Work

Survival Prediction in Glioblastoma Using Combination of Deep Learning and Hand-Crafted Radiomic Features in MRI Images

Year: 2022

Type: article

Abstract: Background and objectives: Glioblastoma (GBM) is the most common malignant primary tumor of the brain. Survival prediction allows for risk stratification impacting all aspects of care. Previous attempts at predicting survival outcomes have relied on clinical parameters that suffer from uneven capture in available data sets impairing transferable survival precision. We propose a method for overall survival prediction that employs brain MRI imaging to classify patients based on survival by employing a combination of deep learning and hand-crafted radiomic features.Methods: (1) 3D brain tumor segmentation based on the nnU-Net model; (2) Patient classification into long survivors (>900 days), short survivors (<300 days), and mid-survivors (300- 900 days) using the Dense-Net model on segmented brain tumor regions; (3) Combination of deep learning features with hand-crafted radiomic features and patient age information followed by feature selection using the least absolute shrinkage and selection operator (LASSO) regression model and DeepSurv for survival prediction.Results: The proposed method was evaluated on BraTS Benchmark 2020 training data sets using nested five-fold cross-validation. Data was divided into 64% (training), 16% (validation), 20% (testing) sets. The resulting CI of training, validation, and testing sets (0.984, 0.821, and 0.821 respectively), carried superior performance vs. random survival forest (0.817, 0.782, and 0.785, respectively).Conclusions: We present a novel survival prediction approach that combines deep learning and hand-crafted radiomic features, with the potential to generate an imaging biomarker for the prediction of the overall survival in patients with GBM with superior transferability as compared to the traditional machine learning-based methods.

Source: Journal of Advances in Information Technology

DOI: 10.12720/jait.14.6.1461-1469