Brain Tumor Classification Based on Machine Learning and Deep Learning Technologies: A Comprehensive Review

Brain tumor classification represents a critical challenge in medical imaging analysis, where accurate and timely diagnosis significantly impacts patient outcomes and treatment planning. This comprehensive review examines machine learning and deep learning technologies applied to brain tumor classification from MRI scans, systematically comparing the evolutionary trajectory from traditional machine learning (SVM/RF/k-NN with handcrafted features) through convolutional neural networks to vision transformers within a unified analytical framework. We systematically analyze traditional machine learning approaches including support vector machines, random forests, and k -nearest neighbors, alongside advanced deep learning architectures such as convolutional neural networks, recurrent neural networks, and transformer-based models. The paper explores preprocessing methodologies, feature extraction techniques, and classification strategies while addressing fundamental challenges including dataset limitations, class imbalance, model interpretability, and clinical deployment barriers. We demonstrate practical deployment feasibility through model compression techniques, achieving 46 × reduction (244 MB to 5.3 MB) via knowledge distillation while maintaining 0.68 s inference time for 3D MRI scans on edge devices. Through comparative analysis of methodologies and performance metrics, we identify research trends and future directions. While deep learning models demonstrate superior performance with accuracy exceeding 95% in controlled settings, significant challenges remain in generalization, interpretability, and clinical integration. This review provides researchers with structured understanding of the field's landscape and identifies opportunities for advancing automated brain tumor classification toward clinical viability.