpectral analysis based on MRI hemodynamic habitat can describe metabolic differences in HGG patients with different molecular subtypes,breaking through the limitations of traditional spectral analysis and predicting the IDH status.Metabolic ratios in various vascular habitats are risk factors affecting patient prognosis and can be used to construct a highly accurate nomogram for predicting the median PFS of HGG patients.The effectiveness of evaluating patient’s OS is also significantly better than that of traditional spectral analysis.