AUTHOR=Zhang Yang , Chen Chaoyue , Cheng Yangfan , Teng Yuen , Guo Wen , Xu Hui , Ou Xuejin , Wang Jian , Li Hui , Ma Xuelei , Xu Jianguo 

TITLE=Ability of Radiomics in Differentiation of Anaplastic Oligodendroglioma From Atypical Low-Grade Oligodendroglioma Using Machine-Learning Approach

JOURNAL=Frontiers in Oncology

VOLUME=Volume 9 - 2019

YEAR=2019

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2019.01371

DOI=10.3389/fonc.2019.01371

ISSN=2234-943X

ABSTRACT=Objectives To investigate the ability of radiomics from MRI in differentiating anaplastic oligodendroglioma (AO) from atypical low-grade oligodendroglioma with enhancement using machine learning algorisms.
Methods A total number of 101 qualified patients (50 participants with AO and 51 with atypical low-grade oligodendroglioma) were enrolled in this retrospective, single-center study. Forty radiomics features derived from six matrices were extracted from contrast-enhanced T1-weighted images. Three selection methods including Distance Correlation, least absolute shrinkage and selection operator (LASSO) and gradient boosting decision tree were performed to select the optimal features for classifiers. Then three machine learning classifiers including linear discriminant analysis, support vector machine and random forest (RF) were adapted to generate discriminative models. Receiver operating characteristic analysis was conducted to evaluate the discriminative performance of each model.
Results All the classifiers represented feasible ability in differentiation with AUC more than 0.850 when combined with suitable selection method. Among the nine models, the combination of LASSO and RF classifier showed the best discriminative performance with AUC of 0.904. 
Conclusion Radiomics with machine learning approach could be potentially served as a feasible method in distinguishing AO from atypical low-grade oligodendroglioma. Moreover, the combination of LASSO and RF classifier was proven to be the optimal model in discrimination.