TransMedNet: A Hybrid Transformer-CNN Framework for Accurate Brain Tumor Classification from MRI Scans

Abstract

The diagnosis of brain tumors through MRI analysis presents significant challenges due to the heterogeneous nature of tumor morphology and subtle differences across tumor types. To address these complexities, we introduce TransMedNet, a hybrid deep learning model that synergizes convolutional and transformer-based architectures for enhanced brain tumor classification. By combining the spatial precision of CNNs with the global context awareness of Vision Transformers, the model effectively captures both localized abnormalities and long-range dependencies within MRI scans. A dual attention mechanism further refines feature selection, while multiscale processing supports representation of tumor patterns at different spatial scales. Experimental evaluation on the public MRI dataset shows that TransMedNet provides competitive classification performance across the reported accuracy, AUC, and F1-score measures. Because the present study is based on a single public dataset and does not include independent external clinical validation, the results should be interpreted as evidence of technical feasibility within the evaluated experimental setting rather than proof of clinical deployment readiness