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

Abstract

Brain tumor detection using Magnetic Resonance Imaging (MRI) is a critical task in medical diagnostics, enabling early intervention and improved patient outcomes. Traditional machine learning approaches rely heavily on handcrafted features and are limited in capturing complex spatial dependencies within MRI images. In recent years, deep learning techniques, particularly Convolutional Neural Networks (CNNs) and Transformer-based architectures, have significantly improved classification performance. Transformers leverage self-attention mechanisms to capture global contextual relationships, overcoming the limitations of CNNs in modeling long-range dependencies.

This review explores recent advancements in brain MRI classification using Transformer models, Group Parallel Axial Attention, and emerging Quantum Self-Attention mechanisms. Axial attention reduces computational complexity while preserving global context, making it suitable for high-resolution medical images. Furthermore, quantum-inspired attention mechanisms enhance feature representation by leveraging quantum state spaces, improving generalization and convergence efficiency.

A comprehensive analysis of literature is presented, highlighting performance improvements, architectural innovations, and key challenges. Comparative evaluation demonstrates that hybrid Transformer-CNN and attention-augmented models achieve superior classification accuracy, often exceeding 98%. The review concludes by identifying research gaps and future directions, emphasizing the need for explainable, efficient, and uncertainty-aware models for reliable clinical deployment.