One of the most deadly neurological disorders is brain tumors, and prompt and accurate diagnosis is necessary to plan appropriate treatment. Recent years have seen the development of algorithms for machine learning (ML) and deep learning (DL) that are helpful in automatically detecting and categorizing brain cancers, particularly using clinical imaging modalities like magnetic resonance imaging (MRI). A thorough comparison of ML- and DL-based techniques used for brain tumor segmentation is covered in this work. Sophisticated DL architectures like CNN, U-Net, and transfer learning models are contrasted with traditional ML methods like SVM, RF, and KNN. Additionally covered are key preprocessing strategies, feature selection strategies, and segmentation strategies. Additionally, available datasets like BraTS and TCIA are used for comparison using common assessment measures including accuracy, precision, recall, and F1-score. In addition to discussing new developments like explainable AI and ensemble deep learning models, the survey is used to evaluate the paradigms' advantages and disadvantages as well as their suitability for use in clinical settings. The lessons learned are meant to direct future research toward more reliable, understandable, andclinically acceptable brain tumor diagnosis methods.