AI-Based Brain MRI Cancer Classification Using Transformers and Quantum Self-Attention Techniques

Abstract

Brain tumor detection using Magnetic Resonance Imaging (MRI) is a critical task in modern healthcare, where early diagnosis significantly improves patient survival rates. Artificial Intelligence (AI), particularly deep learning, has revolutionized this domain by enabling automated and accurate classification of brain tumors. Traditional Convolutional Neural Networks (CNNs) have shown promising performance; however, they are limited in capturing long-range spatial dependencies due to localized receptive fields. Recent advancements have introduced Transformer-based architectures that leverage self-attention mechanisms to model global contextual relationships within MRI images. Vision Transformers (ViTs) process images as sequences of patches, enabling efficient global feature extraction. Additionally, Group Parallel Axial Attention has emerged as an efficient alternative to conventional attention mechanisms by reducing computational complexity while preserving both local and global dependencies. Furthermore, Quantum Self-Attention introduces a novel paradigm by incorporating quantum-inspired operations, enhancing feature representation and generalization capabilities. This paper presents a comprehensive review of AI techniques for brain MRI classification, focusing on recent trends and identifying key challenges. Comparative analysis reveals that hybrid CNN–Transformer and attention-based models achieve superior accuracy (often exceeding 98%), although challenges such as computational cost, data scarcity, and interpretability remain significant barriers.