Glioma is one of the most common primary tumors among adults. Glioblastoma multiforme (GBM) is the most aggressive form of glioma with very poor prognosis. The median patient survival is about 15 months. Treatment of glioma requires a complex approach combining surgical resection, chemotherapy and radiation therapy. Definition of the tumor border is the important step of radiation therapy treatment planning. The rapid development of the diagnostic methods made it possible to address this challenging task. However, the optimal treatment volume is still a matter of debate due to the complex biological behavior and high invasiveness of the tumor. Furthermore, the subjective interpretation of the visual information by the expert existed. So far, the application of machine learning in image analysis is a promising tool for glioma segmentation in multimodal MRI imaging. This review aims to summarize recent works using machine learning in high — and low — grade glioma segmentation.