Prediction value of preoperative findings on meningioma grading using artificial neural network

Abstract

ObjectivesMeningioma is the most common brain tumor in adults. Grade 1 meningiomas have excellent prognoses, but grades 2 and 3 usually have worse outcomes, higher recurrence rates, and higher mortality rates. Preoperative determination of tumor grade may be helpful in deciding the type of surgery and the rate of resection. Blood markers have been used to predict the rate of malignancy and prognosis of tumors in different regions, including the brain. The current study investigated the use of blood markers on predicting meningioma grade. Patients and methodsPatients with newly diagnosed meningiomas were retrospectively reviewed. Data on the patients’ demographics, tumor locations, blood markers, and tumor pathology grades was extracted. The relationship between preoperative findings and tumor grade was statistically analyzed, and using the same findings and an artificial neural network, the accuracy of tumor grade prediction was evaluated. ResultsThis study included 95 patients, 69 cases (72.4 %) of grade 1, 23 cases of grade 2 (24.4 %) and 3 cases of grade 3 (3.2 %) meningiomas. Monocyte and neutrophil counts as well as lymphocyte-to-monocyte ratio (LMR) were significantly different between low grade and high grade meningiomas, with higher monocyte and neutrophil counts and higher LMR associated with high grade meningiomas (p < 0.05). Evaluation of the data with an artificial neural network using RBF with 5 variables (age, monocyte count, LMR, platelet-to-lymphocyte ratio (PLR), and neutrophil count) indicated that tumor grade can be determined with 83 % accuracy using an artificial neural network. ConclusionA preoperative high monocyte count and high LMR are associated with high grade meningioma. An artificial neural network using preoperative data can acceptably be used to characterize meningioma tumor grades.