An Approach to Survival Analysis in Glioma using Intra-Regional Ratiometric MR Image Features

In this study, brain tumor sub-regions are analyzed using ratiometric features from Magnetic Resonance (MR) images for the survival analysis of glioma sub-types. Structural MR images of high-grade and low-grade glioma patients are considered from a public database. From these images, segmented necrotic core, edema and enhancing tumor sub-regions are obtained. Ratiometric volume and mean intensity features are extracted from the segmented glioma sub-regions. Statistical significance of these features in differentiating high-grade and low-grade glioma is analyzed. Further, the significant features are employed for the survival analysis using Cox-Regression model. Results indicate that the formulated volume and intensity-based ratiometric features are able to differentiate the glioma sub-types (p-value < 0.05). The ratio between enhancing tumor and necrotic core features is associated with the overall survival of glioma patients (Model coefficient estimate, B > 1). Higher volume of enhancing tumor and lower necrotic core volume results in decreased survival of glioma patients. These findings indicate that the formulated ratiometric features could be useful in analyzing the biological behavior of glioma growth and aid in the early clinical decision making.