Brain Tumor Detection and Classification Using Hybrid CNN–GNN Framework with Multi-Modal MRI: A Review

Brain tumor detection and classification using magnetic resonance imaging (MRI) is a critical task in neuro-oncology due to the complex structure and heterogeneous nature of tumors. Early diagnosis significantly improves treatment outcomes. Deep learning approaches, particularly convolutional neural networks (CNNs), have shown promising performance in automated brain tumor detection and classification tasks [10], [18]. Several CNN architectures such as ResNet, EfficientNet and YOLO-based models have demonstrated high detection accuracy on MRI datasets [1], [13], [16]. However, CNN models mainly focus on local spatial feature extraction and fail to capture long-range contextual dependencies between tumor regions [12], [17]. Recent studies have explored vision transformers, attention mechanisms, and graph neural networks (GNNs) to improve contextual modeling [3], [6]. Additionally, multi-modal MRI analysis (T1, T2, and FLAIR) improves diagnostic performance by leveraging complementary anatomical information [4], [7].