Neuro-Oncology Research Updates

In 34 pediatric patients (ages 0–17) with low-grade gliomas, glioneuronal tumors, and high-grade gliomas, immunohistochemistry showed CAIX expression in 58.8% of tumors with no significant enrichment in HGGs versus LGG/GNT. Tumors frequently expressed monocyte/macrophage markers (CD68, CD44, CD163, CCR2) but had low PD-1/PD-L1 expression; CAIX correlated positively with CD44 (p=0.01) and CD206 (p=0.009). Strong CD68 expression (>30%) was associated with better 2-year event-free survival (91.7% vs 53.8%; p=0.011), while CAIX and other immune markers were not significantly associated with EFS.

This PROSPERO-registered systematic review included 30 studies of AI, machine learning, and deep learning using MRI-derived and/or multimodal perioperative data in glioblastoma for prognosis, treatment stratification, and related tasks; survival prediction was the most common focus (20/30 studies). The most frequent approaches were radiomics with conventional machine learning (13/30), deep learning (8/30), and hybrid methods (4/30), but validation was mainly internal with external validation uncommon (5/30); the authors conclude AI shows promise yet is limited by heterogeneity, scarce external validation, and inconsistent methodological reporting.

A systematic review and meta-analysis of 83 studies (3,666 patients) estimated the pooled incidence of symptomatic radiation necrosis after intracranial brachytherapy at 5.67% (95% CI 3.84%–8.29%). Sensitivity analyses showed higher pooled incidence when RN definitions were clear (8.71%) and subgroup rates varied by tumor type and isotope (e.g., 8.44% for high‑grade gliomas, 2.07% for brain metastases), but subgroup differences were not statistically significant.