Voxel-based spatial distribution of intracranial meningioma subtypes and their relationship to radiogenomic maps

Meningiomas are histologically and genetically heterogeneous tumors with varying anatomical distributions. While distinct genetic mutations have been associated with specific tumor locations, the spatial distribution of histological subtypes and their relationship to radiogenomic profiles remains poorly defined. Moreover, the predictive value of spatial information for histopathological classification and tumor grading has not yet been systematically explored. This study aimed to systematically analyze the anatomical predilection of histological meningioma subtypes and their concordance with known mutation-specific spatial patterns, and the predictive potential of voxel-based spatial features. We retrospectively analyzed 737 patients undergoing surgical resection of intracranial meningiomas. Preoperative magnetic resonance images were normalized to a common stereotactic space, and tumors were semi-automatically segmented. Voxel-based lesion-symptom mapping (VLSM) was performed to identify subtype-specific spatial clustering. Spatial distributions were compared with mutation maps from current literature using receiver operating characteristic analysis (AUC-ROC). Additionally, multinomial logistic regression models were applied to evaluate whether tumor localization could predict histological subtype and World Health Organization (WHO) grading. Histological subtypes showed distinct spatial preferences. Meningothelial meningiomas clustered in the anterior and middle skull base; fibrous and transitional types predominated in the convexity, falx, and tentorium; secretory tumors localized to the sphenoid wing and petroclival region; and atypical meningiomas were common in the anterior falx and frontal convexity. Psammomatous meningiomas displayed a broader distribution with involvement of the petrous bone and foramen magnum. AUC analysis revealed strong concordance between histological subtypes and mutation maps, confirming known histogenomic associations (e.g. KLF4/TRAF7 with secretory; NF2 with fibrous and transitional; SMO with meningothelial). No associations to any mutation map were observed for angiomatous, microcystic and metaplastic meningiomas. Predictive modeling based solely on spatial features achieved moderate accuracy for subtype classification and higher accuracy for WHO grade prediction. Meningioma subtypes show distinct, statistically robust anatomical predilections that align with known genetic mutation maps. Predictive modeling highlights that spatial features themselves hold diagnostic and prognostic value, linking anatomical localization to tumor biology and aggressiveness. The study introduces anatomically precise voxel-based templates that may improve radiogenomic classification and non-invasive genotype prediction.