A Systematic Review of Brain Tumor Classification from MRI Scans Through the Last Decade

Brain tumors develop as a result of abnormal cell growth in the brain or adjacent tissue, and if not treated early on, they can be fatal. In this context, neuroimaging is critical for disease detection. The magnetic resonance imaging (MRI) scan of an individual’s brain comprises multiple slices spanning the three-dimensional anatomical perspective, and because of its non-intrusive and non-ionizing modality, it eliminates the need for a biopsy and makes the procedure safer. However, this mainstream technique of manually classifying brain tumors is a labor undertaken by clinicians and radiologists, further complicated by the variations in tumor shape, size, and location. Image processing methods coupled with machine learning, deep learning, or transfer learning techniques can help clinicians accurately diagnose many forms of brain tumors and analyze microscopic changes. This systematic review intends to provide a complete overview of the domain, publicly available datasets, and has been conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines, and examines peer-reviewed studies from the last decade (2015–2024) on AI applications in brain tumor classification. The protocol for this review has been filed in the International Prospective Register of Systematic Reviews (PROSPERO), and can be identified by the registration number CRD420251047174. It encompasses studies using machine learning and deep learning methods on MRI sequences in this domain, and headlines their strengths, limitations, and the key obstacles faced by researchers such as dataset variability, integration of multiple modalities, generalizability and explainability of AI-driven frameworks. By accentuating these issues, this review offers expedient trajectories for deployable, ethical and transparent research in classifying brain tumors using end-to-end AI techniques.