The Diagnostic Value of MRI-based-Radiomics before Surgery of Patients with Meningioma

Background: Meningioma is the most common primary brain tumor in adults, mostly seen in middle-aged women. Magnetic resonance imaging (MRI) is the primary diagnostic and follow-up tool. Recently, radiomics have been introduced to predict prognosis in patients with brain tumors, yet studies are limited. Aim: The aim of this study was to investigate the diagnostic value of radiomics based on MRI images before surgery in patients with meningioma. Methods: In this cross-sectional study, meningioma patients who were operated on between 2018 and 2024 were evaluated. MR imaging of the patients were observed and Radiomics features were extracted by an expert radiologist. Finally, the images were analyzed based on radiomic algorithms. Results: LASSO test was performed and among 851 features, 10 features with non-zero correlation were selected. The accuracy of SVM was 81.25%, Precision was 0.85 with validity of 0.814±0.052. According to findings from Logistic Regression, the accuracy of the test was 87.50%, the precision was 0.89, and the validity of the test was 0.943 ± 0.070. Random Forest algorithm was also performed on the samples and it was seen that its accuracy was 87.5%, Precision 0.89, and validity 0.889 ± 0.056. The area under the curve (AUC) for SVM, Logistic Regression, and Random Forest were 0.62, 0.75, and 0.75, respectively. Conclusion: The use of artificial intelligence based radiomic algorithms can help radiologists diagnose meningioma more accurately, but it cannot replace it.