Abstract

Info

Title:: Prediction of glioma-associated epilepsy based on radiomics and multi-sequence MRI

Author(s):: YANG Hongxi1; GAO Ankang2; WANG Yida1; BAI Jie2; ZHANG Yong2; CHENG Jingliang2; YANG Guang1; 1. Shanghai Key Laboratory of Magnetic Resonance, School of Physics and Electronic Science, East China Normal University, Shanghai 200062, China 2. Department of MRI, the First Affiliated Hospital of Zhengzhou University, Zhengzhou 450000, China

Keywords:: Keywords: glioma-associated epilepsy magnetic resonance imaging radiomics

Abstract:: Abstract: A non-invasive, accurate and objective model is developed based on multi-sequence MRI for diagnosing glioma-associated epilepsy. The T1WI, T2WI, T1Gd and T2-FLAIR images of 403 glioma patients are collected retrospectively. A pre-trained deep learning model is used to segment the region of interest containing the tumor and peritumoral edema, from which the first-order statistical characteristics, morphological features and texture features are extracted. After feature selection using Pearson correlation coefficient, recursive feature elimination and other methods, a scout model is built for each group of features, and finally a radiomics model containing 15 features is established for epilepsy diagnosis. The radiomics model achieves an AUC of 0.836 on the independent test set. A clinical-radiomics model containing 14 features is further built by incorporating basic clinical information (age and gender) to the radiomics model for remodeling, and it achieves an AUC of 0.872 on the independent test set. The model combining basic clinical information and multi-sequence MRI radiomics signatures can serve as an effective tool for the diagnosis of glioma-associated epilepsy.