Application of MRI Radiomics Combined With Deep Learning Technology in Glioma Segmentation and Survival Prognosis

Abstract

 Accurate automatic segmentation and prognostic diagnosis based on three-dimensional magnetic resonance imaging
(MRI) are essential for the proper treatment of gliomas. We developed a rapid, automated pipeline to segment gliomas and can
accurately predict patient survival prognosis based on pretreatment MRI in this paper. T1-Gado MRI sequence images of gliomas
are utilized to automatically segment gliomas using deep convolutional neural networks in this study. Nine machine learning
models that combine radiomics features and clinical characteristics are leveraged to predict and compare the survival and
prognosis of glioma patients. The results of the experiments show that all nine well-known learning model classification
architectures can achieve accurate classification and reliable prediction results. The clinical decision curves show that, except for
k-nearest neighbor and decision tree, all models perform well at various threshold probabilities.