Preoperative identification of intracranial meningiomas with aggressive behaviour may help for better risk stratification and choosing the optimal treatment strategy. In this study, we developed a structured semantic scoring system to evaluate potential aggressive tumour features. We compared its predictive value to conventional radiological analyses, to determine WHO grade and early tumour relapse in intracranial meningiomas.

A single-centre retrospective analysis of intracranial meningiomas treated between 2007-2018 was conducted. Inclusion criteria were: surgical removal and sufficient imaging data. Recurrences within 5 years after Simpson Grade I-III resection were considered as early. Preoperative T1 CE MRI sequences were analysed conventionally by two board-certified neuroradiologists. While being blinded for other clinical data they were asked to give a guess on the WHO grade and tumor recurrence within 5 years for each MRI. Additionally we developed a structured semantic scoring system with 11 items (1 point per item): Intratumoral heterogeneity, Multifocality, Midline shift, Sinus invasion, Necrosis/Hemorrhage, Mass effect, Cystic component, Bone invasion, Hyperostosis, Spiculation and Edema. The score was evaluated for each patient by 2 neurosurgeons. The predictive values for determination of WHO grade and early recurrence was analysed with a ROC analysis and the respective AUCs.

In total, 262 patients matched the inclusion criteria. There were 21% WHO grade 2 tumours and 40 (15%) tumour recurrences, Grade 3 tumors were omitted due to low numbers. Neuroradiologists classified 64.1% / 59.5% of meningiomas as WHO grade 1 and 26.7% / 31.3% as WHO grade 2 (ICC 0.93). The AUC for determination of WHO grade by radiologists compared to histological WHO grade was 0.66 and for determination of recurrence 0.54.

The semantic score showed statistically significant association between the number of items and tumour recurrence within 5 years (logistic regression, OR 1.28 (95% CI 1.003-1.62) Wald 3.95). Prediction of WHO grade 2 using the semantic score resulted in an AUC of 0.76 (optimal cut-off ≥ 3 points). The AUC for prediction of tumour recurrence was 0.62.

Structured image analysis e.g. by the proposed score could further improve prediction of tumour aggressiveness and WHO grade and might be a useful tool for risk stratification.