Comparative Evaluation of the Possibilities of Radiomic Аnalysis of Magnetic Resonance Imaging in the Differential Diagnostics of Primary Extra-Axial Intracranial Tumors

Abstract

Purpose of the study. Comparing magnetic resonance imaging (MRI) abilities in differential diagnostic of three types of primary extra‑ axial brain tumors (benign and malignant meningiomas, and neuromas) based on standard semiotics and radiomic features.Patients and methods. Retrospective research included 66 patients with primary extra‑a xial tumors who were divided into two groups: the instructional (39 patients) and the valid (27 patients). MRI was used towards all patients before surgery. The one method of statistical modeling – discriminant analysis – was used to compare the abilities of differential diagnostic based on semiotic features and radiomic parameters.Results. The features of tumor semiotics MRI didn’t allow to differentiate effectively benign and malignant meningiomas. Several parameters were certainly varied for all those tumor types (neuromas, benign and malignant meningiomas). The modelling based on the discriminant analysis demonstrated that radiomic features can be used for primary extra‑a xial tumors differential diagnostic. The area of the radiomic model ROC‑curve took 0.86 which exceeds the result of the model based on semiotic features (AUC 0.78).Conclusion. The best results of the tumors classification by radiomic model demonstrate expediency to continue research the primary extra‑ axial tumors differential diagnostic with support of histogram and textural parameters of MRI imaging.