Diffusion–Relaxation Correlation Spectroscopic MRI in Adult Gliomas: Spectral Heterogeneity and Associations With Histomolecular Features

Background Preoperative MRI characterization of glioma histomolecular features is limited by intratumoral heterogeneity and the scalar nature of conventional diffusion and relaxation metrics. Purpose To characterize glioma heterogeneity using diffusion–relaxation correlation spectroscopic imaging (DR‐CSI) and evaluate associations with glioma grade, isocitrate dehydrogenase (IDH) status, and Ki‐67 index. Study Type Prospective. Population Fifty Three treatment‐naïve adults with histopathologically confirmed gliomas (mean age, 49.0 ± 13.9 years; 34 men). Field Strength/Sequence 3 T; multi‐b, multi‐TE diffusion–relaxation spin‐echo echo‐planar imaging sequence. Assessment Voxelwise D–T2 spectra were reconstructed from multi‐ b ‐value, multi‐echo‐time data and clustered into seven spectral subregions (V1–V7). Spectral fractions and conventional metrics were normalized to normal‐appearing white matter (NAWM): nV1–nV7, nADC 1 , and nT2 1 were obtained from the entire non‐necrotic tumor, whereas nADC 2 and nT2 2 were obtained from enhancing tumor (ET) in substantially enhancing lesions or non‐enhancing tumor (NET) in lesions with little or no enhancement. Associations of these metrics with glioma grade (high or low), Ki‐67 index, and IDH status (wild type or mutant) were investigated. Statistical Tests Mann–Whitney U test, Spearman rank correlation coefficient ( ρ ), logistic regression, likelihood ratio testing, receiver operating characteristic (ROC) analysis, bootstrapped area under ROC curve (AUC). Significance: p < 0.05. Results DR‐CSI showed distinct D–T2 spectral patterns across ET, NET, necrosis/cystic change, and NAWM. nV3 significantly correlated with Ki‐67 index ( ρ = 0.401) and remained significantly associated with high‐grade status after age adjustment (odds ratio [OR] per standard deviation [SD] increase = 6.020). nV5 was significantly associated with IDH‐mutant status after age adjustment (OR per SD increase, 10.980). Including nV3 in addition to age significantly improved grade model fit (bootstrap‐corrected AUC, 0.805 vs. 0.709). Including nADC 2 and nV5 in addition to age significantly improved IDH model fit over age alone (bootstrap‐corrected AUC, 0.953 vs. 0.877). Data Conclusion DR‐CSI characterized D–T2 compartment spectral heterogeneity in gliomas, and DR‐CSI metrics were associated with grade, Ki‐67 index, and IDH status. These findings support DR‐CSI as a composition‐sensitive MRI approach for glioma heterogeneity assessment. Evidence Level 2. Technical Efficacy Stage 2.