A Systematic Review of Machine Learning and Deep Learning Techniques for Brain Tumor Detection and Classification in MRI Images (2020-2026)

Uncontrolled cell development in the brain causes brain tumors, which are serious neurological illnesses that frequently result in serious health issues like memory loss and motor dysfunction. Improving patient survival and treatment planning requires early and accurate diagnosis as well as precise tumor detection, classification, and segmentation from magnetic resonance imaging (MRI). However, sophisticated computer-aided diagnostic methods are required due to the growing workload of radiologists and the intricacy of MRI interpretation. The accuracy and dependability of automated brain tumor analysis have been greatly increased by recent developments in deep learning (DL), especially with multi-modal MRI, which supports better clinical decision-making. With an emphasis on recent contributions from 2021 onward, this study offers a comprehensive overview of DL-based research articles published between 2021 and December 2026. In addition to assessing cutting-edge DL designs like CNNs, U-Net variants, transfer learning, autoencoders, attention-based models, hybrid architectures, and transformer-based networks, the paper looks at popular benchmark datasets like BraTS, TCGA-GBM, and Figshare. Although CNNs and U-Net models are excellent in localizing tumors and extracting features, hybrid and transformer-based methods show better contextual modeling and generalizability. Data asymmetry, low contrast, annotation bias, limited interpretability, and clinical implementation continue to be problems in spite of these advancements. The results offer a thorough overview of current approaches and pinpoint important research avenues for reliable and practically applicable AI-based brain tumor analysis systems.