Susceptibility-Weighted MRI for Predicting NF-2 Mutation and S100 Protein Expression in Meningiomas

Abstract

Abstract Purpose To investigate non-invasive biomarkers of neurofibromatosis type 2 (NF-2) mutation and S100 protein expression in meningiomas using morphological and radiomics features of susceptibility-weighted MRI (SWI) with deep learning. Methods Ninety-nine meningioma patients, who were pre-operatively scanned at a 3T clinical MRI scanner, underwent genetic analysis for NF-2 mutation and immunohistochemistry for S100 protein expression. Tumor growth pattern, peritumoral edema, sinus invasion, hyperostosis, bone destruction, and intratumoral calcification were semi-quantitatively assessed. The differences of radiomics and deep learning features of SWI were compared between NF-2 mutation- and S100 protein expression-based subgroups using either Mann–Whitney U or χ² tests. Logistic regression and machine learning techniques explored the relationships between the molecular characteristics and the features obtained by radiomics and deep learning. Results NF-2 mutation was associated with a higher risk of developing high-grade tumors (P = 0.01). Additionally, elevated signal intensity (P = 0.002) and a decrease in entropy (P = 0.049) within the tumoral region on SWI were observed in meningiomas with S100 protein expression. NF-2 mutation was associated with lower SWI signal intensity (P = 0.015), a growth pattern described as "en plaque" (P = 0.023), and the presence of calcification within the tumor (P = 0.021). Logistic regression models achieved accuracies of 0.74 for predicting NF-2 mutation and 0.80 for identifying S100 protein expression. Deep learning features demonstrated strong predictive capability for S100 protein expression (AUC = 0.85) and had reasonable success in identifying NF-2 mutations (AUC = 0.74). Conclusion SWI shows promise in identifying NF-2 mutation and S100 protein expression by revealing neovascularization and microcalcification characteristics in meningiomas.