AUTHOR=Sun Chen , Fan Liyuan , Wang Wenqing , Wang Weiwei , Liu Lei , Duan Wenchao , Pei Dongling , Zhan Yunbo , Zhao Haibiao , Sun Tao , Liu Zhen , Hong Xuanke , Wang Xiangxiang , Guo Yu , Li Wencai , Cheng Jingliang , Li Zhicheng , Liu Xianzhi , Zhang Zhenyu , Yan Jing 

TITLE=Radiomics and Qualitative Features From Multiparametric MRI Predict Molecular Subtypes in Patients With Lower-Grade Glioma

JOURNAL=Frontiers in Oncology

VOLUME=Volume 11 - 2021

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2021.756828

DOI=10.3389/fonc.2021.756828

ISSN=2234-943X

ABSTRACT=Background：Isocitrate dehydrogenase (IDH) mutation and 1p19q codeletion status have been identified as significant markers for therapy and prognosis in lower-grade glioma (LGG). This study aims to construct a combined machine learning model for predicting the molecular subtypes of LGG, including: (1) IDH wild-type astrocytoma (IDHwt), (2) IDH mutant and 1p19q non-codeleted astrocytoma (IDHmut-noncodel), (3) IDH-mutant and 1p19q codeleted oligodendroglioma (IDHmut-codel), based on multiparametric magnetic resonance imaging (MRI) radiomics, qualitative features and clinical factors. 
Methods：A total of 335 patients with LGG (WHO grade II/III) were enrolled retrospectively. 5,929 radiomics features were extracted from multiparametric MRI. Selected robust, non-redundant, and relevant features were used to construct a random forest model based on a training cohort (n =254) and evaluated on a validation cohort (n = 81). Meanwhile, preoperative MRIs of all the patients were scored in accordance with the Visually Accessible Rembrandt Images (VASARI) annotations and T2-FLAIR (fluid attenuated inversion recovery) mismatch sign. By combining radiomics features, qualitative features (VASARI annotations and T2-FLAIR mismatch sign) and clinical factors, a combined prediction model for the molecular subtypes of LGG was built.
Results：The 18-feature radiomics model achieved area under the curves (AUCs) of 0.8127, 0.6592 and 0.7698 for IDHwt, IDHmut-noncodel, and IDHmut-codel, in the validation cohort, respectively. Incorporating qualitative features and clinical factors into the radiomics model resulted in improved AUCs of 0.8422, 0.8166 and 0.8195 for IDHwt, IDHmut-noncodel, and IDHmut-codel, while the accuracies were 81.48%, 76.54% and 75.30%, respectively. 
Conclusion：The combined machine learning algorithm can provide a method to non-invasively predict the molecular subtypes of LGG preoperatively with excellent predictive performance and robustness.