A Systematic Review of Machine Learning and Deep Learning Approaches for Brain Tumour Analysis Using Magnetic Resonance Imaging

Analyzing brain tumours from magnetic resonance imaging data is inherently challenging because tumour characteristics such as shape, size, location, and visual appearance vary considerably among patients to overcome the constraints of manual image interpretation, researchers have increasingly investigated automated computational methods grounded in machine learning and deep learning. This systematic review examines peer-reviewed journal-articles published between 2015 and 2026 that focus on MRI-based brain tumour detection, segmentation, and classification. Both traditional ML methods relying on handcrafted feature representations and modern DL approaches, including convolutional-neural networks, transfer learning frameworks, and region-based detection models, are critically analysed. The reviewed studies are compared in terms of methodological design, dataset utilization, and evaluation strategies. Furthermore, emerging performance trends, computational demands, and practical deployment challenges are discussed. Finally, unresolved research gaps and prospective directions are identified from a computer engineering perspective, with emphasis on model explainability, data efficiency, and scalable system implementation.