A Survey of Methods and Architectures for Brain MRI Image Classification for Cancer Detection Using Transformer and Group Parallel Axial Attention with Quantum Self-Attention

Abstract

Brain tumor classification using Magnetic Resonance Imaging (MRI) has become a critical application of artificial intelligence in healthcare. Traditional approaches based on Convolutional Neural Networks (CNNs) have achieved significant success in extracting spatial features; however, their limited ability to model global dependencies restricts performance in complex tumor classification tasks. Recent advancements have introduced Transformer-based architectures, axial attention mechanisms, and hybrid deep learning models that significantly enhance classification accuracy and efficiency. Transformers leverage self-attention mechanisms to capture long-range dependencies across MRI images, enabling improved feature representation and classification performance. Hybrid CNN-Transformer models further enhance results by combining local and global feature extraction. Additionally, axial attention reduces computational complexity by decomposing attention operations, making these models more efficient for high-resolution medical images. Emerging research on quantum self-attention introduces a novel paradigm, integrating quantum computing principles to improve computational efficiency and feature learning. This survey reviews recent literature from 2020 to 2024, analyzing key architectures, methodologies, and performance trends. Comparative analysis shows that hybrid and attention-based models achieve classification accuracy exceeding 99%. Despite these advancements, challenges such as data scarcity, computational cost, and interpretability remain. Future research directions include lightweight architectures, explainable AI, and quantum-enhanced models for clinical deployment.