Brain Tumor Classification Using Deep Learning: A Comprehensive Literature Review

Brain Tumor classification using Magnetic Resonance Imaging (MRI) plays a critical role in early diagnosis, treatment planning, and patient survival improvement. However, manual interpretation of MRI scans is time-consuming, subjective, and prone to inter-observer variability. In recent years, deep learning—particularly Convolutional Neural Networks (CNNs) and transformer-based architectures—has revolutionized automated brain Tumor diagnosis by enabling end-to-end feature extraction and high-precision classification. This paper presents a comprehensive literature review of deep learning approaches for brain Tumor classification and segmentation. It systematically examines CNN-based architectures such as VGG, ResNet, Dense Net, Efficient Net, Inception, and Xception, along with advanced models including U-Net, 3D CNNs, attention mechanisms, Vision Transformers, and hybrid ensemble frameworks. Comparative analysis reveals that transfer learning and ensemble methods significantly enhance performance, with recent state-of-the-art models achieving classification accuracies exceeding 99% and Dice similarity coefficients above 0.90 for Tumor segmentation. The review also highlights key challenges, including limited annotated datasets, class imbalance, computational constraints, and the need for explainable AI in clinical settings. Emerging solutions such as federated learning, self-supervised learning, uncertainty quantification, and lightweight deployment models are discussed as promising future directions. Overall, deep learning has transformed brain Tumor diagnosis from subjective manual assessment to highly accurate, automated decision-support systems. Continued advancements in multimodal integration, interpretability, and real-world clinical deployment are expected to further enhance reliability, scalability, and patient outcomes in neuro-oncology diagnostics.