State-of-the-Art in Brain Tumor Segmentation and Current Challenges.

Brain tumors are the third most common type of cancer among young adults and an accurate diagnosis and treatment demands strict delineation of the tumor effected tissue. Brain tumor segmentation involves segmenting different tumor tissues, particularly, the enhancing tumor regions, non-enhancing tumor and necrotic regions, and edema. With increasing computational power and data sharing, computer vision algorithms, particularly deep learning approaches, have begun to dominate the field of medical image segmentation. Accurate tumor segmentation will help in surgery planning as well as monitor the progress in longitudinal studies enabling a better understanding of the factors effecting malignant growth. The objective of this paper is to provide an overview of the current state-of-the-art in brain tumor segmentation approaches, an idea of the available resources, and highlight the most promising research directions moving forward. We also intend to highlight the challenges that exist in this field, in particular towards the successful adoption of such methods to clinical practice.