Classification of Brain Tumours in MRI Images Using VGG 19 Algorithm

Abstract

For early identification and efficient treatment planning, brain tumour categorization is crucial. Deep learning in medical image analysis, models have lately shown remarkable outcomes, especially regarding MRI scan tumour categorization. This study proposes VGG-19 as the primary model and evaluates its performance against ResNet-50 and Google Net by comparing F1-score, AUC, precision, recall, and accuracy. Consequently, the suggested VGG-19 model had a maximum classification accuracy of 97.2%, outperforming GoogleNet at 96.4% and ResNet-50 at 95.3%. It proves from the confusion matrix that the classification of all kinds of tumours is superior, and minimal misclassifications are available with VGG-19. It is proven further by the ROC curve since AUC scored 0.97, giving excellent discrimination capability. GoogleNet was found to be competitive, especially in Glioma detection, while ResNet-50 provided a balanced classification in all classes. This outcome suggests that deep learning models, particularly VGG-19, efficiently classify brain tumour. Further studies can investigate transformer-based architectures, hybrid deep learning techniques, and the expansion of datasets to further enhance model generalization. Integration of explainable AI into medical imaging systems further improves transparency and clinical trust in using AI-assisted diagnosis in healthcare.