Prediction of brain invasion in patients with meningiomas using preoperative magnetic resonance imaging

// Alborz Adeli 1 , Katharina Hess 2 , Christian Mawrin 3 , Eileen Maria Susanne Streckert 4 , Walter Stummer 4 , Werner Paulus 2 , André Kemmling 1 , Markus Holling 4 , Walter Heindel 1 , Rene Schmidt 5 , Dorothee Cäcilia Spille 5, * , Peter B. Sporns 1, * and Benjamin Brokinkel 4, * 1 Institute of Clinical Radiology, University of Münster, Münster, North Rhine-Westphalia, Germany 2 Institute of Neuropathology, University Hospital Münster, Münster, North Rhine-Westphalia, Germany 3 Institute of Neuropathology, Otto-von-Guericke University, Magdeburg, Saxony-Anhalt, Germany 4 Department of Neurosurgery, University Hospital Münster, Münster, North Rhine-Westphalia, Germany 5 Institute of Biostatistics and Clinical Research, University of Münster, Münster, North Rhine-Westphalia, Germany * These authors have contributed equally to this work Correspondence to: Benjamin Brokinkel, email: benjamin.brokinkel@ukmuenster.de Keywords: brain invasion; grading; meningioma; magnetic resonance imaging; radiology Received: July 22, 2018     Accepted: October 25, 2018     Published: November 13, 2018 ABSTRACT Brain invasion (BI) in meningiomas impacts WHO grading and therefore adjuvant treatment. However, BI is rare and neurosurgical sampling and neuropathological analyses are not standardised. Moreover, associations with imaging findings are sparsely known. Associations between BI and findings on preoperative MRI were investigated in 617 meningioma patients. BI was strongly correlated with other high-grade criteria (p<.001). Presence of a contrast enhancing tumour capsule, disruption of the arachnoid layer, intratumoural calcifications and T2-intensity were not related to high-grade histology or BI (p>.05, each). High-grade histology (p=.033) but not BI (p=.354) was associated with tumour location. Irregular tumour shape (OR: 3.33, 95%CI 1.33-8.30; p=.007), heterogeneous contrast enhancement (OR: 2.82, 95%CI 1.19-6.70; p=.015) and peritumoural edema (OR: 1.005 per ccm, 95%CI 1.001-1.008); p=.011) were associated with BI. Multivariable analyses identified only increasing edema volume (OR: 1.005 per ccm, 95%CI 1.002-1.009; p=.010) as a predictor for BI, independent of other histopathological high-grade criteria. We finally provide a new model to estimate the risk of BI using routine preoperative MRI. Several imaging characteristics were identified as predictors for BI. Consideration in clinical routine can increase the accuracy of the detection in neuropathological analyses.