Innovative CNN Approaches for Brain Tumor Detection: A Comparative Analysis of Model Architectures

This research presents a comparative analysis of three state-of-the-art convolutional neural networks (CNNs)—InceptionV3, ResNet50, and VGG19—for classifying brain tumors from MRI images. Utilizing a diverse dataset divided into four categories (glioma, meningioma, notumor, and pituitary), the study evaluates model performance based on precision, recall, F1-score, and overall accuracy. VGG19 emerged as the best performer, achieving an impressive accuracy of $\mathbf{9 6\%}$, excelling in the precise identification of glioma and notumor cases with minimal misclassification. ResNet50 also performed well with 96% accuracy, showing balanced results across most categories but slightly underperforming in meningioma classification. InceptionV3, with a lower accuracy of 93%, struggled with meningioma detection. These findings highlight the growing potential of deep learning in medical diagnostics, positioning VGG19 as the most reliable model for brain tumor classification. In subsequent study, these models can be improved by using hybrid architectures and datasets that are more diverse to achieve better clinical outcomes in the real world.