Glioblastoma Survival Prediction

Glioblastoma is a high-grade invasive astrocytoma tumor. The highly invasive nature makes timely detection and characterization of the tumor critical for the survivability prediction of patients. This work proposes MRI- and clinical information-based automated pipeline that implements various state-of-the-art image processing, machine learning, and deep learning techniques to obtain robust tumor segmentation and patient survival estimation. We use 163 cases from the training dataset, and 28 cases from the validation dataset provided by the BraTS 2018 challenge for the evaluation of our model. We achieve an accuracy of 0.679 using the validation dataset and that of 0.519 for the test dataset.