CNN and Transformer based Brain Tumor Detection and Classification System

A brain tumor, also recognized as an intracranial tumor, develops as an anomalous mass of tissue wherein cells proliferate and multiply without restraint. Detection and classification of tumor in the brain is of paramount importance for effective treatment of the patient. CNNs and Vision Transformers are the popular techniques that have been used in the past for classification of images. Both the techniques have performed extraordinarily in the task of image classification. So here we have introduced a new approach where we fused VGG16 with the Vision Transformer (ViT) for brain tumor detection and classification. We have also compared the performance of our approach with other robust approaches such as ResNet50, VGG16, InceptionV3 and a pre-trained Vision Transformer that is GViT. Our approach has outperformed the other approaches during the comparison. Testing accuracies of pre-trained ResNet50, VGG16, InceptionV3, GViT are 90.18%, 90.04%, 74.96%, 72.38% respectively. The testing accuracy of our approach is 93.31%.