Abstract—Glioblastoma (GBM) is the brain’s most common malignant primary tumor. Survival prediction is crucial for risk stratification and impacts all aspects of care. However, previous attempts at predicting survival outcomes have relied on clinical parameters that suffer from uneven capture in available data sets, limiting the precision of transferable survival predictions. In this study, we propose a novel method for overall survival prediction in GBM patients that combines deep learning and hand-crafted radiomic features using brain Magnetic Resonance Imaging (MRI) images. The proposed method involves three main steps: 3D brain tumor segmentation using the nnU-Net model, patient classification into long, short, and mid-survivors using the Dense-Net model on segmented tumor regions, and the combination of deep learning features with hand-crafted radiomic features and patient age information. Feature selection is performed using the Least Absolute Shrinkage and Selection Operator (LASSO) regression model and the DeepSurv is utilized for survival prediction. The proposed method was evaluated on the BraTS Benchmark 2020 training data sets using nested five-fold cross-validation. The resulting C-index values for the training, validation, and testing sets were 0.984, 0.821, and 0.821, respectively, outperforming the random survival forest method. Our findings suggest that the proposed method has the potential to serve as an imaging biomarker for predicting overall survival in GBM patients, with superior transferability compared to traditional machine learning-based methods.
 
Keywords—survival prediction, deep learning, neural networks, Random Survival Forest (RSF), Magnetic Resonance Imaging (MRI) image


Copyright © 2023 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.