Noninvasive Profiling of the Glioma Vascular Microenvironment via 7T MRI : Decoding Angiogenic Signatures for Isocitrate Dehydrogenase and World Health Organization Grade Differentiation

Vollständiger Abstract

Worum geht es in dieser Arbeit?

ABSTRACT Background Accurate preoperative glioma grading and molecular subtyping are important for treatment. The vascular microenvironment promotes tumor progression. A noninvasive screening tool capable of mapping tumor vascularity may assist in preoperative grading and subtyping of gliomas. Purpose To evaluate 7T susceptibility‐weighted imaging (SWI) for differentiating glioma isocitrate dehydrogenase (IDH) status and World Health Organization (WHO) grade based on vascular microenvironment features. Study Type Retrospective. Population Among 218 patients with histologically confirmed gliomas, 152 (70%) were assigned to training and 66 (30%) to validation by stratified random sampling. The training/validation sets included 61/26 IDH‐mutant, 91/40 IDH‐wildtype, 57/25 Grade 1–2, and 95/41 Grade 3–4 gliomas. Field Strength/Sequence 7T; T1‐weighted Magnetization prepared 2 rapid acquisition gradient echo (MP2RAGE) and SWI. Assessment Two independent neuroradiologists delineated tumors on 7T SWI. Vascular topology, density, and intensity were extracted following Frangi filtering and 3D skeletonization. Following feature selection, logistic regression (LR), support vector machine (SVM), naive Bayes (NB), and random forest (RF) were developed to differentiate IDH mutant from IDH wildtype, and WHO Grade 1–2 from WHO Grade 3–4. Statistical Tests Continuous and categorical variables were compared using Student's t ‐test/Mann–Whitney U test and chi‐square test, respectively. Cohen's kappa statistic and intraclass correlation coefficient (ICC) were used to assess the interobserver agreement. Features were selected by t ‐test (or Mann–Whitney U test) and Spearman's rank correlation. Sensitivity, specificity, accuracy, F1‐score, and area under the curve (AUC) were used to evaluate model performance. A p value < 0.05 was considered significant. Results Nine robust vascular features were retained. Imaging models achieved AUCs of 0.816–0.843 for IDH status and 0.834–0.874 for WHO grade in the internal validation set. Integrating clinical factors improved performance, with optimal AUCs of 0.888 and 0.889, respectively. Data Conclusion Integrating 7T SWI vascular features with machine learning may aid noninvasive differentiation of glioma molecular status and grade. Evidence Level 4. Technical Efficacy Stage 2.