评估CBV在识别胶质母细胞瘤早期复发风险中的价值

Impact: Quantifying the fraction of high-CBV regions within glioblastoma may provide an early, practical imaging-based estimate of risk of recurrence, addressing a key unmet challenge in anticipating tumor regrowth and guiding timely clinical intervention. Introduction: Glioblastoma (GBM) is highly aggressive, with recurrence common despite maximal standard therapy. Cerebral blood volume (CBV) from DSC-MRI quantifies tumor vascularity and helps distinguish true progression from pseudoprogression. Standardized relative CBV (sRCBV) thresholds validated by Anil et al., 2024, against image-localized histopathology differentiate recurrence (sRCBV >1) and indicate viable tumor presence (sRCBV >1.37, ~88% probability). This study evaluates whether the fraction of high-sRCBV voxels serves as an early indicator of recurrence, identifying thresholds linked to recurrence onset. Methods: Fifteen post-therapy MRI time points from six IDH1-wild-type GBM patients (five female, one male; age 49-81 years) were retrospectively analyzed under IRB approval (2012-0441, PI Dr. Puduvalli). All received Stupp-protocol chemoradiotherapy (VMAT 6000cGy/30 fractions) and temozolomide, followed by surveillance until recurrence. Only DSC-MRI scans acquired after chemoradiation were included. All studies used the consensus single-dose, low-flip-angle (30°) protocol without preload. Standardized RCBV maps were generated using IB Neuro with BSW leakage correction and registered to contrast-enhanced T1-weighted (T1CE) images. Tumor masks were defined semi-automatically on T1CE. The fraction of high-sRCBV voxels (fH) was calculated as voxels with sRCBV >1.37 divided by total tumor voxels. The interval between each scan and radiologic recurrence (Δt) was recorded. Associations between fH and Δt were tested using quadratic regression, ROC analysis for early (≤120 days) vs late recurrence, Youden-index thresholding, and Cox proportional-hazards modeling. Results: Across 15 time points, fH showed a borderline-significant inverse relationship with time to recurrence (R² =0.35, p = 0.073). ROC analysis yielded AUC = 0.74 (95% CI 0.44-0.96) with an optimal threshold of fH ≈ 0.39. Higher fH correlated with increased recurrence hazard (beta = 4.26, HR = 70.47 per +1.0 fH). Tumors with ~27% high-sRCBV voxels corresponded to a 120-day median recurrence-free survival. Conclusion: Preliminary findings indicate that the fraction of high-sRCBV voxels within the enhancing tumor may serve as a quantitative indicator of glioblastoma recurrence risk. Although limited by small sample size and retrospective design, consistent trends suggest that higher vascular fractions correspond to shorter recurrence intervals and increased risk of regrowth. Ongoing work is expanding this analysis to a larger, multi-timepoint cohort to improve statistical power and define clinical thresholds for integration into CBV visualization tools.