Deep Learning Approach for Radiogenomic Classification of Brain Tumor

Medical imaging of brain tumors is critical for tumor pathology characterization and early tumor detection. Current genetic analysis techniques take a long time and necessitate the surgical removal of samples of brain tissue. Multimodal brain tumor images can be accurately classified to speed up identification and lessen patient anxiety. In this research a novel deep-learning-based classification model is proposed that checks for the genetic subtype of glioblastoma to detect MGMT (O6-methylguanine-DNA methyltransferase) promoter methylation, the presence of which has demonstrated to be a positive prognostic factor and a powerful predictor of responsiveness to chemotherapy. It allows for less invasive diagnosis and treatments. A scenario-specific medical images dataset, the RSNA-MICCAI dataset, is used to test our model which comprises structural multi-parametric MRI (mpMRI) scans, in DICOM format. This work aims to investigate the feasibility of applying deep learning techniques for predicting the methylation state of the MGMT promoter.