优化时间依赖性扩散MRI用于成人弥漫性胶质瘤术前分子分型

BACKGROUND AND PURPOSE: Isocitrate dehydrogenase (IDH) mutation and 1p/19q codeletion guide preoperative management in adult diffuse gliomas but are usually unknown before tissue sampling, and conventional ADC-based surrogates lack specificity. We evaluated whether an optimized time-dependent diffusion MRI (TDD-MRI) protocol integrating oscillating-gradient spin-echo (OGSE) and pulsed-gradient spin-echo (PGSE) acquisitions with Bayesian IMPULSED mapping can predict these markers preoperatively and validated imaging indices against digital histopathology.

MATERIALS AND METHODS: In this study (February 2023-December 2024), adults with suspected intracranial tumors underwent preoperative 3.0-T MRI with high-performance gradients. TDD-MRI included OGSE (20 and 40-Hz) and PGSE acquisitions. Bayesian IMPULSED generated voxelwise extracellular diffusivity (Dex), intracellular volume fraction (fin ), cell diameter (Diameter), and Cellularity; multi-time ADC and ADCratio were also computed. Diagnostic performance was evaluated using logistic regression and receiver-operating-characteristic analysis. Imaging-pathology associations were evaluated using digital hematoxylin-and-eosin metrics.

RESULTS: Among 119 adult diffuse gliomas (mean age, 51.9±9.1 years), tumors were classified as IDH-wildtype glioblastoma (IDHwt; n=53), IDH-mutant oligodendroglioma with 1p/19q codeletion (IDHmut-codel; n=33), or IDH-mutant astrocytoma with intact 1p/19q (IDHmut-intact; n=33). The intra- and inter-observer reproducibility of ROI measurements by the radiologists was excellent (intraclass correlation coefficients, 0.89-0.97). All microstructural parameters showed significant group differences on overall analysis (P<0.05). In Bonferroni-adjusted post hoc comparisons, IDHwt exhibited higher ADCratio, fin , and Cellularity and lower Dex than other groups, whereas IDHmut-codel had the lowest Cellularity. Among single parameters, ADCratio and Cellularity performed best. Multivariable models achieved AUC 0.95 (95% CI: 0.91-0.99) for identifying IDHmut-codel and 0.95 (95% CI: 0.92-0.99) for IDHwt, each outperforming single-time ADC (P<0.05); performance for IDHmut-intact was moderate (AUC 0.68; 95% CI: 0.59-0.78). Bayesian fitting yielded higher AUCs and more stable maps than nonlinear least squares. Imaging-derived fin , Diameter, and Cellularity correlated with pathology (r = 0.684-0.769, all P<0.001).

CONCLUSIONS: Optimized TDD-MRI with Bayesian IMPULSED provided quantitative microstructural mapping with strong histopathologic concordance and showed promise for the preoperative molecular subtyping of adult diffuse gliomas.