Brain Tumor Segmentation and survival prediction using 3D U-net

Abstract

Abstract Detecting Brain Tumor is a critical problem in the medical field and analyzing its location and size has been a challenging task for doctors. Designing automated brain tumor systems for MRI is one of the important factor in this decade, to avoid high expenses, low accuracy due to parallax errors and time consumption. Automated systems assist in diagnosis and early detection of tumors, which allows physicians to focus more on treating a patient. A method based on Convolutional Neural Network (encoder decoder architecture) captures 3D Flair (MRI) for glioblastomas as an input, which could be anywhere, any shape or size of the brain. These factors made us to investigate for an artificial intelligence system that utilizes the flexibility of a highly efficient neural network with an astonishing efficiency. The U-Net model is essential for achieving optimal performance in segmenting brain tumor and patient stage. The current approach uses 3d image instead of 2d image for proper diagnosis.