Abstract
Meningiomas are the most prevalent benign intracranial tumors, and surgical intervention is the primary treatment. The physical characteristics of meningiomas, such as tumor stiffness or consistency, play a crucial role in the surgical approach. This study introduces a machine learning-based MR Elastography (MRE) inversion method, employing an artificial neural network trained with model-based synthetic displacement fields data to estimate mechanical properties of meningiomas. This framework enables accurate stiffness estimation by reducing partial volume effects near the tumor boundary and eliminating the assumption of material homogeneity in simulations (R(2)=0.93), which is often inaccurate for heterogeneous meningiomas. The study investigates the association between post-operative extent of resection (EOR) and pre-operative MRE-based tumor consistency. A significant correlation (p=0.024) was observed within a subset of patients with skull-based meningiomas in a cohort of 52 patients.