Ultrasound Localization Microscopy Reveals Microvascular Heterogeneity for Accurate Stratification of Glioblastoma Malignancy

Background: Intertumoral differences in glioblastoma (GBM) malignancy pose significant challenges for clinical management. Distinct microvascular growth patterns contribute substantially to tumor heterogeneity. Ultrasound localization microscopy (ULM) enables microscale mapping of microvascular network remodeling by tracking individual microbubble trajectories in vivo. This study evaluated whether ULM-derived microvascular heterogeneity metrics can facilitate histopathology-based stratification of GBM malignancy. Methods: An orthotopic glioblastoma model was established in 113 Sprague–Dawley rats, and ULM-derived heterogeneity parameters were extracted from tumor regions of interest. Spearman’s rank correlation coefficients were used to assess associations between microvascular heterogeneity metrics and histopathological indices. The Kruskal–Wallis H and Mann–Whitney U tests were used to compare metrics across different levels of microvascular maturity and cell proliferation. A decision tree–based diagnostic model was developed using ULM-derived microvascular features. Results: Microvascular heterogeneity was significantly negatively correlated with the vascular maturity index (p < 0.001) and positively correlated with the cell proliferation index (p < 0.001), supporting the biological and pathological relevance of ULM-derived heterogeneity metrics. Compared with transitional microvessels, mature microvessels exhibited significantly lower tortuosity (p = 0.002). Orientation variance, fractal dimension, connectivity, local thickness, and the spatial distribution index also tended to decrease but did not reach statistical significance (p = 0.074–0.529). In contrast, all corresponding metrics were significantly higher in immature microvessels (p ≤ 0.007). Compared with the low-proliferation group, all heterogeneity-related parameters were significantly higher in the high-proliferation group (p < 0.001). The decision-tree model based on microvascular heterogeneity demonstrated high performance at the sample level in predicting microvascular maturity and cell proliferation status, achieving accuracies of 90.29% (p = 0.029) and 92.23% (p = 0.026), respectively. Conclusions: We developed a clinically implementable decision-tree diagnostic model to support GBM malignancy stratification. As super-resolution ultrasound advances toward clinical translation, our findings may help inform future clinical decision-making.