An Optimized Novel Hybrid CNN-Vision Transformer (ViT) Architecture for Brain Tumor Classification

Brain tumor is the uncontrolled division of mutated cells. If not detected and treated timely, it may ultimately cause death. According to data collected from different sources it is evident that, in past few decades brain tumor has badly affected a significant population worldwide and due to its expensive diagnostic and treatment strategies it has various associated socio-economic consequences. AI has played an important role in the medical imaging industry for the detection of various diseases. In this study a novel hybrid CNN-ViT architecture has been proposed for classification of Giloma, Pituitary and Meningioma tumors in a publicly available dataset ‘SARTAJ’. As compared to the conventional designed CNN architectures our proposed hybrid architecture achieved 98.1% accuracy. The proposed model has the potential to achieve the significant milestones in the medical imaging industry specially for detection of the brain tumors.