Fractal analysis improves the preoperative identification of atypical meningiomas
Marcin Czyż, Hesham Mohamed Abouelela Abdelmawla Radwan, Jian Li, Chrisopher Filippi, Michael Schulder · Neuro-Oncology · 2018
There is no objective and readily accessible method for the preoperative determination of atypical characteristics of a meningioma grade. The aim of our study was to evaluate the feasibility of using fractal analysis as an adjunctive tool to conventional radiological techniques in identifying meningiomas with higher grade of malignancy. A group of 54 patients was enrolled into the study. 27 patients diagnosed with atypical (WHO Grade II) meningioma and a second age and sex matched group of 27 patients with benign (WHO Grade I) meningioma were analysed. Preoperative brain MR (T1W, post-gadolinium) were processed and analysed to determine the average (FDa) and maximum (FDm) fractal dimension of the contrast-enhancing region of the tumor. Box-count method and ImageJ 1.49 software were used. FDa and FDm as well as volume of the peritumoral oedema, incomplete peritumoral band, male sex and skull base localisation were included into the logistic regression model as possible predictors of malignancy. The cohort consisted of 34 women and 20 men, mean age of 62 ± 15 years. Fractal analysis showed good inter-observer reproducibility (Kappa >0.70). Both FDa and FDm were significantly higher in the atypical compared to the benign meningioma group (P < .0001). Multivariate logistic regression model reached statistical significance with P = .0001 and AUC = 0.87. The FDm, which was greater than 1.31 (odds ratio [OR], 12.30; P = .039), and non-skull base localisation (OR, .052; P = .015) were confirmed to be statistically significant predictors of the atypical phenotype. Fractal analysis of preoperative MR images appears to be a feasible adjunctive diagnostic tool in identifying meningiomas with potentially aggressive clinical behaviour.