Review of Brain Tumor MRI Image Segmentation Methods for BraTS Challenge Dataset

Abstract

Brain Tumor MRI segmentation is a crucial task in biomedical imaging. Early discovery of brain cancer can help with improving the quality of life and survivability posttreatment. In the case of children affected with brain tumors, early detection can determine what therapy would be required and early treatment in most cases will increase the longevity of their life. Brain Tumor segmentation has been done manually by an expert operator who is clinically trained. But this is very time-consuming and also the rating by these trained operators has intra-operator variability in most cases. Also, the different machines from different places have few variations in imaging which brings the rating variations. A need for Fully automated segmentation models is required for overcoming these issues. In recent years we have seen a trend of deep learning models being in heavy use in medical imaging tasks. These deep learning models have exhibited state-of-the-art performance by self-learning features. In this paper we focus on different machine learning, deep learning models being used on the Brain Tumor Segmentation (BraTS) challenge datasets. This paper tries to give the overall work-flow required for Brain Tumor MRI segmentation and gives a comparison on different models centered around deep learning as well as machine learning models. Lastly, an evaluation of the present state is shown and future improvements to standardize MRI-based brain tumor segmentation techniques.