The human brain is an incredible and wonderful organ that governs all body actions. Due to its great importance, any defect in the shape of its regions should be reported quickly to reduce the death rate. The abnormal region segmentation helps to plan and monitor the treatment. The most critical procedure is isolating normal and abnormal tissues from each other. So far, remarkable imaging modalities are being used to diagnose abnormalities at their early stages, and magnetic resonance imaging (MRI) is renowned and noninvasive among those modalities. This paper investigates the current landscape of brain tumor segmentation (BTS) by exploring emerging deep learning (DL) methods for brain MRI analysis. The findings offer a comprehensive comparison of recent DL approaches, emphasizing their effectiveness in handling diverse tumor types while addressing limitations associated with data scarcity and robust validation. DL has shown a vital improvement for BTS, so our primary focus is to include significant DL robust models to analyze the brain MRI. However, DL outperforms traditional methods; still, there are several limitations, especially related to the diverse tumor types, lack of datasets, and weak validations. The future perspectives of DL-based BTS present significant potential for revolutionizing the diagnosis and treatment of brain tumors.